Wanted to share something I've been working on for the past few weeks. It started as a personal tool because I got tired of writing and also remembering the prompt structures and remembering every space in them over and over when using LLMs.
The Problem
Every time I wanted to generate an image, I'd spend minutes for thinking about lighting, camera angles, composition, negative prompts... and then I'd forget half of them. I had a messy Notes pages with "prompt templates" but it was getting out of hand.
What I Built
Prompy is basically a structured form that helps you build detailed prompts. You pick your subject type, and it shows you relevant options - portrait gives you ethnicity/expression/pose options, product photography gives you background/material options, etc.
The cool part (at least for me) is the output formats. You get:
- Natural language prompt (ready to paste into any AI tool)
- JSON format (if you're building something that needs structured data)
- TOON/TONL formats (experimental compact notations I've been playing with)
I also added an "Objects & Relationships" feature because I was frustrated that AI tools often ignore spatial relationships. Now I can specify "person is holding the coffee cup" or "cat is sitting on the windowsill" and it gets included in the prompt structure.
The Tech Stack
Nothing fancy - vanilla JS frontend with TailwindCSS, Python Flask backend, SQLite for storage. I specifically avoided React/Next.js because I wanted something simple that loads fast. The whole thing is under 200KB.
Why Login?
I know some people will ask about this, so let me explain upfront. The login (Google OAuth) is 100% optional. You can use the entire generator without signing in - generate prompts, copy them, use all presets, everything works.
Login is only for:
- Saving your prompts to a personal gallery
- Creating lists to organize prompts by project
- Seeing generated prompts by other users
I added these features because I personally wanted to keep track of what prompts worked well for specific styles. Nothing is sold, no emails are sent, I'm not collecting data for any commercial purpose. The whole thing is open source so you can verify this yourself. I just want to make tights with and between users.
Current State
This is very much a beta. I've been using it myself and a limited circle whom in my Twitter list for a few weeks and it's stable, but there are definitely rough edges. The UI could use some polish, and I'm sure there are bugs which haven't founded yet.
Wanted to share something I've been working on for the past few weeks. It started as a personal tool because I got tired of writing and also remembering the prompt structures and remembering every space in them over and over when using LLMs.
The Problem
Every time I wanted to generate an image, I'd spend minutes for thinking about lighting, camera angles, composition, negative prompts... and then I'd forget half of them. I had a messy Notes pages with "prompt templates" but it was getting out of hand.
What I Built
Prompy is basically a structured form that helps you build detailed prompts. You pick your subject type, and it shows you relevant options - portrait gives you ethnicity/expression/pose options, product photography gives you background/material options, etc.
The cool part (at least for me) is the output formats. You get:
- Natural language prompt (ready to paste into any AI tool)
- JSON format (if you're building something that needs structured data)
- TOON/TONL formats (experimental compact notations I've been playing with)
I also added an "Objects & Relationships" feature because I was frustrated that AI tools often ignore spatial relationships. Now I can specify "person is holding the coffee cup" or "cat is sitting on the windowsill" and it gets included in the prompt structure.
The Tech Stack
Nothing fancy - vanilla JS frontend with TailwindCSS, Python Flask backend, SQLite for storage. I specifically avoided React/Next.js because I wanted something simple that loads fast. The whole thing is under 200KB.
Why Login?
I know some people will ask about this, so let me explain upfront. The login (Google OAuth) is 100% optional. You can use the entire generator without signing in - generate prompts, copy them, use all presets, everything works.
Login is only for:
- Saving your prompts to a personal gallery
- Creating lists to organize prompts by project
- Seeing generated prompts by other users
I added these features because I personally wanted to keep track of what prompts worked well for specific styles. Nothing is sold, no emails are sent, I'm not collecting data for any commercial purpose. The whole thing is open source so you can verify this yourself. I just want to make tights with and between users.
Current State
This is very much a beta. I've been using it myself and a limited circle whom in my Twitter list for a few weeks and it's stable, but there are definitely rough edges. The UI could use some polish, and I'm sure there are bugs which haven't founded yet.
Some things I'm considering adding:
- More preset styles
- API for integrations
- Prompt generating for other purposes
Links
- Live: prompy.xyz
- Source: github.com/niyoseris/photoprompt (MIT license)
Would love to hear feedback.
What features would actually be useful? What's missing that would make you use this over just typing prompts manually?
---
Built this because I needed it. Sharing it because maybe someone else does too.
Have a great day.