Reverse-engineering a picture
The vision model reads the image the way a prompt-writer would. What is the subject, what style is this, where does the light come from. How is it framed, what lens look is at play. And writes it as one dense, reusable prompt.
Why prompts beat guessing
A borrowed aesthetic is faster than a described one. Analyze a reference you love, keep the style words, swap the subject. That is the quickest route to on-brief generations.
From words back to pixels
Generate this sends the prompt straight into the image generator. Create variations opens image-to-image with your reference. Animate this hands it to video. Zero copy-paste.
FAQ
What images can I analyze?
Photos, renders, artwork, screenshots - anything visual. Rights to the analyzed image remain with its owner. Use references you may use.
How accurate is the prompt?
It captures style, light and composition very reliably. Exact identities and tiny text are summarized, not named.
What does it cost?
2 credits per analysis - the cheapest step in the workflow, designed to be run often.
Can it write a ChatGPT photo prompt?
Yes. The prompt it writes is plain text, so it runs in ChatGPT, Midjourney or here. Copy it wherever you need it.