Convert video speech to text

Run local Whisper transcription and export an editable TXT transcript without sending speech to a processing server. The media stays on this device during normal processing.

Practical browser-local workflows for cutting, framing, compression, audio and subtitles.

Recommended flow

Open Convert video speech to text from this guide; the Video Editor starts with the relevant mode and preset already selected.

Choose the source video from the browser file picker. Inspect its duration, dimensions and available audio track before changing settings.

Adjust the prepared settings for this particular source. Use the player and timeline to check the beginning, middle, transitions and ending.

Export the result, open the downloaded file once and confirm duration, framing, sound and format before publishing or sharing.

What to watch

Run local Whisper transcription and export an editable TXT transcript without sending speech to a processing server.

Keep the original file unchanged so you can return to it if another format or framing is needed.

Browser codec support and available memory depend on the current device; use the compatibility report shown by the editor.

Choose settings from the destination backwards

Run local Whisper transcription and export an editable TXT transcript without sending speech to a processing server.

Automatic speech recognition is a draft, not final copy. Select the spoken language when automatic detection is uncertain, review names and numbers, and repair segment boundaries before exporting. WebVTT is a W3C timed-text format; SRT is widely used but is not presented here as a W3C standard.

Quality control before publishing

Open the exported file outside the editor. Confirm duration, orientation, framing, audio synchronization and container. Check the first frame, every edit boundary and the final frame; a technically successful export can still contain a clipped subject, silent section or unwanted padding.

Compare output duration with the intended active timeline.

Check the result at normal size and full-screen.

Keep the untouched source until the destination accepts the export.

Repeatable preset for agents

Agents can use the official recipe `transcribe_txt` through window.QuokkaPixVideoAgent or the Video MCP tools. The recipe prepares settings only: the source file is still selected locally in the browser, the exact payment tier is measured after selection, and the result manifest remains the source of truth for completion and QA.

recipeId: transcribe_txt

Official sources and scope

These links support the technical method or destination profile. Strict requirements and recommendations are intentionally kept separate. Sources were checked on 13 September 2026.

Transformers.js documentation

FAQ

Does convert video speech to text upload my video?

No. QuokkaPix performs the editing in the browser. Normal media processing does not send source or output bytes to a QuokkaPix processing server.

Why does the guide open the editor with settings already filled in?

The guide passes documented URL parameters to the same Video Editor. It selects a useful starting configuration; you can still change every setting after choosing the file.