Most "AI does your content" promises fail the same way: it's a demo, not a system. This is the opposite — real runs of my own pipeline, with the costs and the limits in plain view. Not a client case, but studio work that shows how I build and run AI content systems.
From text to video (cloned voice)
One blog post goes through the video pipeline and comes out as a short video. Measured from one real run:
- 8 scenes from one source post
- 8 AI images (Gemini), brand-consistent via a fixed style contract
- ~1,250 characters of text-to-speech in my cloned voice
- 1 script call (gpt-5.2) that splits the text into scenes
- 3 languages (NL/EN/DE) from the same source; language-independent images are shared via caching
The output is live under my blog posts. See a working example on AI experiments.
From theme to social posts (with governance)
The same discipline on the distribution side. One real run of the campaign system:
- 1 AI call (
claude-sonnet-4-6) → 2 LinkedIn posts with hashtags and CTA - automatically in an approval flow (draft → approved → publish)
- 274 input / 1,038 output tokens → $0.0164 (~1.6 cents) for 2 posts
- cost recorded per AI call; publishing defaults to dry-run (no unwanted posts)
The costs, in view
No "AI is free" illusion. Every call is traceable and usage is measured per run, so cost scales with your own rates instead of being a black box. Per-call cost tracking isn't a detail — it's what makes AI accountable in production.
What this proves
Not "I have a tool". Rather: I build AI content systems that run the whole chain — generation, governance, cost, multilingual, publishing — and that actually run. The same approach I apply in client work around AI in production.
An honest note: these are two real, measured runs — the video from a blog post, the social posts from the campaign system — not one seamless chain on the exact same input.