Text to image on the deep-imager AI image generator works when the brief is a scene, not a slogan.
Subject, light, ratio, negative — then spend the credit.
What belongs in the first sentence of a text to image prompt?
The subject and the frame: who or what, and 9:16 / 1:1 / 16:9.
Style words come second. If style leads, the model may invent a different subject.
How do you keep a second run from drifting?
Reuse the paragraph and change one clause, or switch to image to image with the first still as reference.
The workspace keeps the prompt when you change models. Use that instead of a blank box.
When is a negative phrase worth the tokens?
When a known failure repeats: extra fingers, watermark, cropped logo.
Do not paste a 40-word ban list on the first credit. Add one ban after you see the miss.
How many credits should prompt practice take?
Two or three: a base text to image, one edit, one image to image. 5 free credits cover that drill.
Plans start at $1/month for 10 credits. Expect 30 seconds–2 minutes per still.
Start in text to image or image to image, pick a model such as GPT-4o image, then check pricing from $1/month and how credits work before a long run.