Record, upload or link an audio file and get a transcript back. No API key and no backend: this page calls puter.ai.speech2txt() in your browser through Puter.js, using the tutorial you shared.
Pick one way in. You can listen back before you transcribe.
Tap the mic and speak. Tap again to stop. Best in Chrome or Edge on a secure page.
Good default for voice notes and short clips.
A hint only. Leave on auto detect if you are not sure, or if the clip mixes languages like Urdu and English.
Test mode returns a sample response without a live call, handy to check the page wiring before you spend anything.
First live call may open a Puter sign-in popup. That is Puter handling payment under its user-pays model, not this page collecting a key.
Copy it, download it as a text file, or run another clip.
A mic recording becomes a Blob, a file stays a File, and a URL stays a link. All three are valid inputs for Puter.
The page calls the function below with your chosen model. Puter may ask you to sign in, because the usage is billed to the user, not to an API key hidden in the page.
The result object carries text, and for some models also speaker segments or word timings, which this page shows under the transcript.
// What the Transcribe button runs
const result = await puter.ai.speech2txt(audioSource, {
model: 'gpt-4o-mini-transcribe'
});
console.log(result.text);
Straight from the Puter tutorial and the current speech2txt docs. One honest note: the tutorial shows a translate: true option, but the current docs say it is not supported and fails on OpenAI and Gemini, so this page does not offer it.
| Model | Best for | Notes |
|---|---|---|
| gpt-4o-mini-transcribe | Voice notes, quick tests | Fast default. Start here. |
| gpt-4o-transcribe | Important recordings | Highest quality OpenAI transcription. |
| whisper-1 | Multilingual audio | Classic Whisper model. |
| gpt-4o-transcribe-diarize | Meetings, interviews | Returns speaker segments, shown above when present. |
| gemini-3.5-transcribe | Word timings | Gemini provider. Audio up to 10 MiB per the docs. |
| grok-voice-transcribe-2.0 | Trying Grok | xAI provider. |