Sathi Transcribe

Beta · 0.0.1

Speech AI for Nepal's multilingual reality.

One sentence in Nepal moves through Nepali, English, and Hindi without pausing — and Japanese follows families abroad. Sathi Transcribe writes down what was actually said, in the scripts it was said in, instead of forcing everything into one language.*

Nepali · English · Hindi · Japanese — and natural mixes of them.

Seeking speaker and evaluation partners for Newar, Gurung, Magar, Tamang, Tharu, Limbu, and other languages of Nepal.

Try it now

Try a short recording or an audio file. Audio is processed externally during beta; read the current terms before continuing. Current processing and privacy terms →

Up to 30 seconds · 1 MiB per clip.

Record a short clip or choose an audio file.

A simple API

POST /v1/audio/transcriptions — batch JSON file transcription through the official OpenAI Python or server-side JavaScript SDK. Beta keys are granted by hand with finite limits.

curl https://api.sathi.ai/v1/audio/transcriptions \
  -H "Authorization: Bearer <your beta key>" \
  -F "file=@conversation.m4a;type=audio/mp4" \
  -F "model=sathi-transcribe-1"

→ {"text": "आज हामी meeting मा जाने हो...",
   "language": "ne", "duration": 12.43}

Pricing

Closed beta — every access grant states its finite allowance.
General-availability pricing will be published after measured cost and quality tests.

Enterprise and early access: info@sathi.ai