CoolFace
Apppublic

kingabzpro/Urdu-STT-with-GPT-OSS

sourceHugging Faceapache-2.0updated 1y agoView on Hugging Face
4likes
App README

🏎️ Faster Urdu ASR

This Space provides state-of-the-art Urdu Automatic Speech Recognition (ASR) built on Faster-Whisper, fine-tuned for Urdu. In addition to transcription, it offers optional polishing with Groq’s `openai/gpt-oss-120b` LLM to improve Urdu grammar, punctuation, and fluency.

✨ Features

  • β€”πŸŽ€ Audio input via upload or direct microphone recording
  • β€”πŸ“œ Multiple output formats: plain text, .srt, .vtt, .json
  • β€”βš‘ Built on Faster-Whisper (CT2) for efficient GPU/CPU inference
  • β€”πŸ€– Optional LLM polishing with Groq API for natural, improved Urdu text
  • β€”πŸ”‘ Works with environment variable GROQ_API_KEY or via UI input

πŸš€ Usage

  1. 1.Upload or record an Urdu audio file.
  2. 2.Choose output format (text, srt, vtt, json).
  3. 3.(Optional) Enable LLM Polishing to improve transcription quality.
  4. 4.Provide a valid `GROQ_API_KEY` if not set in your environment.
  5. 5.Adjust temperature and system prompt as needed.
  6. 6.Click Transcribe and view/download your results.