Upload a picture and get 3 ready-to-paste prompts that recreate it. The AI describes the subject, setting, style, lighting, camera angle, mood and colours, your browser reads the colour palette and aspect ratio, and the result is written for Midjourney, Stable Diffusion, ChatGPT, Flux, Veo or Sora. Free, no sign-up.
a red vintage scooter, a quiet street lined with old shophouses, potted plants by the doors, long shadows on the pavement, photo, eye-level side view, warm afternoon sunlight, nostalgic mood, red, cream and teal --ar 3:2
Negative prompt:
a red vintage scooter, a quiet street lined with old shophouses, potted plants by the doors, long shadows on the pavement, cinematic film still, eye-level side view, 35mm lens, warm afternoon sunlight, nostalgic mood, red, cream and teal, film grain --ar 3:2 --stylize 150
Negative prompt:
a red vintage scooter, a quiet street lined with old shophouses, potted plants by the doors, long shadows on the pavement, digital painting, eye-level side view, warm afternoon sunlight, nostalgic mood, red, cream and teal, intricate textures --ar 3:2 --stylize 400
Negative prompt:
Advertisement
Example: for a photo of a red scooter outside old shophouses, the Midjourney prompt is a red vintage scooter, a quiet street lined with old shophouses, potted plants by the doors, long shadows on the pavement, photo, eye-level side view, warm afternoon sunlight, nostalgic mood, red, cream and teal --ar 3:2.
--ar.Advertisement
Note: AI descriptions can miss or misread details, so check the fields before you copy. Only upload pictures you have the right to use; the tool never identifies people.
Your browser first reads the picture's main colours and aspect ratio. The picture, shrunk to 560 pixels, is then described by an AI vision model (Llama 3.2 11B Vision) as subject, action, setting, style, lighting, framing, mood and colours. Those fields are written as 3 prompts (as described, cinematic and artistic) in the format of the app you choose.
No. A finished image does not contain its prompt, and different prompts can produce similar pictures. The tool describes what is visible so you can make something close, then refine it. Midjourney's own /describe command works the same way.
No. The picture is shrunk in your browser and sent once to the AI service for a description, which is returned and not saved by this site. If you use the in-browser AI option, the picture never leaves your device at all.
The online description runs on a free daily allowance (10,000 Cloudflare Neurons, reset at 8am Malaysian time). When it runs out or the service is busy, the colours and aspect ratio are still filled in, and you can choose the in-browser AI: a short caption model of about 250 MB downloaded once, then cached.
Yes. Switch to Video, pick Veo, Sora, Kling or Runway, and add a camera movement such as slow dolly in or orbit. For image-to-video apps, Runway advises describing the motion you want rather than describing the picture again.
Midjourney (with --ar for the picture's shape), Stable Diffusion (with a negative prompt), ChatGPT/DALL·E, Flux, Gemini, Leonardo and Canva for images, and Google Veo, Sora, Kling, Runway, Hailuo, Pika and Luma for video. Press Edit in Prompt Generator to fine-tune with style chips.
The AI is told to describe only visible features such as clothing, pose and expression, and never to identify who someone is. Use pictures you have the right to use, and ask before turning a recognisable person's photo into AI art.
Advertisement