> For clean Markdown of any page, append `.md` to the page URL.
> For a complete documentation index, see https://docs.sarvam.ai/llms.txt.
> For full documentation content in one file, see https://docs.sarvam.ai/llms-full.txt.
> For AI client integration (Claude Code, Cursor, etc.), connect to the MCP server at https://docs.sarvam.ai/_mcp/server.

# Speech to Text

Speech to Text transcribes audio or video files into text, with speaker diarization for multi-speaker recordings across regional Indian dialects. Upload a recording and get back a readable transcript - no manual re-listening and typing required.

## Popular Use Cases

* Transcribing meetings and interviews for notes or records.
* Reviewing call-center recordings for QA with speaker-separated transcripts.
* Generating captions or subtitles from raw audio/video footage.

## Key Facts

Uploads accept **MP3, WAV, M4A, FLAC, OGG, and AAC**, up to **500 MB** per file.

Diarization works best when you tell it how many speakers to expect. If you're unsure of the exact count, round upward rather than down - it's easier for the model to merge extra speaker slots than to split an undercounted one.

Code-mixed speech (e.g. a regional language with English financial or technical terms) transcribes as-is - a real example output includes Gujarati text with embedded English terms like "IDBI bank," "share," and "disinvestment."

## In This Section

Upload or record audio, set the language and speaker count, and export the transcript.

Set up for accurate transcription and understand the known hard cases.