yt2bb - YouTube to Bilibili Video Repurposing

A Claude Code skill that repurposes YouTube videos for Bilibili with bilingual (EN/ZH) hardcoded subtitles.

Compatible with Claude Code, OpenClaw, Hermes Agent, Pi (pi-mono), and indexed by SkillsMP.

Workflow

YouTube → yt-dlp → whisper → validate → translate → merge → ffmpeg → publish_info → Bilibili

Step	Tool	Output
Download	`yt-dlp`	`.mp4`
Transcribe	`whisper`	`_{lang}.srt`
Validate/Fix	`srt_utils.py`	`_{lang}.srt` (fixed)
Translate	AI	`_zh.srt`
Merge	`srt_utils.py`	`_bilingual.srt`
Burn	`ffmpeg`	`_bilingual.mp4`
Publish Info	AI	`publish_info.md`

Usage

/yt2bb https://www.youtube.com/watch?v=VIDEO_ID

Installation

Claude Code

git clone https://github.com/Agents365-ai/yt2bb.git ~/.claude/skills/yt2bb

OpenClaw

git clone https://github.com/Agents365-ai/yt2bb.git ~/.openclaw/skills/yt2bb

Hermes Agent

git clone https://github.com/Agents365-ai/yt2bb.git ~/.hermes/skills/media/yt2bb

Pi (pi-mono)

git clone https://github.com/Agents365-ai/yt2bb.git ~/.pi/agent/skills/yt2bb

Prerequisites

Python 3
ffmpeg
yt-dlp
openai-whisper
YouTube account logged in via Chrome browser (yt-dlp extracts cookies automatically)

Utility Script

# Detect platform and recommend whisper backend + model
python3 srt_utils.py check-whisper

# Merge EN and ZH subtitles
python3 srt_utils.py merge en.srt zh.srt output.srt

# Validate timing issues
python3 srt_utils.py validate input.srt

# Lint against Netflix Timed Text Style Guide (CPS, duration, line length, gaps)
python3 srt_utils.py lint bilingual.srt

# Fix timing overlaps
python3 srt_utils.py fix input.srt output.srt

# Convert to styled ASS (presets: netflix, clean, glow)
# Presets stay bottom-aligned and scale with resolution
# `netflix` = broadcast-grade: white text, thin outline, soft shadow, no box
python3 srt_utils.py to_ass bilingual.srt bilingual.ass --preset netflix
python3 srt_utils.py to_ass bilingual.srt bilingual.ass --style-file custom.ass

# Generate slug from title
python3 srt_utils.py slugify "Video Title"

All subcommands support --format json for structured agent-friendly output. merge and to_ass support --dry-run to validate inputs without writing files.

Subtitle Preset Previews

All three presets rendered on the same 1920×1080 background, so you can compare typography, layout, and contrast at a glance.

Preset	Preview	Use case
`netflix`		Default for professional content. Pure white text, thin black outline, soft drop shadow, no box. Modeled on the Netflix Timed Text Style Guide. Best for documentaries, interviews, and long-form video.
`clean`		Readability safety net. Golden-yellow text on a semi-transparent gray box. Use when `netflix`’s outline could get visually lost on busy or bright-heavy footage — the box guarantees a contrast pad.
`glow`		Entertainment / vlog. Yellow ZH + white EN with a colored outer glow. Most eye-catching, least subtle — best for high-energy edits and B站-style content.

To regenerate these images after changing a preset, run:

bash docs/presets/render_previews.sh

The script renders each preset against a neutral gradient background using the committed docs/presets/sample.srt fixture and writes docs/presets/{preset}.png.

License

MIT License

Support

If this project helps you, consider supporting the author:

WeChat Pay

Alipay

Buy Me a Coffee

探索未至之境