Skip to main content
Back to trending

yapsnap

Transcribe any video URL or audio file into plaintext. No GPU. No cloud. One command.

  • ai-tools
  • asr
  • cli
  • cpu
  • diarization
  • speech-recognition
  • speech-to-text
  • subtitles
  • transcription
View on GitHub
Stars
539
Forks
16
+ today
—
Created
5mo

Ranking data as of November 4, 2026 (UTC).

Overview

Transcribe any video URL or audio file into plaintext. No GPU. No cloud. One command.

The project is written in Python and has 16 forks. It was created 5mo ago.

AI-generated summary of the repository README — may contain errors.

What it is

kouhxp/yapsnap is a Python command-line tool that turns a video URL or an audio file into plaintext. Its description stresses that it needs no GPU, no cloud and just one command. It is licensed under Apache-2.0.

Capabilities described

- **CPU-only speed:** the README says a streaming Zipformer transducer (Kroko) runs several times faster than realtime on a laptop, without CUDA. - **Many sources:** YouTube, X, TikTok, Instagram Reels, direct .mp4/.mp3 links, or local files. The README credits yt-dlp for fetching and ffmpeg for decoding. - **Offline after first run:** models download once to a cache, roughly 80 MB per language, according to the README preview. - Topic tags also mention diarization, subtitles, speech-to-text and transcription, though the supplied text does not describe those features in detail.

Getting started

Installation commands and usage syntax were not included in the supplied data. See the repository README for setup steps and the exact command.

Installation
git clone https://github.com/kouhxp/yapsnap.git
cd yapsnap
# see README for setup

Frequently Asked Questions

Does yapsnap need a GPU or cloud service?

No. The description says it needs no GPU and no cloud, and the README says it runs on ordinary CPU cores without CUDA.

What inputs can it transcribe?

Video URLs from YouTube, X, TikTok and Instagram Reels, direct .mp4 or .mp3 links, and local files, according to the README preview.

Does it work offline?

After the first run, yes. The README says models download once to a cache, about 80 MB per language.

What licence and language does it use?

The metadata lists Apache-2.0 as the licence and Python as the language.