GitHub radar
Turn audio into text on your own laptop, for free
Parakeet Redux is a shrunk speech-recognition model that transcribes audio right on your machine, no GPU and no cloud service required. You get text with timestamps for each sentence and word.
It's a model that listens to an audio file and writes down what's said — like Shazam, but instead of a song title you get the full transcript.
When it helps
Useful for podcasts, interviews, lectures — local transcription without sending your audio to someone else's cloud. For very noisy audio (street, construction site), accuracy drops noticeably, so look elsewhere for that.
Pros
- Runs with no GPU, on a laptop's ordinary CPU
- Only 178 MB instead of the original's 1.2 GB
- Transcribes 113x faster than real time
- Accuracy close to the original, even better across 25 languages
Cons
- Noisy recordings produce noticeably more errors than the original model
- Requires a Python package and basic command-line comfort
How to set it up — step by step
- 1Open your AI agent (Claude Code, Codex) in your project folder.
- 2Send the agent the text from the block below.
- 3Give the agent a test audio file, or ask it to find a short sample.
- 4Check the result: the transcript text and timestamps per sentence and word.
- 5If the audio is noisy, double-check the result by hand — the model errs more on noise.
Text for your agent
Copy this and send it to your agent — Claude Code, Codex, any of them:
Set up the local speech-to-text model Parakeet Redux from Moondream: https://huggingface.co/moondream/parakeet-redux — install it via the moondream Python package, run a transcription on a sample audio file and show me the result with timestamps.
A regular laptop, no GPU needed — the model is about 180 MB and runs fast on CPU alone.
Open on Hugging Face▌ More finds