Spaces:
Running
Running
| title: README | |
| emoji: 🧇 | |
| colorFrom: yellow | |
| colorTo: green | |
| sdk: static | |
| pinned: false | |
| # Fritter | |
| Speech models for English dictation, fine-tuned on one Irish-English speaker's recorded voice: AI model names, | |
| Linux and developer vocabulary, brand names, acronyms, and numbers written as digits. | |
| **Try it:** [Fritter Demo](https://huggingface.co/spaces/Fritter-ASR/Fritter-Demo) runs Fritter in your browser, on your own device. | |
| **Putting Fritter in an app?** [Fritter Models](https://huggingface.co/spaces/Fritter-ASR/Fritter-Models) lists every build with direct download links, the licence (free for non-commercial use with attribution) and the attribution line to use. | |
| ## Models | |
| | Model | What it is | | |
| |---|---| | |
| | [Fritter](https://huggingface.co/Fritter-ASR/Fritter) | Fine-tune of NVIDIA Parakeet TDT 0.6B v2 for dictation. NeMo, GGUF, ONNX and sherpa-onnx builds | | |
| | [Fritter Nemotron 0.6B](https://huggingface.co/Fritter-ASR/Fritter-Nemotron-0.6B) | The streaming version, for live captions and push-to-talk. NeMo and GGUF | | |
| | [Fritter Nemotron 0.6B (sherpa-onnx int8)](https://huggingface.co/Fritter-ASR/Fritter-Nemotron-0.6B-sherpa-onnx-int8) | Streaming sherpa-onnx build, 560 ms chunks | | |
| | [Fritter TTS](https://huggingface.co/Fritter-ASR/Fritter-TTS) | A text-to-speech voice trained on the same recordings, built on Qwen3-TTS | | |
| | [Fritter TTS GGUF](https://huggingface.co/Fritter-ASR/Fritter-TTS-GGUF) | GGUF builds of Fritter TTS | | |
| ## Spaces and data | |
| Collections: [Fritter-ASR-Parakeet](https://huggingface.co/collections/Fritter-ASR/fritter-asr-parakeet-6af1b23cae51f0c28e3a352d), [Fritter-ASR-Nemotron](https://huggingface.co/collections/Fritter-ASR/fritter-asr-nemotron-6af1b7b1db780deb8706be17), [Fritter TTS](https://huggingface.co/collections/Fritter-ASR/fritter-tts-6af1b23c68e089a2acf1acdb). | |
| - [Fritter Demo](https://huggingface.co/spaces/Fritter-ASR/Fritter-Demo): speech recognition in the browser | |
| - [Fritter Models](https://huggingface.co/spaces/Fritter-ASR/Fritter-Models): downloads, licence and attribution for every model | |
| - [Fritter TTS Samples](https://huggingface.co/spaces/Fritter-ASR/Fritter-TTS-Samples): listen to the TTS voice | |
| - [Tracker](https://github.com/Fritter-ASR/Fritter): releases, data growth and a dated log, on GitHub | |
| - [Training Sources](https://huggingface.co/datasets/Fritter-ASR/Training-Sources): the datasets each release trained on | |
| ## History | |
| Fritter started as a proof of concept on [Daniel Rosehill's](https://huggingface.co/danielrosehill) personal | |
| account (Fritter v1.0 to v1.3, Fritter Nemotron v1.0, Fritter TTS v1). It moved to this organisation in | |
| November 2026 so the models, the demo and the data index sit in one place and are easier to maintain. New | |
| releases are published here; the original repositories stay online so existing links keep working. | |
| ## Licence | |
| The weights are released under CC-BY-NC-4.0: free for non-commercial use with attribution. Base models keep | |
| their own licences (NVIDIA Parakeet TDT 0.6B v2, NVIDIA Nemotron streaming, Qwen3-TTS). Not affiliated with NVIDIA | |
| or Qwen. | |