Video Auto Creator

Problem It Addresses

When reworking a video URL into new material, the bottleneck is often not editing but turning the source into reusable structured artifacts: title, description, tags, key frames, script, voiceover, subtitles, and timing are usually scattered across manual steps. video-auto-creator targets that workflow by chaining “read video → sample frames → generate script → TTS voiceover → subtitle alignment → composition” into a debuggable pipeline. It is suitable for promo shorts, content summaries, demo videos, or tutorial narration.

How It Works

• Video analysis: opens Bilibili, YouTube, or Douyin pages, extracts metadata, and saves JSON as the input for scene planning.
• Frame sampling: captures stable frames by --interval or count, then identifies key scenes and visual assets.
• Script generation: outputs scene descriptions, shot copy, voiceover text, subtitle text, and optional Midjourney / Stable Diffusion prompts.
• Voiceover and subtitles: uses edge-tts for multilingual speech and produces SRT subtitles aligned to audio duration.
• Composition: uses FFmpeg to combine image sequences, audio, and burned-in subtitles into H.264/AAC output at 720p/1080p.

It depends on browser control, edge-tts, and FFmpeg; Windows setups may need extra handling for encoding and Chinese file paths. Check copyright, platform scraping limits, and TTS licensing before use. It fits semi-automated production rather than unchecked bulk distribution.