Skip to the workspace

TextFlowKit

Local transcription workspace — runs on this machine only.

1 · Choose media

Drop a media file here

or choose a file

Or paste a URL

2 · Progress

No job running.

3 · Results

Download

Recent jobs

Language Engine Model Device Output folder (optional)

Output formats TXT SRT JSON VTT Markdown DOCX PDF

Label speakers (diarize)

Needs the optional diarize extra and a Hugging Face token with access to the gated pyannote model. Presence of the package is not proof the model is downloadable; the job fails loudly if either is missing.

Translate to (optional)

Uses the configured Ollama backend. The job fails loudly if the backend is unreachable. If that backend is remote or cloud-hosted (not on this machine), the transcript text is sent to it for translation — choose a local model if transcripts must not leave this machine.

Local interface — this page loads no analytics and no external assets; every page asset is served by this process. Transcription runs on this machine by default.

Network is used in a few cases. Choosing a URL source downloads that media from the internet. The first run of an engine may download its model. And if you configure remote/cloud translation, the transcript text is sent to that translation backend. With a local translation model and local media, no transcript leaves this machine.

Stop the local server?

This shuts down the local web server. It waits for the current jobs to finish, then stops; it does not interrupt a job that is still running. If you want to interrupt a job, press Cancel on that job first. Saved progress is kept, so a job can be resumed the next time you start the UI. Recent jobs are not deleted.

First use will download a model