Skip to main content
Framesail AI
Docs

Pipeline steps

Scene renders: images, clips, and models

Scene rendering turns each storyboard scene into an image, and any rendered image can then be animated into a video clip. Model choice is per job, per run — there is no global image or video provider setting — and every shot can be regenerated on its own without touching the rest. You can also upload your own image for any scene.

Image models

Video models

A clip is made from a rendered image: open Animate scenes… from the menu beside Generate images and pick the scenes to animate. The cost is shown before anything renders.

Current options: H3 Max Turbo (the default), H3 Max, FLUX 3 Max, Gemini Omni Flash, Kling v3 Standard, Kling v3 Pro, Seedance 2 Pro, Seedance 2 Fast, and Seedance 2.5, plus VEED Fabric 1.0 for Talking Head clips. Each model exposes its own settings:

Clip duration is not a setting: it is derived automatically from the scene's narration length and snapped to the model's allowed durations (Gemini Omni Flash clips run 3 to 10 seconds, for example).

Prompts and the prompt editor

Every generation step's prompt is built from named sections. In the prompt editor you can override the editable sections per run — overrides are validated server-side — and reset any section back to its default. The prompt editor is also where the model picker and per-model settings live. Overrides are saved on the item you set them for: the script, the storyboard, a style template, a reference image, a scene render, or a voice block.

Regeneration rules

Each render draws on the scene's tagged reference images, which is how a shot fifty scenes in still matches shot one — see how character consistency works.