
AI Video Prompts, Motion Budgets, and Clip Costs
AI video prompts fail on motion, not description. Vendor duration limits side by side, the cost of a kept clip, and the one-action rule that fixes most of it.
A small shelf of prompts that were actually run, labelled with the model and the reason they work.

AI video prompts fail on motion, not description. Vendor duration limits side by side, the cost of a kept clip, and the one-action rule that fixes most of it.

An ai image prompt changes shape with the image type. The class vocabulary table, the size limits both vendors publish, and why editing beats regenerating.

Most ai prompt examples show the good version only. These show the bad one, the single edit, and what changed, across text, image, and video prompts.

The fun ChatGPT prompts worth keeping double as experiments. A probe ledger, why the personality reads are rigged, and where the fun moved to images.

Funny AI prompts fail in predictable ways. The mechanism table, why straight-faced rendering beats cartoons, and the sign problem nobody warns you about.

Every prompt engineer course on page one teaches the same text syllabus and skips images and video. Here's the coverage table and the syllabus I'd write.

Claude prompts have their own shape. XML structure, examples, the backwards long-document rule, and how I use Claude to write my image and video prompts.

Most chat gpt prompts break the moment your situation differs. Here are reusable templates with the variable slots exposed, and what each swap does to output.

An AI photo prompt is a caption, not an instruction. Here's the slot order, the lighting vocabulary that does the work, and which word causes each failure.

AI prompts don't transfer between models. Here's one brief rewritten three ways, what carries across, and the word-order rule that only applies to images.