3 ready-made prompts
AI Image and Video
Fix the classic failures of image and video models: broken hands, unreadable type, inconsistent characters and copyright risk.
Image model failures are systematic, not random. Hands break because the model learned the texture of a hand rather than finger anatomy; type is unreadable because it sees letters as shapes rather than characters. These prompts do not hide the error — they build a composition that routes around the model's weak spot.
Midjourney Hands Fix: Prompt Structure That Avoids Bad Fingers
Shrinks the hands, moves them out of frame or poses them deliberately, with the right negative terms for six fingers and fused digits.
Intermediate MidjourneyIdeogramAI Image Text Unreadable? Prompt for Legible Type
Two to four words, in quotes, on a flat empty surface, with the type style stated separately. Legible lettering for logos and posters.
Intermediate MidjourneyFluxConsistent Character Prompt: Same Face Across Every Image
Describes the face by immutable features, ties it to a reference image and holds identity constant while the scene changes. For comics, ad series and social content.
AdvancedFrequently asked
About AI Image and Video
Why are hands still broken?
Hands are small, moving and frequently overlapping regions in training data. What the model learned is the texture and shadow pattern of a hand, not finger anatomy. The fix has two parts: shrink the hands in the composition or move them out of frame (in a pocket, behind the back, gloved), and put explicit terms in the negative prompt — extra fingers, fused fingers, six fingers. For work where hands must be visible, do not leave the image model unsupervised.
Is there a way to get clean text in an image?
For short, common words yes; for long sentences no. The most reliable method is to generate the image without text and set the type in a design tool. If you must have it in-image, quote the word, keep it to three to five words, describe the area as a blank sign, and state the type style separately (serif, handwriting, neon).
How do I keep the same character in the second image?
Not with text alone; every generation starts from zero. Use a reference image (image prompt or character reference), describe the face by immutable features (face shape, hairline, eye colour, a mark) and paste the same description block verbatim every time. Lock the wardrobe and the lighting too — the model reinterprets the scene and the character together.
Other categories