Prompt Engineering
The craft of the prompt — structure, syntax, weighting, and the hard-won tricks that bend image and video models to your will.
Newest posts
- What finally made you switch your go-to image model?
- How do you run a fair comparison between image models?
- Do role lines like "you are an expert" still help image and video prompts?
- What do you get from a single-word prompt?
- Do layout boxes give you more control than prompt wording?
- Have you tried guiding an image model with a marked-up map or sketch?
- Does telling the model it's fine to find nothing stop it from inventing problems?
- How do you tell if a prompt change made things better when you have no labeled data?
- Question: is there any REAL weighting syntax that works in video models, or is it all placebo?
- Do negative prompts actually do anything in video models? I ran a blind test and I'm annoyed
- WIP technique: 'prompt laddering' to reverse-engineer what a model actually understands
- Discussion: 'prompt engineering' as a term is on borrowed time and that's fine
- Natural-language sentences vs keyword/tag style — what actually wins for you in 2026?
- Token order matters more than weighting syntax in most modern models — a demo
- Hottest prompt-engineering debate: are long detailed prompts actually better than short ones?
- Underrated technique: use SYNONYMS to bias a concept instead of weighting syntax
- Technique: how to prompt CAMERA MOVEMENT in models that have no dedicated camera mode
- Resource dump: a structured prompt template that ports across image AND video models
- Which prompt lever gives you the biggest quality jump per unit effort?
- I built a model-agnostic prompt linter that flags dead tokens and contradictions
- Community spreadsheet: tested dead-tokens and working levers per model (contribute yours)
- Deep-dive video: how modern text-to-video models actually interpret your prompt (not marketing)
- PINNED: The prompt-engineering starter guide — read this before asking 'why doesn't my prompt work'