25/06/2026 UPDATE - Replaced the models with versions that work with the new native implementation of INT8 in comfyui, for use with the native Load Diffusion Model node, Make sure you update you ComfyUI (Still works with the custom node but using native loads the model a bit faster for me)
Files gathered from https://huggingface.co/bertbobson/Ideogram-4-INT8-ConvRot
This is an unavoidable double quantization due to the release state of Ideogram4.
The FP8 weights were cast to FP32 with the FP8 scales, then downcast to BF16 before being converted to INT8.
For use in ComfyUI with https://github.com/BobJohnson24/ComfyUI-INT8-Fast
Speed is 1.78x faster(2.03s/it) than FP8(3.62s/it) on a 3090, without compile.
4.4-6.2s/it on my 3060 12gb
Description
FAQ
Comments (2)
For a fully clothed woman I got a blocked by filter error. Is there enough interest in Ideogram 4 to have it fine tuned to remove the filter?
Ideogram 4 expects its trained structured JSON prompting; ComfyUI-KJNodes has an Ideogram Prompt Builder that organises into the expected JSON prompt format. Using the prompt builder rarely returns an image blocked by a safety filter. Alternatively, you can ask an LLM of your choice to write a prompt in the expected JSON format, just copy and paste this into it and provide as much detail as you want into your initial prompt and ask it to structure and expand your prompt with the format. Hope this helps :) ( official prompting guide by Ideogram https://github.com/ideogram-oss/ideogram4/blob/main/docs/prompting.md )
Also, using any lora tends to remove the safety filter from happening with Natural Language Prompts
{"high_level_description":"...","style_description":{"aesthetics":"...","lighting":"...","photo":"...","medium":"...","color_palette":["#RRGGBB","#RRGGBB","#RRGGBB"]},"compositional_deconstruction":{"background":"...","elements":[ ... ]}}