> For the complete documentation index, see [llms.txt](https://docs.ideogram.ai/llms.txt). Markdown versions of documentation pages are available by appending `.md` to page URLs; this page is available as [Markdown](https://docs.ideogram.ai/prompting/fix-common-problems.md). # Fix Common Problems What to do when an image shows something you didn't want, ignores part of your prompt, or gets text, framing or style wrong. Each problem below lists the usual cause and the fix. If you're exploring rather than aiming for one exact image, some of these "mistakes" can lead somewhere interesting. ## Something you said you didn't want shows up **Why:** Image models don't handle "no", "without" or "not" well. They latch onto the word that follows. "A man without a beard" can give you a beard. A word elsewhere in the prompt, or a detail Magic Prompt added, can also suggest the thing you're trying to avoid. **Fix:** Describe what you'd see if the thing weren't there. > **Instead of** "no people in the room", **write** "an empty room with chairs neatly arranged" > > **Instead of** "a robot with no eyes", **write** "a robot with a smooth, featureless face" Then check the rest of the prompt for anything that hints at the unwanted item. If Magic Prompt added it, turn [Magic Prompt](/create/image-settings.md#magic-prompt) off; it can also help in the other direction, since its rewrite often turns a negative into a positive. On Ideogram 4.5 and 4.0, rephrasing is the only option. On Ideogram 3.0 you can also list what to avoid in the [Negative prompt](/create/image-settings.md#negative-prompt) setting. ## Extra objects, people or text appear **Why:** A vague or crowded prompt leaves gaps, and Ideogram fills them. Magic Prompt can add details too. **Fix:** Cut ideas you don't need and describe the main subject more fully. Use positive phrasing ("an empty field" rather than "no one around"). Turn Magic Prompt off if it keeps adding things. ## Text is misspelled, broken or missing **Why:** The text isn't marked clearly, there's too much of it, or the scene around it is busy. Non-Latin scripts, such as Arabic or Chinese, render less reliably than English. **Fix:** Put the exact words in quotes, mention them early, and keep them short. Fix a wrong letter afterward with **Select area** in [Studio](/edit/studio.md) (paid plans), or add long copy as a [text layer](/edit/text-and-layers.md) in Studio. Keep non-Latin text to a word or two. See [Text in Images](/prompting/text-in-images.md). ## Faces, hands or limbs look distorted **Why:** When a person is small in the frame, there are too few pixels for fine detail. This is a limit of the model more than of your wording. **Fix:** Move closer with framing words such as "close-up" or "portrait", and name the details you care about (hands, eyes). If you like the rest of the image, use **Select area** in Studio (paid plans) on the flawed part instead of starting over. ## The subject is cropped, or too far away **Why:** Ideogram fills the whole frame, so the [aspect ratio](/create/image-settings.md#aspect-ratio-and-size) shapes the framing. "A woman walking on a busy city street sidewalk" may show her head to toe at 1:2 and from the waist up at 2:1. **Fix:** * Name the framing: "full body", "head and shoulders", "wide view". * For a full figure, describe what's near the feet (shoes, the sidewalk). For a close-up, describe the face and upper body only. * Make the setting the main subject for a wider shot; make the person the main subject for a tighter one. * Pick an aspect ratio that suits the subject, or use **Reframe** or **Extend** (paid) in Studio on an image you already have. * For exact placement on 4.5 or 4.0, use a [JSON prompt](/prompting/json-prompting.md). ## Ideogram ignores part of your prompt **Why:** The detail comes late in a long prompt, gets little weight, or conflicts with something else. **Fix:** Move it earlier, describe it in more detail, or mention it more than once. See [Emphasize what matters](/prompting/refine-and-iterate.md#emphasize-what-matters). ## The prompt contradicts itself **Why:** Conflicting details, such as "a minimalist sculpture with fine and intricate details", force Ideogram to pick one or blend both badly. **Fix:** Choose one direction. > **Instead of** "a clean, empty room cluttered with artifacts", **try** "a clean, empty room with plain white walls and a single wooden chair" **or** "a room filled with ancient artifacts on simple white pedestals" ## The style or mood is off **Why:** Words like "beautiful", "cool", "artistic" or "modern" don't point to anything Ideogram can draw. **Fix:** Name a medium, technique or movement, and describe what you'd see. > **Instead of** "a modern painting of a landscape", **try** "an impressionist painting of a rolling countryside with thick brushstrokes and pastel tones" > > **Instead of** "a beautiful dress", **try** "a red satin evening gown with intricate lace details" On 4.5 and 4.0, set style and color in the prompt; you can give exact hex colors in a [JSON prompt](/prompting/json-prompting.md). On Ideogram 3.0 you can also use style presets, style reference images and color palettes. ## The emotion or idea doesn't come through **Why:** Abstract ideas like "hope" or "regret" have no single look, so results vary widely. **Fix:** Tie the idea to something visible: a pose, an expression, an object, a place. > **Instead of** "a symbol of hope", **try** "a single flower blooming through a crack in the concrete" > > **Instead of** "an old man lost in regret", **try** "an old man sitting alone on a park bench, staring down at a faded photo in his hands" ## A word or phrase isn't working **Why:** Some words aren't tied to a clear look, or other words nearby pull against them. **Fix:** Try a more visual synonym, or change the related words together. See [Try other words](/prompting/refine-and-iterate.md#try-other-words). *** If you're still stuck, strip the prompt back to its core and rebuild it one detail at a time, as in [Build up in steps](/prompting/refine-and-iterate.md#build-up-in-steps).