Is this Thai speech-to-text really free?

Yes. The page is public, no signup is required, and there are no length limits. The recognition model runs entirely in your browser. For higher accuracy on noisy or accented audio, run your file through Subformer's cloud transcription in Subtitles-only mode.

How does the Thai transcription work in the browser?

It uses an open-source neural speech recognition model compiled to WebAssembly and run via ONNX Runtime Web. The model (~75 MB) is downloaded on first use and cached in your browser. Subsequent visits skip the download entirely and can transcribe offline.

Is my audio uploaded to a server?

No. Recognition runs locally in your browser. The only network calls are to download the model and the WebAssembly runtime - both are static assets. Your audio file and the generated transcript never leave your device.

What audio formats are supported?

Anything your browser can decode: MP3, WAV, M4A, AAC, FLAC, OGG/Opus, plus video containers like MP4, MOV, and WEBM (the audio track is extracted automatically).

Can I download SRT or VTT subtitles?

Yes. After transcription you can download the result as plain text (.txt), SubRip subtitles (.srt), or WebVTT (.vtt) - both subtitle formats include segment-level timestamps so you can drop them straight into any video editor or player.

Tiny vs Base - which model should I pick?

Tiny (~40 MB) is fastest and great for clean studio audio. Base (~75 MB) is roughly twice as accurate on noisy or accented audio and only slightly slower. If you're not sure, start with Base - it's the default. For the highest accuracy on hard audio (accents, noise, overlap), use Subformer's cloud transcription in Subtitles-only mode.