skills.rest
Backtthefuture
BacktthefutureCOMMUNITY
Explore all skills by this author →
GitHub

video-transcript

Transcribe videos and podcasts into timestamped transcripts using local FunASR models.

108|12|Updated Aug 9, 2026
Search Tags:#subtitles#transcription#podcast#asr#funasr#speaker diarization#video to text
One-click install
npx skills add https://github.com/Backtthefuture/video-transcript --skill video-transcript-backtthefuture
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: video-transcript
Source: https://github.com/Backtthefuture/video-transcript
Command: npx skills add https://github.com/Backtthefuture/video-transcript --skill video-transcript-backtthefuture

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill requires yt-dlp, playwright, funasr, torchaudio, and includes scripts (resource) components.

What problem does it solve? Turning online videos and podcast episodes into usable, citable text normally requires manual note-taking or paid cloud transcription APIs. This Skill downloads audio from major platforms and runs ASR transcription entirely on your own machine, producing structured transcripts without any API key. ## Core Features & Use Cases - Multi-platform video transcription: Paste links from Bilibili, Douyin, Xiaohongshu, YouTube, or WeChat Channels and receive a cleaned transcript with semantic section headings and timestamps, powered by the local SenseVoice-Small model. - Podcast speaker diarization: Transcribe Xiaoyuzhou, Ximalaya, or Apple Podcasts episodes with paraformer + CAM++ to automatically separate host and guest, outputting a speaker-blocked Markdown transcript plus an SRT subtitle file. - Local file support and WeChat Channels delivery modes: Transcribe local mp4/m4a/mp3/wav files, and for WeChat Channels choose between an in-chat verbatim transcript, a text-only PDF, or a screenshot-illustrated PDF. - Use Case: Paste a one-hour podcast episode link and receive a speaker-labeled transcript with host and guest names mapped, plus an SRT file ready to burn as subtitles. ## Quick Start Paste a video or podcast link into the chat and ask the agent to transcribe it with the video-transcript skill, for example by saying "transcribe this Bilibili video into a cleaned transcript".

Frequently Asked Questions about video-transcript

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I transcribe a Bilibili or YouTube video to text?▼

Paste the video link and run the transcript.py script with the URL as input. The skill resolves the direct media link, extracts 16kHz audio with ffmpeg, and transcribes it locally with FunASR SenseVoice-Small, producing a sectioned Markdown transcript.

How to transcribe a podcast with speaker separation?▼

Pass a Xiaoyuzhou episode link or add the --speakers flag for other audio sources. The skill uses paraformer with CAM++ speaker embeddings to separate speakers, and you can set real names with --host and --guest flags.

Does video transcription require an API key or cloud service?▼

No API key is needed because ASR inference runs locally on your machine with FunASR models. Only link resolution and the first-time model download require internet access; models are cached locally afterward.

Which platforms are supported for video transcription?▼

Dedicated parsers cover Bilibili, Douyin, Xiaohongshu, YouTube, WeChat Channels, and Xiaoyuzhou episodes. Other sites like Weibo and Zhihu fall back to yt-dlp, while Spotify and Kuaishou are not supported.

Why does WeChat Channels transcription fail with auth errors?▼

Errors like WECHAT_AUTH_REQUIRED mean the local Tencent Yuanbao login session is missing or expired. Run sph_resolver.py --login and scan the QR code once with WeChat to establish the local login state, then retry.

How do I fix wrong speaker names without re-transcribing?▼

Use the --reformat flag with corrected --host and --guest names. It reuses the cached transcription.json and only reruns post-processing in seconds, avoiding the full ASR pass that --force would trigger.

Related Skills

Explore More Content & Communication Skills→
LogarnLogarn
1

transcribe

Transcribe audio and video files to text using a configured speech-to-text provider.

Community
Intermediate
SaberAlouiSaberAloui

podcast-marketing

Repurpose podcast episodes into social posts, show notes, quote cards, and video clips.

Community
Advanced
helix4uhelix4u

yt-dlp

Download YouTube videos, audio, transcripts, and metadata using yt-dlp.

Community
Intermediate
borealBytesborealBytes
7

openai-whisper

Transcribe spoken audio locally using the Whisper CLI without an API key.

Community
Basic