Ringg Parrot STT V1.0 🦜
📁 Upload an audio file for transcription (supports WAV, MP3, FLAC, M4A, etc.)
🎯 Performance Benchmarks
Ringg Parrot STT V1 Ranks 1st Among Top Models, Outperforming OpenAI Whisper Large-v3 and Other Leading Solutions.
Parrot STT (Ringg AI) | 15.00% | 15.92% |
⚠️ Benchmark Disclaimer
- Evaluated on a modified FLEURS subset to ensure consistent Hindi coverage
- Dataset issues include inaudible segments and repeated sentences caused by interruptions
- Background noise is prominent across many clips, impacting recognition quality
- Mixed Hindi-English speech often provides Hindi-only transcripts
- Currency, time, and year normalization is inconsistent with spoken forms
- Original transcripts lack punctuation, increasing WER for models that predict it
🙏 Acknowledgements
- Special thanks to @jeremylee12 for their contributions
- Built with NVIDIA NeMo models