agentleFS
lgldlklgldlk/lgldlk-agent-skills/skills/video-to-text/SKILL.md

video-to-text

Skill3 starsChanged 5 months agoMITView on GitHub

Transcribe local video/audio files or public media download URLs with Tencent Cloud ASR. Use for speech-to-text, timestamped transcripts, subtitles, speaker separation, or reusable MP4/MOV/M4V/WAV/MP3 transcription without hardcoding API keys.

5 other projects use a version of this.

Adopted by

Most recently taken up by

The file

We split this skill into the rules it gives an agent

10 rules. About 0 are common, 3 are distinctive, and 7 are very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈3

    Distinctive · estimate

    This author's own choices. Worth a look.

  • 7

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  • could be automatic: a setting could enforce it
  1. Very specific:

    Default model: 16k_zh, suitable for ordinary Mandarin content.

    line 19
  2. Very specific:

    Default result format: 3, with punctuation-based segments and timestamps suitable for subtitles.

    line 21
  3. Very specific:

    Use --speaker-diarization for supported models and --sentence-max-length 20 to control subtitle line length.

    line 23
  4. Very specific:

    Use --hotword-list '产品名|10,人名|8' for request-scoped vocabulary, or --hotword-id for a saved Tencent hotword table.

    line 24
  5. Very specific:

    Read configuration when credentials, defaults, or environment variables need to be set.

    line 38
  6. Very specific:

    When the source is a Douyin, Xiaohongshu, Bilibili, X/Twitter, or other share page, parse it first with social-content-parser.

    line 42
  7. Very specific:

    Prefer parse_social.mjs --transcribe so the parser selects the first normalized video.

    line 48
  8. Distinctive:

    Do not print media URLs because signed URLs may contain credentials.

    line 55
  9. Distinctive:

    Do not start a paid transcription without the user's media and configured Tencent credentials.

    line 56
  10. Distinctive:

    If credentials are missing, ask for a private config or environment setup.

    line 58

≈ “Common” is an estimate: at least 1 in 5 of the 364,980 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
SKILL.md

Video To Text

Transcribe local media or a public media URL with scripts/transcribe_media.py. Keep credentials outside the skill in a private config file or environment variables.

Choose The Input Mode

  • Local file: use --input. Files up to 5 MB are uploaded directly; larger files are transcoded to mono 16 kHz AAC when ffmpeg is available. If the result still exceeds 5 MB, use URL mode.
  • URL: use --url with an HTTP(S) media address that Tencent Cloud can download without cookies or custom request headers. Signed COS URLs are supported while valid.

A video webpage such as a normal v.qq.com, YouTube, or Bilibili page is not a media download URL. Do not pass webpage URLs to Tencent ASR. Obtain an authorized downloadable media URL or local file first; do not bypass access controls.

Tencent Cloud ASR

  • Default model: 16k_zh, suitable for ordinary Mandarin content.
  • For noisy, dialect-heavy, or mixed Chinese/English recordings, consider --engine-model-type 16k_zh_en_2.0; it uses a different billing tier.
  • Default result format: 3, with punctuation-based segments and timestamps suitable for subtitles.
  • URL recordings must be no longer than five hours and must remain downloadable while Tencent processes the task.
  • Use --speaker-diarization for supported models and --sentence-max-length 20 to control subtitle line length.
  • Use --hotword-list '产品名|10,人名|8' for request-scoped vocabulary, or --hotword-id for a saved Tencent hotword table.

Run

python3 scripts/transcribe_media.py --input ./video.mp4 --output ./transcript.txt
python3 scripts/transcribe_media.py --url 'https://example.com/video.mp4' --output ./subtitle.srt --sentence-max-length 20
python3 scripts/transcribe_media.py --url 'https://example.com/audio.m4a' --output ./transcript.md --speaker-diarization

The output format is inferred from .txt, .json, .srt, .md, or .markdown. Override it with --output-format. JSON preserves the full Tencent response, including word timing and speaker data.

Configuration

Read configuration when credentials, defaults, or environment variables need to be set. Copy config.example.json to an ignored user or project config; never commit real credentials. Keep reusable Skill code separate from ignored project-local or user-local configuration.

Companion: social-content-parser

When the source is a Douyin, Xiaohongshu, Bilibili, X/Twitter, or other share page, parse it first with social-content-parser. Tencent ASR needs a direct media file URL or a local file, not the share page.

Recommended order:

  1. Parse the share URL and inspect normalized.json.
  2. Transcribe only when the user explicitly asks.
  3. Prefer parse_social.mjs --transcribe so the parser selects the first normalized video.
  4. Or call this script on normalized.primaryVideo / a downloaded media/video.* file.

Install the two skills as siblings so the parser can discover scripts/transcribe_media.py automatically.

Operational Boundaries

  • Do not print media URLs because signed URLs may contain credentials.
  • Do not start a paid transcription without the user's media and configured Tencent credentials.
  • Polling can take minutes. Results are written only after Tencent reports success.
  • If credentials are missing, ask for a private config or environment setup; never place secrets in the skill directory.

Download now (ZIP)

Does it work?

Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.

No reviews yet

People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.

  • It worked
  • Partly
  • It didn't
What did you use it for? e.g. Reviewing pull requests in a Rails app

Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.

Your trail

Files you open will collect here, so you can find them again.

Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.

Now do this for your own.

Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.