Pipeline steps

Storyboard: the AI director's plan

The storyboard is the AI director's shot-by-shot plan for your video. It is generated from the finished script and the rendered voiceover — voiceover completeness is a hard prerequisite — so every segment carries real narration timing, not guesses. Each segment gets its slice of narration, an image or video prompt, transitions, focus animation, and tags naming exactly which characters, environments, and objects are in shot.

Pacing comes from the director style

There is no segment-length dial. How densely the video cuts — long held shots vs. rapid inserts — is decided by your style's director field, which is the pacing authority. To change pacing, edit the director style or pick a different director preset.

Focus tracks

Image segments can carry a focus track: a series of keyframes that re-frames the same still — pans, snaps, and zooms between focal points — in sync with specific narration words. Focus tracks are authored by the storyboard itself; there is no keyframe editor in the UI.

Continuation between segments

A segment can be marked as continuing from an earlier one, in which case the earlier segment's rendered image becomes this shot's primary reference — how the storyboard holds one moment across consecutive shots. The storyboard authors these chains itself, and you can set or clear them per segment.

Per-segment operations

Each segment card's menu offers:

• ·Split — cut one segment into two at a point you choose.
• ·Combine — merge with an adjacent segment.
• ·Change type — switch between generated image, generated video, fetched real photo, fetched b-roll, and overlay scene. Image⇄video swaps can carry the rendered frame across.
• ·Rewrite content — give free-text creative direction and the segment regenerates around it.

Real photos and b-roll

When your style's real-media policy is switched on (see style fields), the storyboard can source real imagery alongside generated shots: entity images via Wikipedia, Wikimedia Commons, and OpenLibrary, and stock photos and b-roll via Pexels. A vision model verifies each pick, and on a miss the segment falls back to a generated image — no beat is ever left broken.

Sound effect planning

The storyboard populates per-segment sound effects according to the project's SFX level setting (none, minimal, or frequent), and you can add or remove SFX on any segment from the audio library — see music & sound effects.

Models

The storyboard model is picked per run — current options include Gemini 3 Flash, Gemini 3.1 Pro, Gemini 3.5 Flash, GPT-5.4 (and Mini), Claude Sonnet 4.6, and Claude Opus 4.8. Gemini 3.1 Pro is the fidelity upgrade when a director style demands dense, hyper-kinetic cutting.

Fresh storyboard generation emits image and video segments; from there, each panel is rendered in the segment renders step. For the product overview, see the Storyboard AI page.