How Pandrator works

A project keeps its source material, settings and results together. Pandrator also calls a project a session, especially in its API and MCP tools.

Sources and results

Import a document, paste text, upload a clip or supply subtitles. Processing stages create separate results, called artifacts. A transcription, corrected text, translation, generated take and export are different artifacts.

Choose the version you want before running the next stage. Changing the text does not automatically regenerate speech that was made from an earlier version.

App and Manager

The browser interface is where you prepare and review projects. Closing its tab does not stop background work. Pandrator Manager installs the app and optional services, launches them and handles updates and repair.

Local speech and transcription models run on your machine. Configured APIs send the relevant input to an external provider. A language model can help with cleanup, correction or translation; it is separate from the model that generates speech.

A useful starting rule

Use one source, one voice and a short output first. Add translation, extra speakers, music or video editing after that result is working.