Pipeline steps
Script: generate, paste, or revise
There are three ways to get a script into a project: generate one from your project brief, paste your own, or revise the current script with a free-text instruction. Scripts are versioned, so you can iterate freely — every past version stays available and can be re-activated at any time.
Generate, paste, or revise
- ·Generate — Framesail writes a script from your project brief, in the channel style's narrative voice.
- ·Paste your own — save a script you wrote elsewhere. Everything downstream works the same whether the script was generated or pasted.
- ·Revise — give a free-text instruction ("tighten the intro", "make the ending land harder") and Framesail rewrites the current script accordingly, as a new version.
Revising, re-generating, and hand-editing are all normal workflow — the script step is meant to be iterated until the words are right, because assets, voiceover, and storyboard are all built from it.
Length control
The project's script length target (in minutes) converts to a word target at roughly 150 words per minute, held within about ten percent. Set it in the project settings — see channels, projects, and settings.
Narration vs. dialogue
The project's voice mix setting controls whether the script is written as pure narration or includes character dialogue, on a range from narrator-only through balanced to character-led. Dialogue is written as Charactername: "line", and each speaking character later gets its own voice in the voiceover step.
Versions
Every save, generation, and revision creates a new script version, and any past version can be re-activated. That makes big swings cheap: try a rewrite, compare, and roll back if the earlier draft was better.
Models and the expert drawer
Script generation and revision each have a user-pickable model — current options include GPT-5.4 (plus Mini and Nano), Gemini 2.5 Flash, Claude Sonnet 4.6, and Claude Opus 4.8. Like every generation step, the script prompt is built from named sections you can override per run in the expert drawer, alongside the model picker.
Once the script is right, the next station is the script scan, which extracts your cast, world, and voice blocks. For the broader picture of how a script becomes a video, see script to video.