Hi Product Hunt 👋

I built Prompset because I kept seeing great AI images and videos with no idea what prompt was behind them. And when I did find the prompt, it was usually written for a different model.

What it does
Upload an image (or a clip of up to 15 seconds), pick the model you'll generate with, and you get a prompt written for that model:

10 image models: Nano Banana Pro, GPT Image 2.5, Midjourney V8.2, Seedream 5.0, FLUX.2, Grok Imagine and more
10 video models: Seedance 2.5, Kling 4.0, Veo 3.1, Wan 3.0, Hailuo 3.0 and more
Each prompt follows that model's official prompting guide, so the same image gets a different prompt for Midjourney (with --ar) than for Nano Banana (plain sentences)

Free, no sign-up to start
5 images a day without an account, 10 with a free one, and 1 video a day. Your image isn't stored or used for training. For videos, only a few still frames leave your device.

The library behind it
Prompset started as a free library of 38,000+ real prompts, each shown next to the image or video it made and the model it was made with. Image to prompt works the other way round: you bring the result, it writes the prompt.

Honest limits

It won't copy an image pixel for pixel. The model draws its own version, so treat the prompt as a starting point.
Video reads frames, not sound, so add dialogue or music yourself.
It refuses explicit or suggestive images and anything showing people who may be under 18.

I'd love your feedback: which model should we add next, and where did a prompt miss? I'll be around all day answering comments.

Hi Product Hunt 👋

I built Prompset because I kept seeing great AI images and videos with no idea what prompt was behind them. And when I did find the prompt, it was usually written for a different model.

What it does
Upload an image (or a clip of up to 15 seconds), pick the model you'll generate with, and you get a prompt written for that model:

10 image models: Nano Banana Pro, GPT Image 2.5, Midjourney V8.2, Seedream 5.0, FLUX.2, Grok Imagine and more
10 video models: Seedance 2.5, Kling 4.0, Veo 3.1, Wan 3.0, Hailuo 3.0 and more
Each prompt follows that model's official prompting guide, so the same image gets a different prompt for Midjourney (with --ar) than for Nano Banana (plain sentences)

Honest limits

It won't copy an image pixel for pixel. The model draws its own version, so treat the prompt as a starting point.
Video reads frames, not sound, so add dialogue or music yourself.
It refuses explicit or suggestive images and anything showing people who may be under 18.

I'd love your feedback: which model should we add next, and where did a prompt miss? I'll be around all day answering comments.

Prompset

Turn any image or video into a prompt for your AI model

Turn any image or video into a prompt for your AI model