AutoTube: turning novels into narrated videos, end to end

AutoTube is one of my earlier projects, and it’s a good one — a pipeline that takes a novel and turns it into narrated, YouTube-ready videos with almost no manual work. It’s built for a YouTube channel where every video is generated this way.

The pipeline

Give it a link to a novel and it runs end to end:

1. Scrape the novel site and pull out all the chapters.
2. Clean the text — authors leave notes, thank-yous, and other noise, so I strip those with a mix of manual rules and regex.
3. Translate where needed, using Google Translate, to produce consistent English content.
4. Voice it with Edge TTS, which gives natural-sounding narration with multiple accent options.
5. Generate images to go with the narration.
6. Stitch the audio and images into a video with FFmpeg, keeping the audio and visuals in sync.
7. Upload the finished video to YouTube automatically through the YouTube API.

novel URL → scrape chapters → clean (rules + regex) → translate
          → Edge TTS voiceover + images → FFmpeg compose → YouTube upload

What I took away

AutoTube was where I first got comfortable chaining a bunch of independent tools into one automated pipeline — scraping, TTS, image generation, FFmpeg, and a platform API — and handling the messy middle steps, like cleaning scraped text, that decide whether the output is watchable. It’s not glamorous AI, but it reliably produces finished videos from a single link, which is the whole point.

The videos live on the YouTube channel, and the code is on GitHub.