Agent Skills
› leamsigc/ShortsGenerator
› ig-research-transcribe
ig-research-transcribe
GitHub使用 Whisper tiny 模型批量转录项目中的音频文件，跳过已有文本的文件，生成对应的 TXT 字幕。
Trigger Scenarios
需要转录音频文件
运行语音转文字任务
Install
npx skills add leamsigc/ShortsGenerator --skill ig-research-transcribe -g -y
SKILL.md
Frontmatter
{
"name": "ig-research-transcribe",
"version": "1.0.0",
"description": "Transcribe state — runs Whisper on all downloaded audio files.\nPlatform-agnostic: works with audio from Instagram, YouTube, Facebook, etc.\nProcesses only files that don't already have a .txt transcript.\n",
"allowed-tools": [
"Bash",
"Read"
]
}
Social Research — Transcribe State
Transcribes all audio files in the project's transcripts/ directory using OpenAI Whisper (tiny model for speed).
Run
bash "$CLAUDE_SKILL_ROOT/scripts/transcribe.sh" <project-name> [sessionId]
What it does
- Finds all
.m4a,.mp3,.webm,.opusfiles intranscripts/ - Skips files that already have a matching
.txt - Runs Whisper
tinymodel on each file - Saves transcripts as
.txtfiles
Performance
| Model | Speed | Accuracy |
|---|---|---|
| tiny | Fastest | Good |
| small | Medium | Better |
| medium | Slow | Best |
Edit transcribe.sh and change --model tiny to --model small or --model medium for better accuracy.
Output
.ig-research/projects/<project-name>/transcripts/<postId>.txt
Version History
- 7b83166 Current 2026-08-29 02:31
