Docs · guides
Movie & Ads Studio
Produce films and video ads from a written brief: AI drafts the scenes, renders a storyboard with consistent characters, waits for your approval, then animates and cuts the final video.
Movie & Ads Studio
The Movie Studio at app.teamday.ai/movie turns a written brief into a finished video through a guided pipeline: storyline → storyboard → approval → animation → final cut. Every layer is regenerable on its own, and nothing expensive renders until you approve the storyboard.
The pipeline
A Next step bar at the top of every movie tells you exactly where you are and what to do next:
- Write the brief. Describe the film or ad — or click a preset (Product ad, Explainer, Brand story, Testimonial) to prefill a proven structure. AI drafts every scene: title, camera motion, narration line, music cue, and the image prompts for the first and last frame.
- Generate the storyboard. One click renders a keyframe for every scene. End frames generate image-to-image from their start frames — automatically, as each start frame finishes — so characters, set, and lighting carry through.
- Review and approve. Open the studio (chat on the left, storyboard on the right), comment on any frame, and click Fix comments to send all open feedback to the producer agent in one message. Nothing animates until you click Approve board.
- Animate. Each scene animates from its first frame to its last (on models that support end-frame conditioning). The next scene's start frame can continue from the previous clip's actual final frame.
- Render the final cut. Clips, per-scene narration, and generated music are muxed into one video with title cards, crossfades, name-tag overlays, and your branded end card.
Formats
Set the Format once per movie — every keyframe, clip, preview, and the final render follow it:
| Format | Render size | Use for |
|---|---|---|
16:9 | 1920×1080 | YouTube, web, presentations |
9:16 | 1080×1920 | Reels, TikTok, Shorts |
1:1 | 1080×1080 | Feed ads |
Keeping characters and products consistent
Keyframes are not independent dice rolls:
- End ← start. A scene's end frame generates image-to-image from its finished start frame (a checkbox on the frame, on by default).
- Scene ← scene. A scene's first frame can continue from the previous scene's end frame — or from the previous clip's actual final frame once it has rendered.
- Product / style reference. Upload a product photo, character sheet, or brand shot in the movie header (≤ 8 MB), or pick a generated image from the library. It is passed as an extra reference into every keyframe generation, so ads feature your exact product.
Image-to-image runs on OpenAI GPT Image 2 and ByteDance Seedream (Auto picks one of them). Selecting an image model that cannot honor references fails with a clear error rather than generating an unrelated image.
Models
Pick per movie from the header, or leave both pickers on Auto:
- Keyframes: Seedream 5.0 / 4.5, GPT Image 2, Wan 2.7, Qwen Image 2.0, Flux 2, Grok Imagine, Gemini Flash Image.
- Clips: Seedance 2.0 (and Fast), Kling 3 Pro / 2.6, Wan, HappyHorse, Grok Imagine Video.
Narration is synthesized per scene with a selectable voice; music is generated to each scene's length and mood.
Working with the producer agent
Start studio creates a production chat scoped to the movie's folder (media/movies/<slug>/), with the storyboard pane beside it. The agent works keyframes-first under the same approval gate, uses image-to-image references for continuity, and addresses your frame comments. Every scene is a plain JSON file in the workspace, so anything the agent does is inspectable and versionable.
Cost control
- Storyboard frames are cheap image generations; video renders only after board approval — a hard gate, not a convention.
- Batch actions skip scenes whose jobs are already queued or running, so re-clicking never double-spends.
- Regenerate a single frame or a single clip instead of the whole movie.