# k2-fsa/sherpa-onnx

> A collection of offline speech processing tools built on next-gen Kaldi and onnxruntime: speech recognition and synthesis, speaker diarization and identification, VAD, speech enhancement, source separation, and keyword spotting. Runs on servers, mobile, and embedded devices without an internet connection.

- Magnitude: 2.0 out of 10 — Steady
- Stars: 15,690 total · +25 stars measured 2026-11-02, 00:13–23:08 UTC
- Star trust: star growth looks organic
- Category: Language models · Language: C++ · License: Apache-2.0 · Created: 2022-09-01 · Last push: 2026-10-29
- GitHub: https://github.com/k2-fsa/sherpa-onnx · Homepage: https://k2-fsa.github.io/sherpa/onnx/index.html · Page: https://gitnova.dev/en/r/k2-fsa/sherpa-onnx

## Useful for

- Add offline speech recognition to an Android or iOS app using the provided bindings
- Generate video subtitles with built-in ASR and punctuation restoration
- Run VAD and speaker diarization on a Raspberry Pi to label meeting recordings

## Why it’s here

- Star-counter measurements on 2026-11-02 (UTC), 00:13–23:08: 15665 → 15690 stars (+25). This is the change over that interval.

## Star trust

Star growth looks organic.

Star-trust labels are heuristics based on the repository’s behavior, not a check of every stargazer.

## Numbers

- Forks: 1,838
- Issues and pull requests: 4,004
- Watchers: 120
- Average over the last week: 25 per day
- Usual pace: 21 per day
- Stars in the last hour (measured): 1
- Latest release: v2.0.1 (2026-10-26)

## Stars per day, last 30 days (oldest → newest, today is partial)

2026-10-04 … 2026-11-02: 20, 10, 19, 20, 15, 16, 24, 14, 25, 22, 20, 24, 18, 13, 20, 32, 22, 25, 20, 16, 26, 20, 20, 29, 22, 32, 29, 14, 23, 25

## Magnitude by day, last 14 days

10-01 1.7, 10-02 1.6, 10-03 1.5, 10-04 1.7, 10-05 1.9, 10-06 1.8, 10-07 1.9, 10-08 2.5, 10-09 2.5, 10-10 2.6, 10-11 2.7, 10-31 2.0, 11-01 1.8, 11-02 2.1

## More in this category

1. **karpathy/nanochat** — 6.7 · Breakout · Language models · Python · +250 stars measured 2026-11-02, 11:38–23:02 UTC
   A minimal experimental harness for training LLMs on a single GPU node, covering tokenization, pretraining, finetuning, evaluation and inference. It lets you train a GPT-2-capability model for about $48 and chat with it via CLI.
   Full card: https://gitnova.dev/en/r/karpathy/nanochat.md
2. **togg53192-cmd/jailbreaks** — 5.7 · Peaking · Language models · +696 stars measured 2026-11-02, 00:07–23:03 UTC
   A collection of 31 system prompts, persona locks, and AGENTS.md files for 25+ models (Claude, Gemini, Grok, GLM, DeepSeek, Kimi, Qwen, Gemma, GPT-OSS). Intended for research on how instruction hierarchies and policy overrides affect model…
   Full card: https://gitnova.dev/en/r/togg53192-cmd/jailbreaks.md
3. **FlashML-org/FreeToken** — 5.6 · Breakout · Language models · Python · +625 stars measured 2026-11-02, 00:07–23:03 UTC
   A Mixture-of-Experts inference engine for running frontier-scale open-weight MoE models (290B+) on consumer hardware — GPUs, CPUs, and host memory. Aimed at users who want to run frontier models locally.
   Full card: https://gitnova.dev/en/r/FlashML-org/FreeToken.md
4. **p-e-w/heretic** — 5.3 · Breakout · Language models · Python · +500 stars measured 2026-11-02, 00:09–23:03 UTC
   A tool for fully automatic removal of censorship (safety alignment) from transformer-based language models using directional ablation and Optuna-based parameter optimization.
   Full card: https://gitnova.dev/en/r/p-e-w/heretic.md
5. **JustVugg/colibri** — 5.1 · Breakout · Language models · C · +703 stars measured 2026-11-02, 00:08–23:03 UTC
   A pure-C, zero-dependency inference engine for running large MoE models (744B–2.8T parameters) on consumer hardware by treating VRAM, RAM, and storage as a single multitier hierarchy and streaming experts from disk.
   Full card: https://gitnova.dev/en/r/JustVugg/colibri.md

---

Magnitude (0–10) measures how fast and how unusually interest in a repository is growing right now. It is not a quality score.
Days are UTC. “So far today” is a fact; “expected by the end of the day” is a forecast.
Summaries and use cases are written by an LLM (DeepSeek V4.1 Flash) from the README and may be inaccurate: verify specific claims (benchmarks, speed, hardware) in the repository itself.
Data as of 2026-11-02 23:16 UTC, updated every 30 minutes.