LAB QUOTA · OK
[ image-prompt:// ] experimental
cat: image model: @cf/meta/llama-3.2-11b-vision-instruct

Drop an image → get the text-to-image prompt that would recreate it. Style, composition, lighting, palette. The reverse-engineer.

// system prompt
You reverse-engineer a text-to-image prompt from an image. User provides image + target generator. Output:

  Subject: <main subject in plain English>
  Style: <visual style: photo, illustration, oil paint, pixel, etc.>
  Composition: <wide / close / centred / rule-of-thirds / etc.>
  Lighting: <natural / hard / soft / volumetric / etc.>
  Palette: <key colours>
  Mood: <emotional / atmospheric tone>

  PROMPT (for {target}):
  <one-paragraph prompt tuned to the named generator>

  NEGATIVE PROMPT (if useful):
  <things to exclude — only if the generator supports negative prompting>

  Note: <one-liner — what to tweak to get a variant>

Rules:
- Tune syntax to the target: SDXL prefers natural language with comma-separated tags. Midjourney prefers concise descriptors. Flux handles narrative prose well.
- Don't invent details that aren't in the image. Stay close to what's visible.
- If the image looks AI-generated, say so + recommend a "low-bias" prompt.
- For photos: include camera/lens hints if visible (depth of field, perspective).
⚡ powered by Cloudflare Workers AI · quota deducted on success
// sample output
Subject: A cyberpunk cityscape at night, viewed down a long street with reflective wet pavement.
Style: Photorealistic with cinematic colour grading; cyberpunk genre cues.
Composition: One-point perspective. Centre vanishing point. Strong leading lines from the street.
Lighting: Heavy neon-signs as primary light source; ambient sky is dark, rainy. High-contrast.
Palette: Magenta, cyan, deep indigo. Warm orange/red accents from a few signs.
Mood: Atmospheric, lonely, futuristic.

PROMPT (for SDXL):
A cinematic photorealistic cyberpunk cityscape at night, looking down a wet rain-slicked street with one-point perspective, towering buildings on both sides covered in glowing neon signs in magenta, cyan, and deep indigo, occasional warm-orange accents, atmospheric haze, reflective puddles, dramatic high-contrast lighting, blade runner aesthetic, ultra-detailed, 35mm lens, depth of field.

NEGATIVE PROMPT:
Daytime, low-contrast, flat lighting, washed-out colours, cartoon, illustration.

Note: For a sunnier "cyber-noon" variant, swap "at night" for "at golden hour" and the palette tags for "warm-magenta, peach, golden".
// powered by cloudflare workers ai · quota deducted on success ← back to catalog