predict-woo

predict-woo / qwen3-asr.cpp

Public

Implementation of Qwen3-ASR-0.6B in GGML

38
10
100% credibility
Found Feb 11, 2026 at 19 stars 2x -- GitGems finds repos before they trend. Get early access to the next one.
Sign Up Free
AI Analysis
C++
AI Summary

A fast program that converts spoken audio into text transcripts or aligns known text to exact word timings in many languages.

How It Works

1
🔍 Discover the speech-to-text helper

You find a friendly tool that turns any audio recording into written words, perfect for family stories or meetings in 30+ languages.

2
📥 Get the program ready

Download the simple app made especially fast for Apple computers and set it up with a quick launch.

3
🧠 Add language smarts

Grab the free thinking files for the languages you need, like English or Korean, and place them in a folder.

4
🎵 Pick your audio

Choose a voice recording, like a podcast or video, and make sure it's in a standard sound format.

5
Choose your magic
📝
Transcribe to text

Turn talking into readable writing instantly.

⏱️
Time the words

Match your known text to when each word is spoken.

6
✨ Watch it work

Hit go and see it listen closely, think smart, and create your results in seconds.

✅ Enjoy perfect results

You now have clear text or timed words to read, share, or subtitle videos with.

Sign up to see the full architecture

5 more

Sign Up Free

Star Growth

See how this repo grew from 19 to 38 stars Sign Up Free
Repurpose This Repo

Repurpose is a Pro feature

Generate ready-to-use prompts for X threads, LinkedIn posts, blog posts, YouTube scripts, and more -- with full repo context baked in.

Unlock Repurpose
AI-Generated Review

What is qwen3-asr.cpp?

qwen3-asr.cpp is a C++ implementation of Alibaba's Qwen3-ASR-0.6B model using GGML, delivering fast, local automatic speech recognition and forced alignment for audio files. Developers get a CLI tool that transcribes 30+ languages, aligns reference text to word-level timestamps, or runs both in one pipeline—no Python needed. Like asr.cpp or llama implementation github projects, it processes WAV files into text or JSON timestamps optimized for Apple Silicon.

Why is it gaining traction?

It stands out with Metal GPU acceleration and vDSP mel spectrogram computation, hitting 45x speedup on Apple hardware for real-time inference (5s for 92s Korean audio on M2 Pro). Quantized Q8_0 models shrink to 1.3GB with minimal quality loss, plus flash attention for 3.7x faster decoding. The pure C++17 binary deploys anywhere, beating Python-heavy alternatives for edge devices.

Who should use this?

Audio engineers building offline transcription apps, podcast producers needing precise word timestamps, or mobile devs on macOS/iOS targeting multilingual voice interfaces. Ideal for prototyping ASR pipelines where latency matters, like real-time captioning or non-English subtitle generation.

Verdict

Grab it for local Qwen3-ASR-0.6B if you're on Apple Silicon—docs and benchmarks are solid, CLI just works. At 19 stars and 1.0% credibility, it's early but promising; test thoroughly before production.

(198 words)

Sign up to read the full AI review Sign Up Free

Similar repos coming soon.