Abiray / ideogram-4-GGUF

Ideogram 4 GGUF (Low-Resource Optimized)

This repository contains GGUF quantized weights for Ideogram 4 (9.3B parameter DiT), compiled to bring state-of-the-art open-weight design and image generation to highly constrained hardware environments.

These quants are specifically targeted for local execution on systems with strict memory ceilings (such as 8GB system RAM) and environments operating purely on CPU inference without dedicated VRAM.

📦 Available Model Files

Note: Each file in this repository includes a custom EOF bypass structural string ( ideogram4_bypass_[filename]_[timestamp] ) injected during processing for unique deployment tracking.

🚀 Quick Start (stable-diffusion.cpp)

You can run these weights completely locally using stable-diffusion.cpp .

Example Command (CPU Only)

This command utilizes the Q4_K model to generate a 1024x1024 image, strictly offloading to the CPU.

.\bin\Release\sd-cli.exe \
  --diffusion-model ideogram4-Q4_K.gguf \
  --uncond-diffusion-model ideogram4_uncond-Q4_0.gguf \
  --llm ..\..\llm\Qwen3VL-8B-Instruct-Q4_K_M.gguf \
  --vae ..\..\ComfyUI\models\vae\flux2_ae.safetensors \
  --offload-to-cpu \
  -H 1024 -W 1024 \
  -p '{"high_level_description":"A square 1024 x 1024 luxury fashion magazine cover featuring exactly one short chubby fluffy cat...","style_description":{"aesthetics":"luxury fashion magazine cover, high-end pet couture campaign...","lighting":"Soft diffused studio lighting...","photo":"high-resolution fashion editorial photography look...","medium":"mixed media fashion photography","color_palette":["#F4EFE7","#111111","#D8B56D","#B73A3A","#FFFFFF","#8A7A6A"]}}'