Spaces:
Running
Running
| title: NepaliConformer | |
| emoji: 🏔️ | |
| colorFrom: red | |
| colorTo: gray | |
| sdk: gradio | |
| sdk_version: "5.44.1" | |
| python_version: "3.11" | |
| app_file: app.py | |
| pinned: false | |
| license: cc-by-nc-4.0 | |
| short_description: Nepali ASR built for real telephone calls, not read speech | |
| # NepaliConformer | |
| Transcribe Nepali speech with a model trained for **real telephone audio**. On the NepTel | |
| real-call benchmark it scores 33.8% WER where Whisper-large-v3 zero-shot scores ~99%. | |
| Full release — models, benchmark, honest results including every measured limitation: | |
| **https://github.com/Ampixa/nepaliconformer** | |