The storyboard

The storyboard is where a film is planned before it is cut. Each scene is a row of shots, and each shot is a card holding its frames, its dialogue and its video. You generate all of them on the card, and send the finished video to the timeline when you are happy with it.

A generation completion notice waits for the finished result: video includes automatic lip sync, dialogue includes its assembled soundtrack, and frames are recorded before their notice appears.

Open the storyboard

Click Storyboard in the rail on the left of the editor. The board takes the place of the preview; click the rail button again to go back.

Switching to another panel and back keeps your place on the board. Refreshing or reopening the project in the same browser also restores the last panel and position. The board remembers the visible shot so loading images or resizing the window keeps you near the same part of your work. An explicit View frame action takes you to that shot instead.

Scenes

At the top of the gap to the right of each shot, use Add shot after or Duplicate shot. These controls are also available after the final shot and in scenes containing only one shot. Hover over the gap or focus its buttons with the keyboard; on touch screens the buttons stay visible.

1. Open the storyboard. On an empty board, click New scene.

   

   

2. Set the scene's heading: interior or exterior, the time of day, and a Location — one of your project's spaces (see Spaces).

Click + after a scene to add another one after it. Drag a scene by its handle to reorder it, or use Duplicate scene and Delete scene on its heading. Collapse all timelines in the toolbar folds every scene to its heading.

Kling 4.0 in Projects

Kling 4.0 is the single Kling video choice in Projects. Choose a whole-second duration from 3 to 30 seconds and 720p or 1080p. A first frame starts the animation; an optional last frame controls its ending. Tagged characters or objects use their approved reference pictures instead. With characters and a last frame together, follow Skip the last frame to continue; the saved last frame stays in the shot.

Older Kling shot selections move to Kling 4.0 when you open the project. Existing generated videos and their original recipes stay intact. Kling no longer offers 4K generation in Projects; use another 4K model or upscale the finished video. There is no separate sound switch for Kling 4.0. Check the Estimated cost before generating. If the shot has generated dialogue, Kling makes the video first and a separate lip-sync pass puts that dialogue into the finished video. The estimate includes both jobs.

Shots

Click Add shot at the end of a scene's row. Each card starts with the shot's number; drag the number to move the shot, or click Delete shot to remove it.

Frames

A shot can have a first frame — the picture it opens on — a last frame it should end on, and, on a supporting video model, a middle frame that guides a moment inside the shot. The tabs at the top of the card run First frame → Middle frame → Last frame → Motion. New shots start with Middle frame and Last frame set to Skipped. Generating, importing or taking an image from another shot turns that frame on automatically.

1. In First frame description, write what the shot shows. Type @ to tag a character, object or space, so it is drawn from your reference.

2. Click Generate. Once a frame exists the button reads Regenerate.

You can also bring a frame in instead of generating it: Import first frame from computer, or Use frame from shot to take a frame from a shot video you have already made. Clicking an empty First, Middle or Last frame preview opens that frame's import picker. Empty Middle and Last previews can be clicked while Skipped, just like their import buttons. In Motion, click the empty preview to upload a motion clip, just like Upload. Click an existing frame to view it at full size. Use Edit on a First, Middle or Last frame, or in its preview, to open the image editor used for character sheets. Save applies your edit only to that frame and keeps its layers for later editing, including after reopening the project. Cancel leaves the frame unchanged. The original image stays available. When no references are tagged, the preview uses the full width. If an image cannot load, use Retry. Editing needs a wider screen, as it does for character sheets. Existing clips keep their usual preview controls. Skip on the first frame excludes both the first and last frames from video generation; Skip on the last or middle frame excludes that frame only. An enabled middle image still guides the video when the first frame is skipped, if the selected model supports that combination. Skipping keeps the image, description and history, so choosing Frame brings it back. When an enabled first, middle or last image is present, video generation prioritizes its visible subjects, appearance and composition over conflicting description text. If a model cannot use an enabled image, the video panel asks you to choose a compatible model or skip that frame before generating.

The Background strip chooses which image of the scene's location the shot is set against. A Continuity chip means the shot carries the previous shot's look forward — its film stock and light, never its framing.

Middle frame

A middle frame is a still image placed at a time inside the generated video. The project choices that support it are H3 Max and Flux 3; the Middle frame tab says why it is unavailable on another model or quality.

1. Choose a supporting Video model in File → Project settings…, and open Middle frame on the card. H3 Max needs both an enabled first and last frame, with a native output of 480p or 768p. At a 1080p project resolution, Draft uses native 768p; Full uses 1080p, which cannot use a middle frame. Flux 3 supports 720p and 1080p, and can use the middle image without a first or last frame.

2. Write the Middle frame description for that moment. Use @ to tag the references it needs, then click Generate. An available first frame can guide the new image; you can also use Import middle frame from computer or Use frame from shot.

3. Set Middle frame time (seconds). It starts halfway through the model's actual output duration, which may differ from your requested duration. H3 Max and Flux 3 round the time to the nearest 1/24 second and keep it strictly between the opening and ending frames.

   

   

Changing the duration or model recomputes the effective time shown and sent, clamping it inside the new output while keeping your authored time. Switching to an unsupported model or quality keeps the middle image, description, time and history; the card shows the first-frame view instead. An enabled middle image blocks video generation until you choose a supporting model and quality or Skip middle frame in the notice.

Middle description, time, image attachment and Skip changes can be undone. Switching frame tabs only changes your view. Undo changes the creative edit; it does not refund a completed generation.

Motion reference

One video clip steers the shot's camera movement, subject action and pacing. Characters, wardrobe, setting and colour still come from your references. Motion is available on H3 Max, Seedance 2.5, Wan 3.0 and Wan 3.0 Prime.

1. Choose a supporting video model, then open Motion at the top of the shot card.

   

   

2. Under Choose a video, pick from Project videos or Generated shots, or click Upload to bring a video into the project. A clip must be at least 2 seconds long.

3. Drag the highlighted window to choose the movement to follow. The Left and Right arrow keys move it by 0.1 second; hold Shift to move by 1 second. The window follows the shot's duration, up to the model's clip limit: 30 seconds for Seedance 2.5 and 15 seconds for the other Motion models. A shorter source clip is used whole, and the tab says so.

4. Describe the shot's motion under Video, check the Estimated cost on Generate video, then generate. Its breakdown shows the effect of the clip as Motion reference.

Choose another video to replace the clip, or Remove to detach it. Attaching, replacing, moving and removing a clip can be undone. Switching to an incompatible model keeps the clip and refuses generation; remove it or switch back to a supporting model.

Dialogue

Open Dialog on the card to give the shot spoken lines.

1. 1. Choose the speaker and write the exact words. Each character or object needs a saved ElevenLabs voice; choose that voice in its Voice tab. The lip-sync model animates the face, while ElevenLabs supplies the voice.
   2. Use Insert timed pause at the start of a line to delay its speaker, or between phrases. Edit the marker, such as [pause: 3s], to set the exact duration. These pauses are inserted locally; they are never read aloud or billed as generated speech. They add silence before the generated take, preserving its natural breaths and phrasing.
   3. Use Delivery to direct the performance. The expression buttons insert vocal events such as sighs or throat clearing. Consistency changes how variable the performance is; it does not control loudness or emotional strength. Keep an expression tag beside spoken words. For example, [pause: 3s] [clears throat] Hello puts silence before the expression and speech. A tag isolated by a timed pause cannot be generated.
   4. Generate line or Regenerate line records just that line. Preview line lets you listen independently. Generate audio builds the complete dialogue. Hover over a generation button for its estimate: every submitted speech segment is priced separately, including tags.
   5. Changing the seconds on existing pauses reuses saved takes. Update timing recomposes them without a new speech charge. Failed segments leave completed takes available, and changed words or voices mark the combined audio out of date. Timestamp ranges such as [5–8s] belong outside spoken text.
   6. Listen to the complete preview. If speech plus pauses exceeds the shot, choose Extend shot or Split shot; speech is never cut automatically. A single overlong line must be divided into shorter lines before using Split shot.

Dialogue-driven video uses the generated ElevenLabs voice and runs a separate lip-sync pass automatically after the video render. This works with video models that accept the dialogue file and with models that have no audio input: the latter render picture first, then sync the saved dialogue. Hover over Generate video to see the video and lip-sync quotes and their combined estimate before starting. The finished result keeps the timing of the lip-sync pass. Its estimate includes a small allowance for an encoded final frame; the completed video is charged by its measured length plus cutipAI's fee. If its measured lip-sync estimate is higher than the amount shown before generation, the saved video continues through lip sync when your balance covers the updated estimate. If your balance is too low, add funds and use Retry lip sync; the paid video does not need to be generated again. If pricing is unavailable, generation waits for a quote. If that pass fails, Retry lip sync uses the saved paid video without generating the video again; it shows a fresh lip-sync quote. A previously submitted pass is recovered rather than submitted twice. Older videos with an unfinished soundtrack still offer Retry soundtrack. Send to selected audio track places the available recording at the playhead on the selected, unlocked audio track. You can send it again as often as needed, even when the video already contains dialogue, the video is not on the timeline, or the script has changed since recording. Each click reuses the existing audio without generating it again. Audio in the destination range is replaced using normal timeline overwrite behavior; later clips stay in place. Use Undo to reverse a placement. The downward arrow's tooltip names the selected track and placement behavior.

Regenerate audio stops and hides the current preview while the replacement is rendering. The new recording appears when ready. If rendering fails or is cancelled, the previous recording is available again; regeneration does not delete it.

Video

1. Open Video on the card. In Shot description, describe the motion, behaviour and sound, then choose a duration and click Generate video. Durations follow the model used for this shot: Grok Imagine Video 1.5 starts at 1 second, while Seedance 2.5 starts at 4 seconds. Models with fixed lengths offer only those lengths.

2. When the video is ready, click Send to timeline. To reuse the video in an existing edit, select one picture clip in the timeline and click the Replace selected clip icon beside Send. The dialogue preview has the same icon for replacing a selected audio clip. A selected picture and its linked sound count as one video target. Replacement keeps the original clip's timing, source trim, effects and other editing attributes, plus its original Storyboard link. Attached sound follows a video replacement; detached sound stays unchanged. Shorter media keeps the existing clip duration without automatic fitting. Undo restores the replacement in one step. No new generation is needed.

Generate video needs a shot description and, ordinarily, a first frame or the first frame set to Skip. An enabled middle image can supply the visual input on a model that does not require a first frame, such as Flux 3. Required endpoint frames, supported quality and any attached motion clip are checked too; the tooltip or notice names what to fix.

If an existing shot has dialogue, Improve lipsync can re-time the mouth to the current recorded line. It is a separate paid pass; review its estimate before clicking. A clear, visible speaker's face gives the best result. A face turned away or covered may not move convincingly. The current lip-sync upload accepts MP4 video so its length can be verified before a paid request starts.

1. Use Improve lipsync for the highest-quality lip sync available in the editor. The same model is used automatically after a dialogue video, with no quality choice to make. Check its live, account-priced estimate for the measured shot length before starting; it includes your account's service fee and rounds the measured video length up to the next whole second for billing. The original dialogue recording and video length are preserved. Existing takes remain in the project, and improving an older lip-sync take starts from its original video. If an older automatic take has lost that link, Improve lipsync checks saved project history and recovers the matching original automatically, continuing past newer history entries that already contain a lip-sync pass. It checks that original and refreshes the estimate before starting. If the original is unavailable, no new job starts; restore the original from History before trying again.

   Automatic routing starts disabled. When enabled for lip sync, you approve the current estimate and maximum before submission. It chooses the cheapest verified connection for the same highest-quality model. Currently only one connection is available, so a confirmed failure ends the job without charging your wallet. Failed attempts and costs above the approved maximum are absorbed by cutipAI. A lost connection or a local timeout keeps the accepted job available for recovery; it never starts another paid copy. Other generation tools retain their own billing terms.

Click Edit under the shot's video to change it with AI. See Editing a video in a project.

Following progress

The footer of the board counts the scenes and shots, and how many are done, generating and empty. Cancel all stops every generation on the board; cancelling is free. Every job also appears in the Job Queue — see Generating in the editor.

Export the board

Export in the toolbar saves the storyboard as Markdown or a Word document, or opens it to Print.

1. A shot's Open in 3D Builder action returns to its linked set. In the builder, choose the storyboard shot and Link blocking reference to associate a saved camera reference. Apply to selected shot explicitly adopts a reviewed image or video; generating alone keeps results in Assets. See 3D Builder.

What's new

27 October 2026: A higher lip-sync estimate for a finished paid video no longer interrupts dialogue generation when your balance covers it. The final charge follows measured usage plus cutipAI's fee. If your balance is too low, add funds and retry lip sync from the saved video.

26 October 2026: Timed pauses are the single silence control for dialogue. Delivery, Consistency and vocal expression buttons are visible without expanding settings, and the dialogue panel shows duration without model names. The Consistency menu uses the same styling as Delivery.