JustScribeTopic added Jan 10, 2026
describe prompt: prompt formula and examples
For “describe prompt”, the prompt should turn a supplied image into a faithful, reusable description. Use observable instructions rather than praise words. The working formula is [subject] + [composition] + [lighting] + [materials] + [camera] + [visible text] + [uncertainties]. Generate a short test, inspect the failed criterion, and revise only the corresponding variable.
Example outputs for this workflow
Each page shows one source image and one short sample video so you can quickly check what to expect before running.
Source: openverse_flickr
Five workflows for this question
Quick take
- For “describe prompt”, the prompt should turn a supplied image into a faithful, reusable description.
- Use the same input to judge prompt adherence and subject and text accuracy.
- Keep the page noindex until locale, intent, factual evidence, uniqueness, and Tool-link checks pass.
Prompt formula
For “describe prompt”, the prompt should turn a supplied image into a faithful, reusable description. Use observable instructions rather than praise words. The working formula is [subject] + [composition] + [lighting] + [materials] + [camera] + [visible text] + [uncertainties]. Generate a short test, inspect the failed criterion, and revise only the corresponding variable.
Formula: [subject] + [composition] + [lighting] + [materials] + [camera] + [visible text] + [uncertainties]
Copyable starting prompts
Prompt 1
Describe only what is visible in the supplied image. Identify the main subject, pose or action, foreground/background arrangement, dominant colors, lighting direction, materials, camera angle, and any readable text. Mark uncertain details as uncertain.
Use example 1 as a starting structure.
Replace the subject, action, setting, and constraints with the real request.
Prompt 2
Convert the image into a reusable generation prompt: subject and attributes; composition and crop; setting; lighting; lens and depth of field; materials; color palette; exact visible text. Do not infer a brand, person, location, or event that cannot be verified visually.
Use example 2 as a starting structure.
Replace the subject, action, setting, and constraints with the real request.
Prompt 3
Compare the source with this draft description. Add omitted spatial relationships and edge details, remove invented objects, preserve exact lettering, and separate objective observations from optional style labels.
Use example 3 as a starting structure.
Replace the subject, action, setting, and constraints with the real request.
How to revise a failed prompt
- If the subject changes, strengthen identity/reference constraints.
- If motion or composition is wrong, describe one observable action and one camera instruction.
- If artifacts appear, simplify the scene and reduce simultaneous changes.
What is the first concrete action for “describe prompt”?
Define the exact image output and its pass condition, then run the smallest representative test with Image To Prompt.
Why are five tools shown?
They cover different input and transformation paths. Choose by the job and capability, not by an unsupported universal ranking.
What information must not be guessed?
Do not guess current prices, credit limits, availability, release dates, live outages, benchmark scores, or product features.