视频字幕自动生成器——免费的才是最好的
v2.1.1自动提取视频音频，识别生成带时间戳的文字稿，输出SRT/VTT字幕及带字幕的视频，并智能提炼视频标题。
⭐ 1· 173·1 current·1 all-time
MIT-0
Download zip
LicenseMIT-0 · Free to use, modify, and redistribute. No attribution required.
Security Scan
OpenClaw
Benign
high confidencePurpose & Capability
The name/description (video→subtitles→burn-in→title extraction) align with the included scripts and SKILL.md. The code uses FFmpeg and Whisper/faster-whisper which are appropriate. Minor inconsistency: SKILL.md and README mention several module files (speech_recognition.py, subtitle_generator.py, title_extractor.py, video_renderer.py) in the example file tree that are not present in the provided manifest — only video_processor.py and subtitle_processor.py are included. Documentation also lists optional features (yt-dlp, stable-diffusion) that are not required by the included scripts. These are likely incomplete/overdocumented rather than malicious.
Instruction Scope
Runtime instructions and the scripts instruct extracting audio, running ASR (faster-whisper or simulated mode), generating SRT/VTT, optionally converting Traditional→Simplified, and calling ffmpeg to burn subtitles — all within the stated purpose. The code reads user-provided video/transcript/style files and writes output files; it does not reference unrelated system paths or unexpected remote endpoints. Documentation suggests optionally setting HF_ENDPOINT for model downloads, but the runtime actions that would reach networks are limited to downloading model files (expected for Whisper).
Install Mechanism
No install spec in the registry (instruction-only). Dependencies are installed via pip/OS package manager according to docs (faster-whisper, openai-whisper, ffmpeg, optionally yt-dlp, stable-diffusion). No arbitrary binary downloads or obscure URLs were found in the provided files. Model weights will be downloaded by the ASR library (faster-whisper/Whisper) from standard model hosting (Hugging Face) which is expected but may fetch large files.
Credentials
The skill declares no required environment variables or credentials (good). Documentation mentions HF_ENDPOINT as an optional mirror variable for model downloads; this is optional and not required. There are no requests for unrelated secrets or cloud credentials. Be aware that the ASR library will perform network downloads for model weights and may use the user's home cache directories (e.g., ~/.cache/huggingface/hub).
Persistence & Privilege
The skill does not request persistent/always-on privileges, does not declare always:true, and does not attempt to modify other skills or system-wide agent settings. It runs as invoked and writes output into user-specified output directories only.
Assessment
This skill appears internally consistent for generating subtitles: it uses FFmpeg and Whisper/faster-whisper (expected). Before installing/running: 1) Review the two included scripts yourself (they are the main runtime) and run them on non-sensitive sample videos first. 2) Expect large model downloads (tens to hundreds of MB or more) when using real ASR models — this requires network access and disk space in your user cache (e.g., ~/.cache/huggingface). 3) The docs reference extra modules and optional features (yt-dlp, stable-diffusion) that are NOT present in the package; treat those as optional planned features. 4) No credentials are requested, and the code does not contain obvious exfiltration; nonetheless run it in a sandbox or isolated environment if you are concerned. 5) If you only want offline/safe testing, use the simulated/mock mode (the scripts include a mock transcript path) rather than downloading models.Like a lobster shell, security has layers — review code before you run it.
ffmpegvk97az084qd3d9cmskny99y7d2s83gv2vlatestvk97az084qd3d9cmskny99y7d2s83gv2vspeech-to-textvk97az084qd3d9cmskny99y7d2s83gv2vsubtitlevk97az084qd3d9cmskny99y7d2s83gv2vvideovk97az084qd3d9cmskny99y7d2s83gv2vwhispervk97az084qd3d9cmskny99y7d2s83gv2v
License
MIT-0
Free to use, modify, and redistribute. No attribution required.