← Back to Skills Marketplace
Audio Video To Text
by
ivan830826
· GitHub ↗
· v1.0.0
1024
Downloads
0
Stars
7
Active Installs
1
Versions
Install in OpenClaw
/install audio-video-to-text
Description
音视频转文字技能，使用 Whisper 进行语音识别。支持多种音视频格式，可输出纯文本、SRT/VTT 字幕或 JSON 格式。适用于会议记录、视频字幕生成、采访整理、播客转录等场景。
README (SKILL.md)
音视频转文字
概述
本技能使用 OpenAI Whisper 模型将音频/视频文件转换为文字。支持自动语言检测和多种输出格式。
何时使用
- 会议录音转文字记录
- 视频内容生成字幕（SRT/VTT）
- 采访/播客内容整理
- 语音备忘录转文本
- 多语言视频翻译准备
快速开始
1. 安装依赖
pip install openai-whisper ffmpeg-python
确保系统已安装 ffmpeg：
# Ubuntu/Debian
sudo apt-get install ffmpeg
# macOS
brew install ffmpeg
# Windows
# 从 https://ffmpeg.org/download.html 下载
2. 基本用法
python scripts/transcribe.py \x3C输入文件> [输出文件] [选项]
3. 示例
# 转录 MP4 视频，输出文本
python scripts/transcribe.py meeting.mp4
# 转录音频，输出 SRT 字幕
python scripts/transcribe.py podcast.mp3 podcast.srt --output-format srt
# 指定中文和较小模型（更快）
python scripts/transcribe.py interview.wav --model tiny --language zh
# 输出带时间戳的 JSON
python scripts/transcribe.py video.mp4 result.json --output-format json
命令行选项
| 选项 | 说明 | 默认值 |
|---|---|---|
--model |
模型大小：tiny, base, small, medium, large | base |
--language |
语言代码：zh, en, ja 等 | 自动检测 |
--output-format |
输出格式：txt, srt, vtt, json | txt |
--device |
运行设备：cpu, cuda | cpu |
--keep-audio |
保留临时音频文件 | false |
模型选择指南
| 模型 | 大小 | 速度 | 精度 | 适用场景 |
|---|---|---|---|---|
| tiny | 39M | 最快 | 一般 | 快速测试、短音频 |
| base | 74M | 快 | 良好 | 日常使用 |
| small | 244M | 中等 | 较好 | 正式场合 |
| medium | 769M | 慢 | 很好 | 高精度需求 |
| large | 1550M | 最慢 | 最佳 | 专业转录 |
输出格式说明
TXT（纯文本）
这是转录的完整文本内容，适合阅读和编辑。
SRT（字幕格式）
1
00:00:01,000 --> 00:00:04,000
这是第一句字幕。
2
00:00:04,500 --> 00:00:07,000
这是第二句字幕。
VTT（Web 字幕）
WEBVTT
00:00:01.000 --> 00:00:04.000
这是第一句字幕。
00:00:04.500 --> 00:00:07.000
这是第二句字幕。
JSON（完整数据）
包含分段、时间戳、置信度等完整信息，适合程序处理。
支持的文件格式
音频： MP3, WAV, FLAC, OGG, M4A, AAC
视频： MP4, AVI, MOV, MKV, WEBM, FLV
性能优化建议
- 短音频优先用 tiny/base 模型 - 速度快，精度够用
- 长内容用 CPU - 避免 GPU 内存不足
- 指定语言 - 可提升准确率和速度
- 批量处理 - 脚本可循环调用处理多个文件
常见问题
转录质量不佳
- 尝试更大的模型（small/medium/large）
- 指定正确的语言代码
- 确保音频质量清晰
处理速度慢
- 使用更小的模型（tiny/base）
- 如有 GPU，使用
--device cuda - 缩短音频长度或分段处理
内存不足
- 使用更小的模型
- 将长文件分割后分别处理
- 关闭其他占用内存的程序
脚本
scripts/transcribe.py- 主转录脚本
参考资料
Usage Guidance
This skill appears to do only local transcription with Whisper and ffmpeg. Before installing/running: (1) verify you trust the skill source and the PyPI package name (openai-whisper) you will install, (2) be aware that Whisper will likely download large model files (especially medium/large) which use network bandwidth and disk space and may require substantial RAM/GPU, (3) install ffmpeg from official sources, (4) run the script in a virtual environment or sandbox and inspect the code if you have concerns, and (5) only run it on files you trust (the script spawns ffmpeg as a subprocess and writes a temp audio file under /tmp by default).
Capability Analysis
Type: OpenClaw Skill
Name: audio-video-to-text
Version: 1.0.0
The skill provides a standard utility for transcribing audio and video files using the OpenAI Whisper library. The code in scripts/transcribe.py uses safe subprocess calls to interface with ffmpeg and contains no evidence of data exfiltration, malicious execution, or prompt injection. All behaviors align with the stated purpose of audio-to-text transcription.
Capability Assessment
Purpose & Capability
Name/description (audio/video → text using Whisper) align with the included script and SKILL.md. Required tools (whisper package, ffmpeg) are explainable and necessary for transcription.
Instruction Scope
SKILL.md and the script limit actions to installing dependencies, extracting audio, loading a Whisper model, transcribing, formatting output, and deleting temporary audio. There are no instructions to read unrelated files, access environment secrets, or send data to external endpoints.
Install Mechanism
This is an instruction-only skill (no install spec). The script depends on the openai-whisper and ffmpeg-python packages and a system ffmpeg binary. Note: loading Whisper models will typically download large model weight files from the network the first time they are used, consuming disk and bandwidth.
Credentials
The skill requires no environment variables, credentials, or config paths. It does not access unrelated secrets or other services.
Persistence & Privilege
always:false and default invocation settings. The skill does not attempt to persist or modify other skills or system-wide agent configuration.
How to Use
- Make sure OpenClaw is installed (local or Docker)
- Run the install command in chat:
/install audio-video-to-text - After installation, invoke the skill by name or use
/audio-video-to-text - Provide required inputs per the skill's parameter spec and get structured output
Version History
v1.0.0
Initial release of the audio-video-to-text skill.
- Converts audio/video files to text using OpenAI Whisper.
- Supports multiple formats: txt, SRT, VTT, and JSON.
- Handles various audio/video types: MP3, WAV, MP4, AVI, and more.
- Allows model selection for speed/accuracy trade-offs.
- Suitable for meeting notes, subtitles, interviews, and podcasts.
Metadata
Frequently Asked Questions
What is Audio Video To Text?
音视频转文字技能，使用 Whisper 进行语音识别。支持多种音视频格式，可输出纯文本、SRT/VTT 字幕或 JSON 格式。适用于会议记录、视频字幕生成、采访整理、播客转录等场景。 It is an AI Agent Skill for Claude Code / OpenClaw, with 1024 downloads so far.
How do I install Audio Video To Text?
Run "/install audio-video-to-text" in the OpenClaw or Claude Code chat to install it in one step — no extra setup required.
Is Audio Video To Text free?
Yes, Audio Video To Text is completely free (open-source). You can download, install and use it at no cost.
Which platforms does Audio Video To Text support?
Audio Video To Text is cross-platform and runs anywhere OpenClaw / Claude Code is available (cross-platform).
Who created Audio Video To Text?
It is built and maintained by ivan830826 (@ivan830826); the current version is v1.0.0.
More Skills