A one-person content pipeline with AI

The pipeline

Step	Tool	Human still required for
Research	Perplexity	Deciding what is worth saying
Outline	Claude	The actual argument
Draft	Claude	Opinion, specifics, anything you know first-hand
Edit	Human + Grammarly	Everything — this step does not delegate
Hero image	Midjourney or Ideogram	Art direction, and text in the image
Short video	Runway or Pika	Deciding whether a video adds anything
Voiceover	ElevenLabs	Script, pacing, pronunciation of names

Where it goes wrong

Drafting the whole thing

Generating a complete article from an outline produces text that is correct, readable and worthless — nothing in it could only have come from you. Use the model for structure and for sections you genuinely have nothing specific to add, and write the parts you actually know.

Skipping the edit

The editing pass is where generated drafts become publishable, and it is the one step that cannot be delegated. Budget as much time for it as for drafting.

Generating images with text in them

Still unreliable in every tool, and completely unreliable in Japanese. Generate the image, composite the text afterwards.

What this costs

Claude or ChatGPT at $20, Midjourney at $10, Runway or Pika at $12, ElevenLabs from $5. Around $50 a month covers the whole chain for one person — which is the real story here, not any individual tool.

Will readers notice AI-assisted writing?

They notice unedited AI writing immediately. An edited draft where the substance is yours reads like you, because it is.

Is the video step worth it?

Only if the platform rewards it. Generating video because you can is the fastest way to spend an afternoon on something nobody watches.

A one-person content pipeline — article, images and video with AI