Most of what I dictate is in English, and not especially exotic English. Emails, messages, prompts for Claude, notes about whatever I'm building. Stock speech recognition handles most of it well, then trips on precisely the words I use most: model names, CLI commands, cloud services, prices, and the odd turn of phrase an Irish speaker brings with him to Jerusalem.
Fritter is my answer to that. It is a fine-tune of NVIDIA's Parakeet TDT 0.6B v2 on my own voice, for English dictation without code-switching. "Fritter" is the code name for the model line (I've taken to naming these after food, as the Hebrish models Jachnun and Sabich show), so later retrains keep the name and bump the version. This is v1.0.
Hugging Face Model Fritter-ASR/Fritter Stock speech models stumble on the words technical people say all day. Fritter is tuned for higher accuracy on technical vocabulary, including AI: model and product names, CLI and cloud terms, brands, acronyms and… ↓ 1.3kAutomatic speech recognitionCC-BY-NC-4.0Updated Nov 2026 View model on Hugging FaceHow it differs from my other models
My Hebrish models are about the Hebrew words that turn up in English speech here. Fritter deliberately leaves that out. It's for when I'm dictating plain English and want tech vocabulary, numbers and brand names to come out right. Boxty, the no-Hebrew sibling of the Hebrish models, covered similar ground. Fritter is my own model line, trained on the speaker takes in my read-aloud dataset that have no code-switching: 320 minutes of my voice, plus 10% LibriSpeech so it doesn't forget general English.
The numbers
On held-out test clips that were never used for training or for picking the checkpoint:
Tech-sentence word error rate drops from 7.53% (stock Parakeet v2) to 4.01%.
Hebrew-free English WER drops from 7.46% to 4.73%.
Recall of the target tech and English terms rises from 65.1% to 79.6%, and numbers from 33.9% to 50.0%.
There's a cost, and I'd rather be upfront about it. On LibriSpeech, which isn't my voice, it is slightly worse than stock (1.47% against 1.23%), and long-form dictation is about a point worse (12.16% against 11.05%). So Fritter is better at *my* dictation and roughly where stock is on everyone else's. The test sets are small too (319 clips for Hebrew-free English), so I'd treat differences under a point as noise.
Try it in the browser
The demo Space runs the int8 ONNX build on your own device, so the audio never leaves your machine:
It is tuned to one voice, accent and phone, so if you try it yourself you will probably see less of an improvement than I do.
Getting it
The .nemo checkpoint is on Hugging Face, along with GGUF builds (Q8_0 recommended, Q4_K_M and F16) for transcribe.cpp, which is what Offline Voice Input on Android and Handy on Linux use. The licence is CC-BY-NC-4.0, so it's free for personal, non-commercial use.
The training data comes from the same pipeline as everything else in this series. New words start as sentences written by the Contextual Sentence Creator, I read them aloud on my phone, and the next Fritter gets retrained on the takes.
Hugging Face Model danielrosehill/Contextual-Sentence-Creator A small model for one job. Given a term, a one-line context and a count n, it returns a JSON array of exactly n natural English sentences (6 to 18 words). Each sentence uses the term once, spelled as given, in the… ↓ 401Text generationAPACHE-2.0Updated Nov 2026 View model on Hugging FaceUpdate, 2026-11-09: Fritter-TTS v1, with before and after samples
Fritter now has a text-to-speech sibling. Fritter-TTS is a separate model from the dictation model above: a voice in my own voice (Irish-accented English) for technical and general English, so AI model names, Linux commands, acronyms and version numbers are read the way I say them.
Fritter-TTS v1 was released on 2026-11-09. It is a full fine-tune of Qwen3-TTS-12Hz-1.7B-Base on 9,125 of my recordings, about 12.4 hours, trained for 3 epochs at a learning rate of 7e-6. Every take was first cleaned with NVIDIA NemotronLabs RE-USE (the RE-USE-GGUF build, run through audio.cpp), then a podcast-style processing chain.
Hugging Face Model nvidia/NVIDIA-NemotronLabs-RE-USE (Note: For overlapping speech, this model is trained to preserve the main speaker while suppressing other speakers.) ↓ 48.7k♥ 231Audio to audioOPENMDW-1.1Updated Oct 2026 View model on Hugging FaceThe model is gated, by request, under CC-BY-NC-4.0: Fritter-ASR/Fritter-TTS. GGUF builds for qwentts.cpp are in Fritter-ASR/Fritter-TTS-GGUF, and all the samples are on the Fritter-TTS-Samples Space.
Hugging Face Model Fritter-ASR/Fritter-TTS A single-speaker text-to-speech fine-tune of Qwen3-TTS-12Hz-1.7B-Base on one voice: Daniel Rosehill, Irish-accented English. It is the text-to-speech side of the Fritter line: technical and general English, with AI… ↓ 49Text to speechCC-BY-NC-4.0Updated Nov 2026 View model on Hugging Face Hugging Face Model Fritter-ASR/Fritter-TTS-GGUF ↓ 4Text to speechCC-BY-NC-4.0Updated Nov 2026 View model on Hugging Face Hugging Face Space Fritter-ASR/Fritter-TTS-Samples Listening samples for Fritter-ASR/Fritter-TTS StaticLiveCC-BY-NC-4.0Updated Nov 2026 Open the SpaceBefore is the untuned base model, Qwen3-TTS-12Hz-1.7B-Base, cloning my voice zero-shot from one reference recording. After is Fritter-TTS v1. Both read the same text, and the samples were not picked by ear. Passages of more than three sentences were generated sentence by sentence and joined. Shortest first:
1. Restart the Docker container and check the logs again.
2. I'll SSH into the NAS and see why the backup job failed overnight.
3. The new release needs Python 3.12 and at least 16 GB of RAM.
4. The VPN drops whenever the router renews its DHCP lease. I've pinned a static IP for now and opened an issue on GitHub.
5. The fine-tune ran on an L40S for about an hour. After that I quantized it to Q4_K_M with llama.cpp and pushed the GGUF to Hugging Face.
6. I tried Claude Sonnet 4.5, GPT-5 and Gemini 2.5 Pro on the same prompt. Sonnet gave the cleanest JSON, so that's the one going into the pipeline.
7. The home server runs Ubuntu 24.04 with everything in Docker Compose. Home Assistant, Jellyfin and a small Postgres database each get their own container. Backups go to a Synology NAS every night, and a copy is synced to object storage once a week. It's not glamorous, but it has run for two years with almost no attention. (sentence by sentence)
8. The build failed on CI but worked on my laptop, which is always a bad sign. It turned out the runner was on Node 18 while I was on Node 22. I pinned the version in the workflow file, cleared the npm cache and reran the job. Twenty minutes of debugging for a one-line fix, which is about the usual ratio. (sentence by sentence)
9. My dictation setup uses a fine-tuned Parakeet model running locally on the GPU. I press a key on the macro pad, speak, and the text lands wherever the cursor is. The fine-tune fixed most of the technical vocabulary, things like Kubernetes, PyTorch and OpenRouter. Numbers were the last thing to sort out, so now I record version numbers the way I actually say them. (sentence by sentence)
10. Most of my day now involves working alongside AI agents in the terminal. I describe what I want, the agent reads the repository, proposes a change and runs the tests. The trick is giving it enough context without burying it in detail. A short CLAUDE.md file with the project's rules does more than a long prompt ever did. When something goes wrong, I ask it to write down what it learned, so the next session starts smarter. It's not perfect, but it has changed how much I can get done in an afternoon. And I dictate almost all of it, which is why the speech models matter so much to me. (sentence by sentence)