Build an image prompt with a clear scene
Write the image idea in a few compatible parts. A specific subject and scene are easier to evaluate than a request that asks for several unrelated pictures at once.
In this article
A prompt built in four parts
- Define the original adult character and stable appearance.
- Choose one outfit and location.
- Specify the framing and a simple action.
- Add lighting, then inspect options and token cost before submission.
Try an example with room for refinement
A 34-year-old ceramic artist wearing a white work shirt, seated at a studio bench beside a clay bowl, waist-up framing, soft light from a high window.
If the room is too busy, simplify the background in the next request. Keep the subject stable so you can tell what that change accomplished.
Finish one request before duplicating it
Inspect the completed image and its available saving controls. If the request fails after a charge, keep the time and error for support. A repeat submission can be another metered action.
Common questions
How do I make an image prompt easier to follow?
Use one adult subject, a location, composition and compatible lighting. Change the detail that matters instead of replacing the entire scene after each attempt.
Does an image request also buy a video version?
No. Image-to-video is a separate action on an eligible result. Check its own conversion control and token requirement.