StreamStack: Automating YouTube Shorts with AI
StreamStack is my attempt to fully automate YouTube Shorts production. Script generation, text-to-speech, video rendering with Remotion, and auto-upload to YouTube — all triggered by a single command.
Why Shorts?
YouTube Shorts is the fastest-growing content format on the platform. The algorithm is generous with new creators, the production quality bar is lower than long-form, and the monetization (while small per video) scales with volume. If you can produce 5-10 Shorts per day at near-zero marginal cost, the math starts to work.
The pipeline
Here's what happens when I run streamstack generate:
- Topic selection — pulls from a queue of trending topics in my niche, seeded by RSS feeds and social monitoring
- Script generation — Claude writes a 45-60 second script optimized for vertical video: hook in the first 3 seconds, value in the middle, CTA at the end
- Text-to-speech — ElevenLabs generates the voiceover with a consistent voice profile
- Visual generation — Remotion renders the video with animated text overlays, background footage, and the audio track synced to visual beats
- Thumbnail generation — AI generates a thumbnail with bold text and high contrast
- Upload — YouTube Data API v3 handles the upload with title, description, tags, and scheduling
The Remotion piece
Remotion is the secret weapon here. It lets you write React components that render to video. Each Short is a React composition with:
- Background clip (stock footage or AI-generated)
- Animated text overlays timed to the voiceover
- Progress bar and captions
- Branded intro/outro bumpers
Because it's all code, I can iterate on the template once and every future video gets the update. No After Effects, no Premiere Pro, no manual timeline editing.
TTS quality
ElevenLabs has gotten scary good. The voice I'm using is a custom clone trained on about 30 minutes of my actual speech. It's not perfect — there are occasional weird inflections — but it's good enough that most viewers can't tell it's AI. The cost is about $0.03 per Short, which is negligible.
Current stats
After two weeks of running StreamStack at 3 Shorts per day:
- 42 Shorts published
- 12,000 total views
- 180 subscribers gained
- Best performing Short: 3,200 views
The view counts are modest, but the cost to produce them is essentially zero (once you account for the API costs). And the compounding effect is real — each Short that performs well drives subscribers who see future Shorts.
What's not automated yet
Human review. I watch every Short before it goes live. About 20% get rejected — usually because the TTS stumbled on a word, the visuals don't match the script, or the hook is weak. I'm working on automated quality scoring to reduce this, but for now, the human check takes about 60 seconds per video.
The endgame
StreamStack is a SaaS play. Once the pipeline is solid for my own channels, I'll open it up to other creators. The value prop is simple: you provide the niche and voice, StreamStack handles everything else. Pricing would be per-video, probably $0.50-1.00 per Short generated.
More updates coming as the view counts grow.