What's new

3 November 2026 - Explore your Spaces

Explore this Space opens a location image as an interactive camera view. Navigate with keyboard or touch controls, direct changes to the scene, and save discovered views as Space areas for the storyboard. Review the maximum price before starting each one-minute session. Saved views persist; new explorations start from a chosen image.

2 November 2026 - Replace a timeline clip and find its Storyboard shot

Right-click a timeline clip to Reveal in Storyboard or Replace with selected asset. Storyboard video and dialogue controls also offer Replace selected clip beside Send. Replacement preserves the existing timing, trim, effects and other editing attributes, including the original shot link. Attached video sound follows the replacement; detached audio stays unchanged. Shorter media keeps the clip duration, and Undo restores the whole replacement.

2 November 2026 - FLUX 3 video resolutions and new video templates

Sandbox FLUX 3 now offers 2K and 4K alongside 720p and 1080p, including image and keyframe workflows. Edit video changes a short clip from written instructions. Upscale video offers Precise and Creative enhancement, predicted output dimensions and source/result comparison. Project rendering now uses Topaz Proteus automatically for 4K upscaling, with no engine menu. Saved results reopen with their source and settings in Templates.

2 November 2026 - Animate a talking image with OmniHuman V2

Sandbox Video now offers OmniHuman V2. Add a person image and 4–300 seconds of WAV or MP3 speech, with an optional performance prompt for expression, movements and camera direction. The audio determines the video length. Saved results retain both inputs for Run again. Project Improve lipsync continues to synchronize existing video with its original motion.

2 November 2026 - Place dialogue audio wherever you need it

The downward Send to selected audio track button places the available recording at the playhead. Reuse it as often as needed, including when the video already contains dialogue or has not been placed on the timeline. The existing recording remains usable after script changes. Placement overwrites audio in the destination range, without shifting later clips or generating audio again.

2 November 2026 - Return to your workspace

Switching editor panels keeps your place, including in Storyboard. Refresh or reopen a project in the same browser to return to the last panel, internal tab, selected item and scroll position. Searches, filters and expanded inspector sections are remembered per project. Playback stays paused, and workspace positions remain local to this browser.

2 November 2026 - Sound Effects v3 throughout

New ElevenLabs text sound effects use v3 in the editor and Sandbox. The v2 choice and fallback are retired. If v3 is temporarily unavailable, new generation reports that its connection or verified price is unavailable. Existing results and submitted jobs remain recoverable.

2 November 2026 - Original recovery after repeated lip sync

Improve lipsync now continues through older saved history when a newer entry contains an already lip-synced take. It restores the exact untouched input, including after an older automatic result received another manual pass. Existing takes are preserved.

2 November 2026 - Completion notices wait for the finished result

Projects generation notices now wait through video lip sync, dialogue assembly, sound-effect cleanup and placement, and image finishing. Intermediate requests no longer announce that the whole action is complete. Multipart audio waits for its batch; regenerated takes notify again when ready. Finishing failures preserve completed paid stages for recovery.

1 November 2026 - Recover original videos for older lip-sync takes

Improve lipsync can recover an older automatic take's exact original video from saved project history. It verifies the source and updates the estimate before proceeding, keeping the old take and saving the recovered original with a successful new result. Missing sources and project changes stop the request before generation.

1 November 2026 - Automatic highest-quality lip sync

Storyboard Improve lipsync now automatically uses the highest-quality model, with a live price estimate and no quality menu. Automatic dialogue post-sync uses the same model. Older lip-sync takes retain their original source video. Optional automatic routing preserves the same model and caps successful lip-sync charges at the approved maximum; failed attempts are absorbed by cutipAI. Routing remains off until verified and enabled.

Timeline previews also prepare more reliably on computers without graphics acceleration. Blur keeps the same appearance when the editor needs to use software rendering.

31 October 2026 - Preview and edit storyboard frames

First, Middle and Last frame previews correctly open cached images. Edit on the frame or in its preview opens the image editor used for character sheets. Saving creates a new take for that frame, preserving its original and editable layers across project reopening. Empty reference columns are hidden, and an unavailable image shows a Retry action.

30 October 2026 - Faster, incremental timeline previews

Space starts within cached footage immediately. Picture and audio wait together at an uncached section, then resume automatically; Space cancels the wait. Adding or adjusting a transition rebuilds only the ten-second sections that overlap it. Generated footage decodes from local cached files, with faster preview encoding and progress for the section being prepared.

30 October 2026 - Smooth persistent timeline previews

Play prepares rendered 720p preview sections with effects, transitions, titles and captions as needed during playback. Finished sections survive panel changes and reopening; edits rebuild only dependent sections. Preferences shows preview storage, project/all clearing and Resume caching. Clearing stays paused across restart. The app never automatically expires these previews; browser site-data clearing and a declined persistent-storage grant still limit retention.

30 October 2026 - Timeline sound generation

Selected clips initialize an editable total duration, and prompts start empty. Scan the video analyzes and generates Music or SFX in one click. Generated sound follows the chosen target track, including overlapping clips; new tracks are created only when chosen. SFX receive automatic human-vocal removal before placement, with both stages included in estimates. Long requests split into parts, progress and stopping are available, and unfinished jobs retain their original receipts for recovery. All-result players and same-clip regeneration remain available.

30 October 2026 — FLUX.3 Klein in the Sandbox

Image adds FLUX.3 Klein and Klein Base for description-based generation and editing with up to ten ordered references. Choose five square sizes through 4K, one to four images, output format and prompt expansion. The visual Layout editor places new objects or keeps, moves, resizes and removes reference objects. Saved recipes restore the full composition.

29 October 2026 - Timeline controls and generation recovery

Press Space after an editor adjustment to return to timeline playback. Pausing on the timeline stops picture and audio together. Waveforms fill resized audio clips without a split. Track focus changes only when you click its name. Projects remember each timeline's playhead, zoom and scroll position, and Storyboard add/duplicate controls sit at the top after every shot, including the last. Regenerating dialogue audio stops and hides the previous preview while the replacement is pending.

Session interruptions reconnect to the existing submitted generation. Recover result in History or the Job Queue checks a finished request and adds its media to the original project's Assets, without generating again or replacing clips. Check History before starting another generation.

29 October 2026 - Preview and regenerate timeline audio

Listen to generated music and sound effects in the generation window, including individual scan cues and sound passes. Regenerate a result to replace the same timeline clip; earlier takes stay in Assets. The previous take stays available during generation. Closing stops and clears the temporary players, while the audio stays on the timeline. See Generating audio.

29 October 2026 - Choose which videos to scan for audio

Music and SFX both offer Scan the video. Choose Selected videos only, or The whole working area to use In/Out marks (the full active timeline when unmarked). Check the video count and time span before scanning. Music proposes editable cues; SFX reviews synchronized sound passes for the included video portions. Results use new audio tracks. See Generating audio.

28 October 2026 - Automatic grading from a reference

The Colour tab can analyse a clip against an image or a chosen frame from a video. Load a reference from your computer, Project Assets or the Generated Library; review the proposed grade and its strength before applying editable colour settings. Copy those settings across the active timeline without another analysis charge, or analyse each clip separately against the same reference. See Automatic grading.

28 October 2026 - Import from empty shot previews

Click an empty First, Middle or Last frame preview to import an image, or the empty Motion preview to upload a motion clip. This also works while Middle or Last is Skipped. Existing preview controls and empty-state text stay the same. See The storyboard.

28 October 2026 - Generation continues without a price confirmation

Clicking Generate starts the job without a second price dialog. New generations settle at the provider's final billed usage plus the service fee, including billed failed or canceled work. Closing a window or a polling timeout leaves submitted work running; History recovers its existing job. Timeline audio now uses ElevenLabs Music v2.5 and Sound Effects v3 through the direct connection, with a priced alternate connection when needed. The audio window's dismissal button is Close. See Generating in the editor.

27 October 2026 - Dialogue videos finish when lip sync costs more

When a finished paid video needs a higher lip-sync estimate than the one shown before generation, cutipAI continues automatically if your balance covers it. The final charge follows the measured result plus cutipAI's fee. If your balance is too low, the video stays saved so you can add funds and use Retry lip sync without generating it again. Generation prices shown in the app are estimates and may finish higher or lower. See The storyboard.

27 October 2026 - Voice library for everyone

Character, object, narrator and Sandbox voice choosers now offer public library voices to all signed-in users. Search, filter and save favorites, then click Use voice to assign one. Assigned voice shows the chosen name on one line. Voice descriptions expand on demand, and each sample plays without a generation. New custom voice creation is retired; previously assigned voices and completed audio remain usable. When the voice connection lacks permission to save library voices, the selected card now says so and lets you try again after the account owner updates its key.

27 October 2026 - More video models for storyboard dialogue

Storyboard dialogue now works with video models that have no audio-file input, including Kling 4.0. Generate the dialogue first; Generate video then makes the picture and syncs it to the saved voice in a separate pass. Its estimate includes both jobs. If lip sync fails, retry it from the saved video without paying to generate the picture again. See The storyboard.

27 October 2026 - Storyboard frame priority

New shots start with Middle and Last set to Skipped. Adding an image to either slot turns it on automatically. During video generation, enabled first, middle and last images take priority over conflicting description text. If the selected model or Motion clip cannot use an enabled frame, the shot explains the conflict before generation. See The storyboard.

27 October 2026 - Storyboard dialogue lip sync

Dialogue videos now show the video and lip-sync estimates together, then automatically run a separate lip-sync pass against the generated voice. The finished video keeps the synchronized soundtrack. If the second pass fails, Retry lip sync resumes from the saved paid video without regenerating it; previously submitted work is recovered. Improve lipsync is also repaired for existing shots. See The storyboard. The lip-sync quote now uses the server's per-second price, and the charged amount follows the measured video result.

26 October 2026 - Owner-only voice creation

The expanded voice library and custom voice design are now available only to the protected owner account. Existing voice engines remain available to other accounts. Previously accepted voice jobs and saved assignments remain readable for recovery. When custom voice slots are full, the app refuses a new design quote before any paid preview starts.

26 October 2026 - Smoother timeline playback

The editor now loads upcoming video and audio into the browser's media cache. A thin green line under the time ruler shows sections ready to play. When a clip stalls, playback holds the playhead and resumes when the media is ready. Moving a video or audio clip to another track leaves its linked partner on its current track. See Editing on the timeline.

Transitions and visual effects are now rendered into short preview segments while the timeline is paused. Once prepared, playback uses the finished picture instead of recalculating the effect on every frame. The green line waits for those segments, and editing a clip refreshes its preview. Finished previews now keep title sizing consistent and avoid a visible flash at segment joins. Cached media and previews survive Storyboard visits and project reopen for at least three days after the last visit. Browser site-data clearing or storage pressure can still remove local data.

26 October 2026 - Timed Storyboard dialogue

Storyboard dialogue uses your speakers' saved ElevenLabs voices. Add silence before a line or insert exact pauses between phrases. Changing pause durations reuses saved speech takes; preview and regenerate individual lines, and extend or split a shot when its dialogue is too long. Delivery remains available, while the advanced Consistency setting replaces Intensity. Dialogue-driven video keeps the original voice soundtrack, with a free retry for local soundtrack assembly after a completed video generation. See The storyboard.

25 October 2026 - Sandbox voice library and audio routing

Sandbox's Eleven speech models now offer Voice library & design in Output settings. Describe a voice, generate candidates, browse filtered ready-made voices, keep favorites and reuse saved voices. Tweak and Run again retain the chosen identity and seed.

Supported Sandbox audio uses the lower verified cost among compatible connections. Projects use the dedicated audio connection for speech, music and sound effects. Confirm the estimate before generation and check History for the final usage charge. See Sandbox.

25 October 2026 - More ways to choose a voice

The new shared voice chooser puts a description box above a searchable library, with language, accent, gender, age and use-case filters. Design voice options, keep favorites, save voices privately across projects, and preview your own dialogue. Only Use voice changes a speaker's assignment. The expanded library requires cutipAI's voice connection; Existing voice engines remain available while it is being connected. See Characters.

25 October 2026 - Import areas into Spaces

Add your own areas to a Space using From computer or From Assets above its image list. Select multiple images and use them immediately in storyboard shots, with Reference enabled by default. Your canonical image stays unchanged. See Spaces.

25 October 2026 - Kling 4.0 video

Make 3 to 30-second videos with Kling 4.0 Text, Image and Reference in the Sandbox, or edit a clip with Kling 4.0 Edit. Projects now offer one Kling 4.0 video choice; earlier rendered videos keep their original recipes. Anime, Claymation, Watercolour and Match a picture in Restyle a video accept up to 30 seconds. Other looks keep their shorter part limit.

24 October 2026 - A chair for Objects

Objects on the editor's rail now has a chair icon, so it no longer looks like the 3D Builder. The same chair marks objects in the Objects panel and in @mentions. See Objects.

24 October 2026 - 3D Builder back, and the whole image in Shoot an area

The 3D Builder is back on the rail and in the storyboard, after recent releases were built without it by mistake. In the storyboard, Shoot an area now shows the whole Space image, so every part of it can be framed. See 3D Builder.

24 October 2026 - A tidier LUT menu

Import .cube… is now the last choice in each LUT menu, and the list of LUTs at the foot of the grade is gone. A project still keeps up to eight LUTs: importing a ninth makes room by dropping the oldest one no clip uses. See Colour.

24 October 2026 - LUTs in the Colour panel

Grade with LUTs. Input LUT, at the top of Basic, converts camera footage before any other correction, and the new Look section applies a ready-made look with an Intensity you can keyframe. Import .cube files into the project with Import .cube…; they travel with the project. See Colour.

24 October 2026 - Grading refinements

In the Colour panel, Reset curve clears just the curve you are looking at, and a clip on a locked track can't be graded until you unlock it. The clip inspector shows Colour grade · off for a grade that is switched off. The colour wheels are read out as the colour they push toward. See Colour.

24 October 2026 - Grading in the Colour panel

Select a clip and grade it under the scopes: Basic correction (white balance, exposure, contrast, tones, saturation and vibrance), Curves including Hue vs Saturation, Shadows, Midtones and Highlights colour wheels, and a Vignette. Every setting can be keyframed, each section switches off for a quick comparison, and the export matches the preview. See Colour.

23 October 2026 - A quieter Colour panel

When the playhead is over a gap, the Colour panel's scope now simply stays black instead of showing a message. The waveform and parade also stay bright and readable for a vertical clip, whose black bars used to dim the trace of the picture. See Colour.

22 October 2026 - Colour scopes

A new Colour panel on the rail measures the frame you are looking at, effects included, with a Waveform, an RGB Parade, a Vectorscope and a Histogram. Use it to spot lost highlights, a colour cast or skin tones that are off, and to match two shots. See Colour.

22 October 2026 — 3D Builder tutorial

The studio's Tutorial button now opens cutipAI Documentation. The expanded guide covers navigation, staging, poses, motion paths, camera moves, sequence exports and a table-conversation exercise. See 3D Builder.

21 October 2026 — Project 3D Builder

Build reusable 3D sets inside projects, with character poses, imported GLB models, cameras, animation and reference exports. The scene assistant previews editable proposals before Apply. Send blocking straight to a compatible video model, or review a finished image first. References link to storyboard shots, and generation results stay in Assets until you apply them. Generated-model previews also offer a larger interactive inspection view. See 3D Builder.

21 October 2026 - Mirelo SFX 2.0 in the Sandbox

The Sandbox adds Mirelo SFX 2.0 for sound from text or video, and video with a replaced soundtrack. Keep dialogue, exclude unwanted sounds, choose audio format and generate one to four samples together. Select each sample in the preview to play, download or send it to a project.

21 October 2026 - Faster caption editing

Caption words now update in the preview while you type, with one undo step per editing session. The caption and transcript actions use compact icon buttons with tooltips. Choose an outline or box colour beside Background in the inspector; the Look selector has been removed. See Captions.

20 October 2026 - Generation recovery controls

Supported models with automatic recovery enabled can try an equivalent alternative after a technical failure. The price is an estimate; the final charge includes provider-billed usage for each attempt and the service fee. Portrait registration is no longer offered in Sandbox or the character inspector. See Generating in the editor.

20 October 2026 - Faster effects

Picture effects are now drawn by your graphics card, so clips with Sharpen or Hue on one colour play smoothly and export faster, with the same look as before. Two things change: effects now work from the top of the clip's list down, as audio effects do, and Film grain is the same in the preview and in every export. See Titles, effects and transitions.

20 October 2026 - Help when a payment fails

Buy credits now has a Contact us link under the Pay … now button. If your card keeps being declined, it opens the contact form with Payment problem already chosen, and we will help you find a way to pay. See Buying credits.

20 October 2026 - Denoise audio from more places

Denoise audio is now on a video clip's right-click menu, where it cleans the clip's sound, and in the Audio section of the clip inspector. The dialog and its price work as before. See Denoise audio.

19 October 2026 - Ducking

The Mixer can now lower the music under speech. Mark your tracks Dialogue and Music under Ducking, choose an Amount and click Duck music: the music dips wherever someone speaks, as volume points you can see and drag. See Mixing sound.

19 October 2026 - Loudness

The Mixer now measures loudness the way video platforms and broadcasters do: Momentary, Short-term and Integrated in LUFS, and True peak. Measure measures the whole timeline without playing it. Choose a Target, Social (−14 LUFS) or Broadcast (−23 LUFS), and the export can bring the file to that level, with a new Loudness choice in the export dialog. See Mixing sound.

19 October 2026 - Audio effects

The Effects panel has a new Audio category: High-pass, Low-pass, Parametric EQ, Compressor, Limiter and De-esser. Drag one onto an audio clip, set it under Audio effects in the clip inspector, keyframe any setting, and reorder them. They are free, you hear them while you edit, and the export sounds the same. See Mixing sound.

19 October 2026 - Mixer refinements

An audio clip's yellow volume line now uses the faders' scale, so a level above 0 dB shows where it is and keeps its boost when you move the point. Volume points move with the arrow keys and stay between their neighbours. Double-clicking a fader or Pan resets it in one step of undo, grabbing a fader's handle no longer makes it jump, and every fader in the Mixer lines up, Master included. See Mixing sound.

18 October 2026 - The Mixer

A new Mixer panel on the rail gives each audio track a fader, pan, Mute and Solo, and a level meter, with a Master fader for the whole timeline. Audio tracks gain a solo button, clips gain Gain in the inspector, and you can add volume points on an audio clip's yellow line with Ctrl-click. Your export now follows volume and pan keyframes exactly as you hear them while editing. Auto-captions, Denoise audio and Insert scene still hear every clip's own sound: the Mixer's faders and Solo never change them. See Mixing sound.

17 October 2026 - Edit with the transcript

Text → Transcript shows every word Auto-captions heard. Click a word to jump there; select words and press Delete to cut that part of the film from every track. Remove filler words and Shorten pauses… tidy speech in one go. It is free: it uses the words Auto-captions already found. See Captions.

17 October 2026 - Auto-captions leaves offline clips out

Auto-captions no longer counts a clip whose media is offline: it is not heard, and its length is not charged. The dialog says how many clips are left out, and asks you to relink them in Assets when all of the sound is offline. See Captions.

16 October 2026 - Auto-captions

Text → Captions → Auto-captions turns the timeline's speech into timed captions, in the language you choose or the one it detects. You see the estimated cost before you click. See Captions.

16 October 2026 - Titles and captions share the Text panel

The Edit and Captions tabs have left the top bar. Titles on the rail is now Text, with a Titles tab and a Captions tab, and the caption style sits in the inspector beside the list. Clicking a caption on the timeline opens the list at that caption; with the Storyboard open, double-click it instead. Another clip selected while Captions is open shows its own settings in the inspector. See Captions.

15 October 2026 - Captions, and a Captions workspace

The editor gains captions. Type them at the playhead, import an SRT or WebVTT file, or make them from your storyboard's dialogue for free. They sit on a C1 row at the top of the timeline and move with your edit. Choose one look for the whole project, then draw the captions into the export with Burn in captions or save them as a file. Two tabs next to the menus, Edit and Captions, switch the panels beside the preview; the preview and the timeline stay in place. A caption over a gap in the picture is now drawn too.

15 October 2026 - Account storage and portrait library

Open Storage or Portrait library from the account menu. Storage brings allowance, purchases and stored-file removal together outside the editor. Portrait references must be active before use; the library displays their availability and verification status.

14 October 2026 - Rodin, Hitem3D and Tripo 2.5

More textured GLB presets cover Rodin 2.5 Text/Image/Fast Image, Hitem3D 3.0 Image/Multi-View and Tripo 2.5 Image/Multi-View. Add object pictures, inspect the original mesh and restore saved views with Tweak. Multi-view uploads retain each model’s view order. Explore the presets.

14 October 2026 - More simple 3D generation

Create textured GLBs with Tripo H3.1 and Hunyuan3D V3 from text or one picture. Tripo Multi-View also accepts two to four ordered object views. Inspect the rotating original, download the GLB or restore writing and references with Tweak. The 3D viewer in Sandbox and History keeps a steady rotation on slower devices and respects reduced motion. Explore the 3D presets.

14 October 2026 - Animate pictures with LTX and Hailuo

Video adds seven first-frame presets across LTX 2.3 and Hailuo 02/2.3. Add one picture, describe its movement and keep the original MP4. Tweak restores the source and writing; saved previews show actual dimensions. Explore picture animation.

14 October 2026 - More native Video presets

Video adds LTX 2.3 22B, Distilled and Quality presets with audio, plus silent Hailuo 02 and 2.3 Standard and Pro presets. Write a description, preview the original MP4, download it or restore the saved recipe with Tweak. Saved previews show actual dimensions; History shows final supplier charges. Explore the Video presets.

14 October 2026 - Transparent images and more BRIA presets

Create an isolated object with Ideogram V3 Transparent and download its original transparent PNG. Image also adds BRIA Fibo/Lite, BRIA Base/Fast/HD and Wan 2.2 A14B Image. BRIA HD keeps native 1536-pixel JPEG output; saved descriptions restore with Tweak. Explore the presets.

14 October 2026 - Edit to music beats

Right-click a timeline audio clip and choose Analyze audio to detect beats locally. Beat guides follow the clip, and Shift snaps the playhead to markers. Clip moves and trims also snap to markers when snapping is on. Beat settings offers estimated BPM, half/double tempo, offset correction, reanalysis and clearing, with undo and project saving. See Analyze audio.

13 October 2026 - Browse community work by model

Choose Model in the Community gallery to see approved creations made with one model. The selection works with other filters and stays active across pages.

13 October 2026 - MAI, Luma Uni and more image engines

Image adds MAI 2.5 and Pro, Phota, Nucleus and BRIA Fibo Gen 1.5 at square 1K, plus Luma Uni 1 and Max at square 2K. Keep original PNG downloads and restore saved writing with Tweak. Explore the presets.

13 October 2026 - Six more square image models

Image adds Gemini 3.1 Flash, Grok Imagine Quality, GPT Image 1 High, Kling Image 3 Text and O3 Text, and MiniMax Image 01. Keep original 1K PNG or JPEG downloads and restore writing with Tweak. Explore the image models.

13 October 2026 - Luma Photon images

Image adds Luma Photon and Photon Flash with square 1536-pixel original JPEG downloads. Restore saved writing with Tweak and see final supplier charges in History. Explore the presets.

13 October 2026 - Ideogram variants and Kontext Max

Image adds Ideogram V2a Turbo, Ideogram V2 Turbo, Ideogram V2a and FLUX Kontext Max Text with square 1K original PNG downloads. Restore saved writing and seed with Tweak, and check final charges in History. Explore the image presets.

13 October 2026 - Find and report community work

The Community gallery gains search across titles, prompts and tags, creator and image/video filters, and newest, oldest or featured sorting. The homepage shows the twelve latest approved shares. Add an optional title and tags when sharing an existing result, then discover related work from its tags. Open an asset and choose Report asset to send a private concern for moderator review.

13 October 2026 - Eight more image presets

Image adds original Nano Banana and Grok models, Nano Banana Lite, GPT Image 1 Mini, Ideogram V2, Krea 2 Medium and FLUX Kontext Pro Text at square 1K, plus FLUX Pro 1.1 Ultra at square 2K. Download original PNG images and restore saved writing with Tweak. Explore the presets.

13 October 2026 - Five more square image workflows

Image adds Nano Banana 2, Nano Banana 2.1, FLUX 3 Image, GPT Image 1.5 and Gemini 2.5 Flash Image with simple square 1K PNG presets. Download original images and restore writing with Tweak. Token-priced workflows show estimates and final supplier charges in History. Explore the image models.

13 October 2026 - Influencer signup codes

Creators can share a personal signup link or code. New accounts created with email or Google can be linked to that creator for their lifetime. Eligible generation spending supports the creator through a commission from our fee; your generation prices stay the same. See Influencer codes.

13 October 2026 - Short dialogue

Voice adds Dia Short Dialogue for English conversations up to 250 characters. Insert two speaker markers in Output settings, keep the original WAV and restore the dialogue with Tweak. The model chooses the voices. Write a conversation.

13 October 2026 - VibeVoice narration

Voice adds six English presets with VibeVoice 0.5B Speech and a 5,000-character limit. Play or download original MP3 speech and restore saved writing and voice with Tweak. Final charges follow actual billed audio. Explore the voices.

13 October 2026 — Delete from Assets

The Assets panel has a bin beside New folder and New timeline. Remove one or several assets, or a selected folder or timeline, after confirmation. Warnings identify affected timeline clips; you can undo the change and original files remain untouched. The bin also activates with Enter or Space when focused.

13 October 2026 - Short narration voices

Voice adds Chatterbox HD Short Speech with nine presets and a 500-character limit. Download original WAV audio or restore saved writing and voice through Tweak. Learn about short narration.

13 October 2026 - Seed Speech voices

The Sandbox adds Seed Speech 2 with native voices labelled by language. Save the original MP3 and restore the text and preset with Tweak. Estimates count the exact spoken characters; final charges follow actual supplier bills.

13 October 2026 - Remove background

The Remove background template cuts out an image automatically, from instructions, or by removing a selected color. Refine transparency with erase and restore brushes, compare before and after, choose a solid background, and download an original-size PNG. Save finished variations separately to your Library; their transparency masks reopen for more editing. Local color removal and refinements have no AI charge. Interrupted AI jobs can be checked without resubmitting.

13 October 2026 - Guided background removal

The Sandbox adds BRIA Guided Background Removal. Upload a picture and optionally describe which subjects to keep. Download the original transparent PNG or restore the picture and instructions with Tweak.

13 October 2026 - More native speech choices

The Sandbox adds Qwen 3 Preset Speech 0.6B, Async Speech Pro and Qwen Audio 3 Flash Speech. Choose native presets, keep original MP3 or WAV files and restore saved speech with Tweak. Unicode characters count correctly at the writing limit. Verified preset speech workflows allow more queue time, and canceled results retain actual supplier billing semantics.

13 October 2026 - ElevenLabs, xAI and Qwen preset speech

The Sandbox adds ElevenLabs Multilingual v2 and Turbo v2.5, xAI Text to Speech and Qwen 3 Preset Speech. Choose native voices, keep the original MP3 and restore saved settings with Tweak. Qwen estimates include both spoken text and delivery instructions in whole thousand-character blocks. Voice previews show Format and omit unsupported seed controls. Final charges follow actual supplier bills.

13 October 2026 - More speech voices

The Sandbox adds six MiniMax Speech models: 02, 2.5 and 2.6, each in HD and Turbo. Choose a voice and delivery, keep the original MP3, and restore your saved settings through Tweak. Final charges follow actual supplier bills.

13 October 2026 - More song and music workflows

The Sandbox adds MiniMax Music 2, Music 1.5 and Reference Song, ACE Step Music and Sonilo 1.1. Each exposes its supported lyrics, style, source or duration controls. Original MP3, WAV and M4A downloads and saved Tweak settings preserve your work. Sonilo estimates include its ten-second minimum; final charges follow actual bills.

13 October 2026 - Edit and extend source music

The Sandbox adds music region editing and extension. Preview an MP3 or WAV, choose the part to replace or time to add at either end, and describe the result. Original WAV downloads and saved source settings remain available when you return. Final charges follow actual supplier bills.

13 October 2026 - Music variations from your audio

The Sandbox adds four Stable Audio music variation models. Upload and preview an MP3 or WAV, describe the change, and choose an output length. Original WAV downloads and saved source recipes remain available when you return. Final charges follow actual supplier bills.

12 October 2026 - Stable Audio Open

The Sandbox adds Stable Audio Open music with duration choices up to 47 seconds, original WAV downloads and saved Tweak settings. Estimates use average generation costs; final charges follow actual compute billing.

12 October 2026 - DiffRhythm songs

The Sandbox adds DiffRhythm for songs from timed lyrics, with 95 or 285 second choices, optional musical style, original WAV downloads and saved Tweak settings.

12 October 2026 - ElevenLabs Music v2.5

The Sandbox adds ElevenLabs Music v2.5 for instrumental music with native duration choices from three seconds to ten minutes. Original MP3 downloads and saved Tweak settings preserve the requested duration. Estimates account for whole-minute supplier billing; final charges follow actual bills.

12 October 2026 - Mureka music and lyrics

The Sandbox adds Mureka 9.5 Instrumental and Song in Music, plus Mureka Lyrics in Text. Songs accept your lyrics or write them from a description, with mode-specific estimates. Original audio, saved writing and Tweak recipes remain available when you return. Final charges follow actual supplier billing.

12 October 2026 - Native Lyria music

The Sandbox adds native Lyria 2 for instrumental WAV tracks at about thirty seconds. Original downloads and saved recipes are preserved, and final charges follow the actual supplier bill.

12 October 2026 - More native music choices

The Sandbox adds MiniMax Music 2.5/2.6 for instrumental tracks or lyrics, and ElevenLabs Music v2 for instrumental music with a requested duration. Original MP3 downloads and saved recipes are preserved. Estimates include the extra MP3 frame before whole-minute billing.

12 October 2026 - Selectable voice presets

The Sandbox adds Gemini 3.1 Flash TTS and MiniMax Speech 2.8 Turbo. Choose a voice preset and delivery in settings, download the original WAV or MP3, and restore both choices with Tweak. Estimates count spoken prompt characters; final charges follow actual billing.

12 October 2026 - Four-picture editing

The Sandbox FLUX 2 workflow accepts up to four reference pictures. Tweak restores every saved picture, and estimates include the input work for each variation. Removing a picture updates the estimate; switching to a one-picture model removes extra references with a notice.

12 October 2026 - Transparent picture cutouts

The Sandbox adds five background-removal choices from BRIA, Ideogram, Pixelcut and BiRefNet. Add one picture without a description, then download the original transparent PNG. A checkerboard makes transparency visible; Tweak restores the source. Each request creates one cutout, with final charges recorded in History.

12 October 2026 - Full-source editing and picture analysis

The Sandbox adds sixteen picture-editing workflows across FLUX 2 and native FLUX Dev and Kontext Dev. Editing preserves the full source picture. Source-size estimates appear after you attach it; final charges follow the supplier's actual bill. Saved results retain the source, original output and recipe.

Picture Analysis adds twelve caption, OCR and question workflows across Moondream, Florence, Isaac, LLaVA and Nemotron. Caption and OCR models need only the picture. Download the saved answer as text, or use Tweak to restore the source and question.

11 October 2026 - Material maps from a picture

The Sandbox adds PATINA Image Maps. Add a picture without a description, then download five labelled original maps. Large sources are fitted without cropping, and saved recipes restore the prepared picture.

11 October 2026 - FLUX Krea LoRA Stream

The Sandbox adds a verified square Image model. Its final original JPEG, returned seed and recipe remain available in saved results and History. Downloads preserve the original file.

11 October 2026 - Seamless material files

The Sandbox adds PATINA Material and Material Extract. Make a texture from a description or source picture, then download its original texture and five labelled maps individually. Saved recipes restore the source and description. Submitted failures preserve the final-billing warning.

11 October 2026 - Keys that follow Premiere Pro

The timeline gains Premiere Pro's navigation and marking keys: jump to the previous or next cut, go to the start or end, go to the In or Out mark, fit the timeline, select the clip under the playhead, and mark a clip. Three existing keys changed to match: Shift with the arrow keys now moves five frames instead of ten, Ripple edit is on B instead of Shift+R, and the Inspector panel is on Ctrl+Alt+9 instead of Shift+I. Keys you set yourself are kept. See Coming from Premiere Pro.

11 October 2026 - Use SVG results in videos, 3D and projects

Saved SVG results now offer Animate this, Make it 3D and Send to project. These actions prepare a PNG copy, with a clear preparation state and retryable errors. Video and 3D still require Generate after reviewing the estimate. Download preserves the original SVG.

11 October 2026 - Original 3D downloads

The Sandbox and History now save original 3D files without navigating away from the preview. OBJ results have separate download buttons for supplied materials and textures.

11 October 2026 - Three-point editing

The timeline gains Insert and Overwrite from a marked source range, Lift and Extract over the timeline's In and Out marks, a target on every track that says which tracks an edit may use, and a Linked selection toggle. The video preview in Assets has a new Overwrite button and remembers your marks.

11 October 2026 - Images from a reference style

In the Sandbox, Generate becomes ready as soon as your selected model’s estimate and required inputs are ready. Other model prices can keep loading in the menu.

The Sandbox adds source-picture editing with Qwen Image Edit, Z Image Turbo Edit and FLUX Kontext, plus SDXL ControlNet Union for outline-guided images. Add one picture and describe your change. Saved recipes restore the source, and downloads preserve the original file.

The Sandbox adds Recraft V4 Style and Pro Style in raster and SVG versions. Attach one reference picture and describe your image. Quotes include style creation, Tweak restores the source picture, and Download saves the original file.

11 October 2026 - Scalable SVG images

The Sandbox adds four verified Recraft SVG models: V4, V4.1 and their Pro versions. Preview your vector picture, restore its description with Tweak, or download the original SVG while keeping your results open. Format shows SVG instead of a raster resolution. Vector results do not offer actions that currently require raster pictures.

11 October 2026 - More verified square Image presets

The Sandbox adds 31 more verified Image choices, including FLUX General, Recraft 20B, Lumina, Sana, Pony, CogView, OmniGen, Janus and built-in style presets. Janus shows its actual 384 × 384 output size. Download now saves the original image, video or audio while keeping your results open.

The Sandbox adds verified raster presets across FLUX, Krea, Qwen, Z Image, Wan, Cosmos, Juggernaut and other Image families. Each exposes its tested square format and supported seed controls. Display names are shorter, favourites keep the same IDs, and compute-priced jobs finalize from actual supplier bills.

10 October 2026 - More Image models and accurate billing recovery

The Sandbox adds 22 verified square-image choices, including GPT Image 2 API, Flare, Seedream 5, Recraft V4 and V4.1, Ideogram V4 Fast and Instant, and Wan 2.7. Format shows the verified 1K or 2K size. Models without seeds restore saved descriptions and settings without inventing one. Variable jobs finalize from supplier bills; refresh can correct an older reserved estimate without running another generation. The model picker stops fetching list prices when it closes.

10 October 2026 - Fixed image sizes made clearer

The Sandbox adds Seedream 4 and 4.5 at fixed 2K square output, plus ten fixed 1K choices including Ideogram V4.5, Qwen Image, FLUX 2 Flex and Hunyuan Image 3. Format displays the output size separately from How many. Saved recipes preserve returned seeds without replacing them with unknown values. Metered jobs show estimates and finalize from actual bills.

10 October 2026 - More square Image choices

The Sandbox adds FLUX 2 Max, Flash, Turbo, Klein 4B and Klein 9B, Krea 2 Turbo and Ideogram V3. The existing FLUX 2 Pro entry retains its picture-editing controls. Each generates a square image from a description. Saved results restore the exact model and description with Tweak. Metered estimates are finalized from actual billed work.

10 October 2026 - More Sandbox Image and Video models

The Sandbox adds SDXL, SDXL Lightning, three HiDream variants, Kolors, AuraFlow and FLUX.1 Schnell for square images from a description. Hunyuan Video 1.5 adds silent video from a description or a first-frame picture. Preview, download or Tweak saved results; final charges follow actual billed work.

10 October 2026 - Cosmos text-to-video

The Sandbox adds Cosmos Predict 2.5 Text and Distilled Text. Describe a scene to make a silent MP4 with fixed dimensions and length, then preview, download or Tweak it.

10 October 2026 - Trim tools

Roll, Slip and Slide tools, frame nudges for a selected edit point, ripple trim and extend to the playhead, and ripple delete. See Trimming. The Export panel moved to Shift+E and the keyframe lane to Shift+U. Every trimming key works the same on a Mac.

10 October 2026 - Relight cover artwork

The Relight template now has a split-light portrait cover, making it easier to find in Templates. See Relight.

10 October 2026 - More picture-to-3D models

The Sandbox adds Pixal3D from one picture and ReconViaGen 0.5 from one to four pictures of the same object. Both return textured GLB with fixed settings. Preview, download or Tweak a saved result to restore its sources.

10 October 2026 - Video relighting

Relight joins Templates and video right-click menus in Assets and the timeline. Arrange lights around a sphere or upload a reference, review the estimate, then compare the result before keeping it. Timeline Apply preserves edit length and original audio; Assets keeps a new video. See Relight a video.

10 October 2026 - CogVideoX and multi-image Trellis

The Sandbox adds CogVideoX-5B Text and Image for silent video in a fixed 3:2 shape, with length chosen by the model. Trellis Multi-image builds a textured GLB from two to four pictures of the same object. Saved results restore their model, description and pictures through Tweak.

10 October 2026 - Frame-exact editing

Moves, trims, splits and pastes now land on whole frames, and the playhead stops on a frame. See Frame-exact editing.

10 October 2026 - Three-view 3D generation

Hunyuan3D v2 Multi-View and Multi-View Turbo join the Sandbox. Add Front, Back and Left pictures in that order, then choose geometry-only or textured GLB. The composer labels each view and waits for all three. Saved results restore the pictures and surface through Tweak.

10 October 2026 - Hunyuan3D Mini and Turbo textures

Mini and Turbo now offer Textured and Geometry only GLB in Sandbox Output settings. Textured output costs three times the geometry price. Tweak restores the selected surface; earlier geometry-only results keep their original settings.

10 October 2026 - Pro video and textured 3D

The Sandbox adds LTX 2.3 Pro Text and Pro Image for 1080p video with generated audio and six, eight or ten second choices. Hunyuan3D v2 Base and Mini-Turbo now offer Textured or Geometry only GLB output. The estimate reflects the surface choice; saved results restore it through Tweak.

9 October 2026 - Video with audio and more 3D models

The Sandbox adds LTX 2.3 Fast Text and Fast Image for 1080p video with generated audio. Choose landscape or portrait and a supported length from six to twenty seconds. Hunyuan3D v2 Base and Mini-Turbo build geometry-only GLBs from one picture. Saved results support preview, download and Tweak, and final charges follow actual billing.

9 October 2026 - Two more Video workflows

The Sandbox adds Wan 2.2 A14B Turbo Text and Turbo Image for silent 480p video. Describe a scene, or attach a first-frame picture for the Image model. Choose a shape; the model chooses the length. Saved videos support playback, download and Tweak, and final charges follow actual billing.

9 October 2026 - Complete model lists

The Sandbox model menu now loads the entire category at once, sorted alphabetically. Scroll or search the list; your Favourites stay at the top.

9 October 2026 - Resizable timeline tracks

Video and audio tracks now sit against the divider, the section titles always stay in view, and the timeline resizes from the line under its tools bar, down to two video and two audio tracks. Each track can be resized on its own, or set to Small, Medium or Large from its header. See Resizing tracks.

9 October 2026 - Mirelo sound effects

The Sandbox adds Mirelo SFX 1.6 under Sound FX. Describe a sound and choose a duration from one to sixty seconds. Its estimate includes the extra second billed for slightly longer output, and final charges follow actual billing. Saved WAV results support playback, download and Tweak.

9 October 2026 - Two more Music models

The Sandbox adds Stable Audio Small Music Base and Medium Music Base for instrumental WAV cues. Choose a duration in Output settings, then play, download or tweak your saved result. Final charges follow supplier billing.

9 October 2026 - Import AVI and Photoshop files

You can now import AVI videos and PSD Photoshop files. cutipAI makes a viewable copy for the editor and keeps your original file unchanged.

8 October 2026 - Four more Voice languages

The Sandbox adds Kokoro Italian, Brazilian Portuguese, Japanese and Mandarin Chinese, each with a matching preset voice. Saved audio and Tweak preserve the original model and text; final charges follow supplier billing.

8 October 2026 - More reliable model estimates

Opening the Sandbox model list now loads estimates more reliably when several models need a price at once. If pricing is temporarily unavailable, use the retry button to refresh the estimate without generating again.

8 October 2026 - More verified Video and 3D models

The Sandbox adds Wan 2.2 A14B Image for silent 720p animation from a first frame, and Hunyuan3D Mini and Turbo for geometry-only GLB from a picture. Saved recipes restore their sources. Wan A14B Text and Image now settle against final supplier bills.

8 October 2026 - Stalled jobs end sooner

A reframe that stalls is now stopped after about ten minutes plus a quarter of a minute per second of clip, instead of an hour and a half, and you are not charged for it. See Templates. And when a generation fails, the amount set aside for it now returns to your balance at the next check, instead of up to a day later.

8 October 2026 - Rapid geometry from a picture

The Sandbox adds Hunyuan 3D 3.1 Rapid Image for untextured GLB geometry from one required picture. Picture size is checked before submission, and Tweak restores the saved source.

8 October 2026 - Templates by category

Templates are now grouped into Image, Video and Audio. Use the switch at the top to show one category, or All to see every template under its category heading. The page address keeps your choice.

8 October 2026 - Verified editing and geometry workflows

The Sandbox adds Kling O3 Pro Edit for changing an uploaded video and Hunyuan 3D 3.1 Rapid for geometry-only GLB objects from text. Kling final charges follow actual billed usage, including fractional seconds. Rapid picture and textured generation remain pending.

8 October 2026 - OBJ geometry previews

Sandbox and History can show OBJ geometry with separate downloads for supplied material and texture files. The viewer states that materials are not shown and supplied maps may be incomplete.

8 October 2026 - More verified Video and 3D choices

The Sandbox adds Hunyuan Video Image for first-frame animation and Meshy 7.1 Multi-image for a textured object from two to four views. Required inputs are checked before Generate is available. Results can be previewed and downloaded; final charges follow the work billed, including texture generation.

7 October 2026 - Easier Sandbox browsing

The Sandbox composer stays at the bottom of the window, and account pages no longer show the website footer. Results have 100-generation pages with navigation above and below, full-history search, date/model sorting and model/date/status filters across media, Text and Analysis. Category buttons are smaller, the prompt starts empty, and image Format and How many are separate controls. Search has more space above the results.

7 October 2026 - More Video and 3D workflows

The Sandbox adds Wan 2.2 5B Text and Wan 2.2 5B Image for silent 720p video, plus Trellis and Trellis 2 for textured GLB files from one picture. Required pictures are clearly labelled, and picture-only 3D workflows hide the unused prompt and surface choices.

7 October 2026 - Results follow your category

Sandbox results now follow the selected category, including preview navigation. Image shows images, Voice shows voice results, and changing category starts a fresh search. Grid and list buttons sit to the left of Search on one toolbar row.

7 October 2026 - More image models

The Sandbox adds Recraft V2, Recraft V3 and FLUX Pro 1.1 for making raster images from a description at 1K. Choose a shape and image count; save favourites to find these models again.

Sandbox model favourites

Browse each category in a single searchable model list. Star any model to keep it in Favourites at the top, including models awaiting workflow verification. Favourites are saved with your account and return when you sign in again. Show more models reveals more of the catalog without replacing the models already on screen.

7 October 2026 - More voice and sound models

The Sandbox adds Kokoro (British English) with the Alice voice and Stable Audio 3 SFX Base for sound effects up to two minutes. Saved audio can be played, downloaded, or sent to a project.

6 October 2026 - Video Analysis

The Sandbox can answer questions about an MP4 video. Preview a clip up to two minutes at 1080p, ask for a summary or details, then copy or download the saved answer. Final charges follow actual usage.

6 October 2026 - Clear pending charges

Completed Text and Analysis results show Final charge pending while billing finishes. Refresh results checks the final bill immediately without generating again. Your saved output remains available to copy and download while you wait.

6 October 2026 - Audio Analysis

The Sandbox can transcribe speech or answer questions about a recording. Upload MP3 or WAV, preview it, then generate saved text that you can copy, download, or revisit. Transcription detects the language automatically.

6 October 2026 — Text and image Analysis

The Sandbox can write text and answer questions about an uploaded image with Qwen 3.8. Choose a response length, copy or download saved results, and restore a prompt for another generation. Charges follow actual token usage. Text also offers Nemotron 3 Nano Omni and Seed 2.0 Mini through the model menu. Saved recipes retain their model, and tiny charges remain visibly nonzero.

6 October 2026 — Image, video and phone improvements

The Sandbox adds FLUX Dev, Sana, LTX Video 0.9.5 and Hunyuan Video. These video engines choose their own length and show Model default in Settings. Phone results use two readable columns, while your saved desktop density is preserved. Mobile Settings stays within the screen, and measured duration labels stay compact.

6 October 2026 — More Sandbox models

The full Sandbox model menu adds FLUX Schnell, Wan 2.2, Stable Audio Small Music, Cassette Sound Effects, Kokoro and Tripo P2. The catalog refresh now separates lightweight discovery from generation verification. The saved-asset @ shortcut has been removed; upload references with +.

6 October 2026 — Sandbox redesign

Sandbox opens after sign-in, with results above a docked prompt composer. Upload references with + and choose essential controls from Settings. Search models and collapse the account sidebar. Catalog categories include models awaiting generation verification.

6 October 2026

• Recast your video is available in Templates and the timeline clip menu, with ordered photos, shot checks, before/after playback and undoable range replacement.

5 October 2026

• Video duration choices now follow each model in Projects and Sandbox, including one-second clips where supported.

• Looping video turns one image into a 5–15 second seamless loop with subtle motion or a full 360° spin. Choose quality and optional audio in Templates.

• Clean audio removes background noise from speech in uploaded audio or video. Compare the original and cleaned recording in Templates, or right-click an audio clip and choose Denoise audio for one undoable replacement.

• Insert scene adds 5–13 seconds of generated footage at the playhead, using both sides of the entire preview as context. Set the point and duration with handles or exact seconds, review both seams, then insert with one undoable ripple edit. Interrupted requests can be checked without generating again. See Generating in the editor.

• Eleven v4 is now the Project and Sandbox Voice default. Saved Project v3 preset voices use v4 for future dialogue, preserving their selected voice and existing audio. Eleven v4 Turbo is also available in the full Sandbox model menu. Dialogue estimates include each line and its delivery tags. See The Sandbox.

4 October 2026

• Seedance 2.0 US 4K joins the Sandbox video models and the model list in 4K Projects. Reducing a project below 4K switches its editable selections to Seedance 2.5. Generation requires verified 4K pricing. See The Sandbox.

• Sandbox Voice now offers Gemini 3.8 Flash TTS and Gemini 3.8 Flash Lite TTS with the Kore voice, Delivery controls and WAV audio downloads. See The Sandbox.

• Extend your video adds footage at the beginning or end from an upload, saved Sandbox video or generated project shot. End extensions use H3 Max and beginning extensions use LTX 2.3, automatically. The timeline's Generative extend uses the same routing. See Extend your video.

3 October 2026

• Storyboard Motion follows one clip's camera movement, action and pacing while keeping your references for the look. Choose project footage, a generated shot or an upload, then move the window to the part you want. Available on H3 Max, Seedance 2.5, Wan 3.0 and Wan 3.0 Prime. See Motion reference.
• Storyboard Middle frame guides a moment inside an H3 Max or Flux 3 video. Generate or import the picture and set its time; unavailable models and qualities keep your work and offer Skip. Middle frames have their own Assets folder and can be chosen as template source images. Its tab sits between First frame and Last frame. See Middle frame.
• An exported film now sounds exactly like the preview. A video placed with Include audio turned off no longer brings its own sound into the export, a video placed with its audio is no longer twice as loud, and a timeline placed inside another keeps its sound. See Exporting your film.
• Exported films now show every frame on time. Videos were often drawn one frame late, and an export could open on a black frame. See Exporting your film.

2 October 2026

• Restyle a video has sharper Anime, Claymation and Watercolour looks. Match a picture gives a clip the visual style of a picture you choose. See Templates.

1 October 2026

• Sandbox 3D now offers Meshy 7.1 in its full model menu: build from a description, one photo, or up to four views, choose Textured or Geometry only, and optionally get a Rigged character. See The Sandbox.

• Sandbox Music now offers Lyria 3.5 in the Model menu. Describe the whole song in the prompt, including any lyrics. See The Sandbox.

• Talking image animates a face image with uploaded or saved speech audio. Choose 480p, 768p, 1080p or 2K, with recordings from 5 seconds to 15 minutes.

• Talking image now has a real generated thumbnail, a silent tile preview, and a full spoken example with reusable settings.

30 September 2026

• Sandbox Voice now offers Grok Voice: upload a recording, choose a voice and receive an audio reply with a transcript. Instructions are optional, and the final charge follows metered usage. See The Sandbox.

• The timeline's scrollbar stays at the bottom of the editor. Resize the timeline and drag its video/audio divider to allocate track space; each project remembers your layout. See Editing on the timeline.

• Unfinished checkouts now get one reminder after a week, with no follow-ups, and close automatically after 14 days. See Buying credits.

• Double-click a video in Assets to mark an In and Out, then Insert selection at the timeline playhead. Existing clips ripple right instead of being overwritten, Include audio can be turned off, and the preview closes after a successful insert. See Organising assets.

29 September 2026

• Projects now expose only the Video model choice. New projects start at Full video quality; image generation uses GPT Image 2.5, and voice, music and sound effects use ElevenLabs.
• The timeline adds keyboard shortcuts for Add text (T), Generate video (G), and Generative extend (Shift+G). The generation keys only select their tools.
• The timeline's Add text tool now places static text, returns to Selection after one placement, and lets the title clip extend beyond its five-second starting length. See Editing on the timeline.

28 September 2026

• Moving several selected timeline clips vertically now keeps their relative track layout instead of merging them onto one track.
• Removed the obsolete stored-footage backup warning from Preferences and upload notifications. The plan summary now separates purchased space and price, the free 3 GB, and total storage.
• Storage purchase controls are now always visible in Preferences → Storage, and purchased space can be any whole number from 1 GB to 1 TB instead of fixed 10 GB steps.
• MP4 and MOV footage can now be stored when the browser reports an empty or nonstandard media type for the file. This commonly affected videos imported on Windows while images stored normally.
• Assets can now be stored individually or as a selection from their right-click menu. A failed upload shows a retry action and can be kept on the current device instead of appearing to wait forever. Source footage is stored on cutipAI's VPS.
• In regular Chrome 153 windows, Enable saved links restores existing file links and remembers your choice across refreshes and browser restarts. If Chrome asks for file access, choose Allow on every visit. See Bringing media in.

27 September 2026

• Imported and relinked files keep their live browser access when you return through the dashboard. Manual relinking now offers Display only files with the exact name, enabled by default, with folder locations to distinguish duplicate names. See Bringing media in.

• Imported MOV files are recognised as video even when the computer reports an image type. Reopening a project also corrects MOV assets previously labelled as images.

• Reframe works again. Reframing an exported film had stopped before it started, saying the price could not be worked out; nothing was charged. You are now charged for your film's exact length.

• Reframe has a new 540p resolution, a lower-cost choice for a film of up to 10 seconds, with 21:9, 4:3 and 3:4 among its ratios. See Exporting your film.

• A reframed video now also lands in Assets, in a Reframe folder, ready to drag onto a timeline.

• Reframe a video in Templates: turn any clip into several shapes at once — vertical, square, feed or cinema — and compare each with the original. See Templates.

• The Reframe a video card now shows three before-and-after examples.

26 September 2026

• A clip longer than the limit is no longer refused. In Restyle a video, and in the Sandbox's A video to edit and Sound FX → A video, a trimmer lets you choose the part to use — up to 14.5 seconds for an edit, 59.5 for sound — and only that part is priced and sent. See Templates and The Sandbox.
• Restyle a video: bring a clip, pick a look, and generate a restyled version. Choose a clip under Upload, Sandbox or From a project — MOV and WebM clips are converted for you — then pick a look such as Anime, Claymation or Film noir, or write your own under What should change?. Try another look keeps the clip so you can run it again. See Templates.
• The Restyle a video card now shows three before-and-after examples.
• Edit a video in a project: choose Edit video on a video in Assets or on a timeline clip, or Edit on a shot's video, then choose the part, pick a look and compare before you keep it. See Editing a video in a project.

25 September 2026

• The Sandbox can now score a video with sound: under Sound FX, set Make sound for to A video, add a clip under 60 seconds — MOV and WebM are converted to MP4 in your browser first — and click Generate. The result is priced by the clip's own length. It offers no Run again, Tweak or Variations; use Add sound on the video it was made from to score it again. A description of the sound is optional. In the editor, the timeline's "sound for the video" scan now shows each clip's Estimated cost in the job queue. See The Sandbox.

24 September 2026

• The Sandbox can now edit a video: under Video, set Start from to A video to edit, upload an MP4, MOV or WebM under 15 seconds, and describe the change under What should change?. MOV and WebM are converted to MP4 in your browser first. The result is a new 720p video, charged for its own length rather than your clip's, and your original clip is kept. Every video result also gets an Edit this video button that opens this mode with that clip loaded. See The Sandbox.
• The image editor gets Remove background: select a picture layer and cut its subject out from its background. The result becomes a mask on the layer rather than a permanent edit, and a free Mask section appears beside Opacity and Blend to edit it — Invert keeps the background instead, Feather softens the cut edge, and Remove mask brings the background back — Undo removes the mask but does not refund the removal. If it could not be applied, choose Apply it again — it is already paid for — or Discard; Save waits until you choose. It shows its Estimated cost on hover, keyboard focus or a touchscreen hold, and Not enough credit shows if your balance is short. See Characters.
• The image editor gets a Text tool and a Shape tool — click to place text and set its size, weight, font and colour, or drag to draw a rectangle or an ellipse with a fill and an outline — plus six filters under Add adjustment: Blur, Sharpen, Noise, Vignette, Black & white and Sepia. Expand canvas… adds free transparent space around a picture on any side, Fill empty areas continues the picture into that space, and Upscale… enlarges the whole picture to 4K (2160 pixels high). Fill empty areas and Upscale show their Estimated cost on hover, keyboard focus or a touchscreen hold. See Characters.
• The image editor gets selection tools — Rectangle, Ellipse and Lasso, with Feather, Select all, Deselect, Invert and Mask selected layer — and AI edits: describe a change and click Generate to redraw the selected area, or Remove to erase it and fill in the background. Both show their Estimated cost on hover, keyboard focus or a touchscreen hold. The rest of the image editor's tools stay free. See Characters.
• The image editor gets a Clone stamp that copies from the original picture or from everything visible, Transform to move, scale and rotate any picture or paint layer, adjustment layers — Exposure, Contrast, Saturation, Temperature and Curves, each able to clip to the layer below — and sharp painting when you zoom in. See Characters.

23 September 2026

• In the image editor, saving with the original picture hidden now asks first, because the hidden picture is left out and what no other layer covers comes out transparent. Choose Save anyway or Keep editing. See Characters.
• The image editor is much faster on pictures with many paint layers: painting starts at once, sliders keep up while you drag, and saving uses far less memory.
• New section, AI creation: Characters, Objects, Spaces, The storyboard, Generating in the editor, The Sandbox, Templates and Community gallery. Every part of cutipAI now has a page.
• In the Sandbox, Run again and Variations now show their Estimated cost when you hover over them, like Generate does. The estimate for making several images at once was too high and is now correct. See The Sandbox.
• The image editor works from the keyboard: it takes focus when it opens, Tab stays inside it, and closing it returns you to the Edit button. A picture that fails to open now offers Try again, and each layer says what it is in plain words. See Characters.
• The image editor gets its first tools: a Brush with pen pressure, an Eraser that removes only paint you added, a Colour picker, a working Layers panel, Undo/Redo, sharp zoom and keyboard shortcuts. Save makes the edited picture the sheet, and Revert brings the previous one back with its layers. See Characters.

21 September 2026

The documentation opens. Find it under Help → Documentation in the editor, or from the Documentation link at the foot of every page.

• New page: Your first project.
• New pages: Managing projects, How credits work, Buying credits, History and Account settings.
• Prices are off the screen. A paid button no longer shows its cost beside it — hover over the button, or press and hold it on a touchscreen, and its Estimated cost appears. Not enough credit still shows without asking. See How credits work.
• New section, The editor: The editor at a glance, Bringing media in, Organising assets, Editing on the timeline, The clip inspector, Titles, effects and transitions, Exporting your film and Preferences.

More Sandbox voice languages

Sandbox Voice now includes Kokoro Hindi, French and Spanish with matching presets, saved results and favourites. Final charges follow supplier billing.