Generation pipeline
From an empty scene grid to a finished video, there are three asset types to generate per scene: images, audio, and motion video. Generate them in any order, re-roll anything that disappoints, and watch the grid fill in.
The three stages
- Images: one still per scene. Usually generated first because the motion generator needs an image as input.
- Audio: the narrator reads each scene's text chunk. Can run in parallel with images.
- Motion: a short video animating each image. Requires the image. Optional if you want a still-only slideshow.
Generating assets
Above the scene grid, a primary Generate button picks the most useful next action:
- If any scene is missing an image, it generates images first.
- Otherwise, if any scene is missing audio, it generates audio.
- Otherwise, if any scene is missing motion, it generates motion.
- When everything is filled in, the button becomes Regenerate all.
A chevron next to the button reveals targeted options: Generate images only, Generate audio only, Regenerate this selection, and so on.
Progress tracking
Each scene tile shows a status badge while its asset is generating. Status pills at the top of the grid summarize overall state: "12 missing images, 8 generating, 2 failed."
The page polls the server for status updates. You can leave the page and come back; generations continue server-side. Push updates appear when you return.
Handling failures
A failed generation keeps the scene in its previous state (its last successful asset or empty). A red badge on the tile shows the failure. Reasons vary:
- Content policy trip: the model's safety filter rejected the prompt. Edit the visual direction or text chunk and retry.
- Provider rate limit: Veo or Imagen is throttling. Wait a minute and retry.
- Credit budget check: you ran out of credits mid-batch. Top up and retry from the failed scenes.
Click the failed tile to open the scene editor; the error message is at the top of the sidebar. Retry with Regenerate.
Applying storyboard defaults
Every scene has its own model and tier settings, but the storyboard also has defaults. In the Storyboard panel, Apply defaults to all scenes copies the current defaults (image model, image tier, motion model, motion resolution, motion audio mode, motion duration) to every scene.
Useful when you decide mid-project to bump everything to 1080p, or downgrade to Fast tier to iterate. A separate Only unset option leaves per-scene overrides in place and only fills in the scenes that were using the defaults.
Costs
Generation costs vary by tier, model, duration, and resolution. The composer previews the total batch cost before you confirm. Typical ballparks for a 15-scene storyboard at default settings:
- Images (Gemini 3, Standard): ~90 credits total.
- Audio (Google Cloud TTS, default length): ~30 credits total.
- Motion (Veo 3.1 Lite, 6s): ~270 credits total.
- Motion (Veo 3.1, 6s, 1080p): ~1,800 credits total.
Play a silent preview first (images and audio only, skip motion) to make sure the structure is right before committing to motion credits.
Recommended order: 1) generate all images, 2) review in the grid and re-roll any that don't land, 3) generate audio, 4) listen end-to-end with still images, 5) generate motion only after the narrative is solid. This keeps motion credits (the most expensive part) pointed at scenes you know are final.
Previewing the full storyboard
The preview player at the top of Scene Mode plays every scene in sequence. Images with narration if no motion exists yet, motion if it does. The player respects per-scene video placement (Loop, Start, End, etc.). Use the scene-marker ticks on the scrubber to jump directly to a scene.
Download the full preview as a single MP4 from the preview's three-dot menu. This is a draft export, good enough for sharing rough cuts.
Where to go next
- Importing to the editor for polish.
- Credits for pricing details.