Prompt-first local generation
Generate AI music in the browser for your video project
Timeline Studio runs its AI music workflow locally through WebGPU. Describe the desired track, select compact style, mood, lead, duration and tempo controls, then keep the completed audio in My assets until you choose where to place it.
How prompting works
The main input is a multilingual description. Non-English descriptions are translated to an English model prompt before inference; the Advanced section exposes that prompt for inspection. Generation stops clearly if browser translation is unavailable rather than silently sending an unsuitable prompt.
Model setup and repeat generation
Model artifacts download in parallel into a versioned browser cache, while WebGPU sessions are created serially for stability. The initialized worker stays alive for repeat generations in the same page, so a second track is not presented as another complete model download.
Editable asset workflow
Completed music is added to My assets and is never inserted into the timeline automatically. AI Music routes to the dedicated Music track when the user adds it, keeping narration and ordinary audio lanes separate.
Device requirements
Generation requires a browser and GPU with adequate WebGPU support and memory. The first model setup can be substantially longer than repeat generation because model artifacts must be downloaded and cached.