/

/

ideogram4

ideogram4

AD

Über dieses Modell

IDEOGRAM · Basismodelle

Ideogram 4 text to image. ! Currently very unstable and censored ! Quality settings Instead of steps, you can control quality with /slot2: /slot2:turbo -> fast and rough /slot2:default -> default quality and speed /slot2:quality -> best quality, slowest render This workflow supports two ways to produce the JSON the model expects: (default) Model-assisted (LLM): Enter a short idea in as positive prompt and the system prompt asks LLM to write schema-compliant JSON. You won't see the generated JSON prompt, it is handled only internally. (/slot1:0) Raw input: You may use plain natural language, but structured JSON gives the most predictable layout and style control. Incompatible prompt WILL trigger "safety filter", even if prompt is SFW. About the model The model is trained on structured JSON captions (scene summary, style block, background, and optional per-object descriptions with bounding boxes and hex color palettes). Official inference validates prompts against that schema. Guidance uses flow-matching with asymmetric classifier-free guidance (the unconditional pass drops text tokens)—not a separate negative prompt string. Practical strengths: flexible resolution (about 256–2048 px per side), layout via normalized bounding boxes [y_min, x_min, y_max, x_max] on a 0–1000 grid, palette control with hex colors, and strong in-image text rendering when described in JSON. NOTE: If you see "Image blocked by safety filter" it is because of safety training in the model itself and there