Free. Yiddish-first.
Turn recordings into text and text into natural speech, built for Yiddish from the ground up. Use it on the web or call the API, both free.
No credit card. Sign in with Google or email.
Two models, one quiet workspace
Everything you make is saved to your private library, so you can come back, fix a word, and download it again.
Speech to text
Devniks ASR: the Yiddish speech model behind Sixday's phone search.
- Upload or record from your mic
- Timestamps for every segment
- Edit right in the transcript
- TXT, SRT, VTT, JSON, DOCX
Text to speech
Devniks TTS: a Yiddish fine-tune of OmniVoice.
- Natural male and neutral voices
- Clone your own voice from a short clip
- Speed and pitch control
- MP3 or WAV
Free API
The same models over plain REST.
- One API key, one header
POST /v1/asrandPOST /v1/tts- Send corrections back for credits
- Clear docs with curl, Python, JS
Why is it free?
Honestly: your recordings and corrections help us make the models better. When you fix a word the model got wrong, that fix teaches the next version. That's the trade, and we want it to be a fair one.
Correct a transcript or rate a voice and you earn extra free minutes.
How credits workBuild on it, free
Add Yiddish transcription to your phone line, app or archive. The free tier covers 120 audio minutes and 50,000 characters of speech a day per key.
# Transcribe a recording curl https://voice.devniks.com/v1/asr \ -H "Authorization: Bearer $DEVNIKS_KEY" \ -F file=@shiur.mp3 -F language=yi # → 200 OK { "id": "asr_8f3k2", "status": "done", "segments": [{ "start": 0.0, "end": 4.2, "text": "שלום עליכם, דאס איז..." }] }