Privacy and architecture
Local-first video editing without pretending the web is offline
Timeline Studio processes selected project media and supported AI workflows in the browser. It does not require uploading the project to an editing backend, but it still downloads application code, model artifacts and runtimes from configured hosts.
What is processed locally?
Timeline operations, preview composition, waveform work, supported inference and browser-side export operate on media selected from the device. Generated results become browser assets and do not automatically replace or publish the source project.
What still uses the network?
The application itself and large model files must be delivered over HTTP. Supported model families use owned Hugging Face and ModelScope mirrors pinned to immutable revisions, with provider-equivalent cache identities where documented. Optional remote generation connectors follow their own provider or loopback-service boundaries.
What is stored?
Preferences, application resources, supported model artifacts and authorized local voice profiles may be kept in browser storage. Clearing site data can remove those caches and profiles. Browser capacity limits can also cause an optional asset to run in memory without being cached.
What are the tradeoffs?
Local processing reduces project upload time and backend exposure, but transfers work to the user's CPU, GPU, memory and storage. First-use model downloads can be large, and performance varies significantly by browser and device.