An easy to use Ideogram 4 workflow with KJ Prompt builder. It also includes chunk feed forward and flash attention nodes. You're able to set scheduler and sampler in the workflow and steps. Settings already set for...
Runtime profile
Source description
An easy to use Ideogram 4 workflow with KJ Prompt builder. It also includes chunk feed forward and flash attention nodes. You're able to set scheduler and sampler in the workflow and steps. Settings already set for your first gen then go from there.
Flash attention nodes are bypassed by default if you have FA installed you can just enable them and run.
Estimated VRAM requirement
Estimate unavailable
15.3 GB across 1 of 5 model files. Model file total + 25% loading overhead + 2 GB execution buffer, rounded up.
Requirements
AI-generated commentary
AI-generated explanation based on source and configuration details. Suggestions are clearly labeled.
Use Ideogram 4 Easy Workflow with KJ Prompt Builder + VRAM Optimizations to build prompts and configure scheduler, sampler, and steps for Ideogram 4 image generation. It accepts those settings and loaded images, then provides preview and saved-image outputs.
Its 28 nodes cover prompt building, scheduling, sampling, image loading and blending, previewing, and saving.
Chunk feed-forward and flash attention nodes are included.
Use the KJ Prompt builder, scheduler, sampler, and steps as the main controls.
The image path includes LoadImage and ImageBlend for inputs, then PreviewImage and SaveImage for outputs.
It is published by Urabewe and available from its Civitai page.
Named model files are flux2-vae.safetensors, ideogram4_fp8_scaled.safetensors, ideogram4_unconditional_fp8_scaled.safetensors, qwen_3_8b.safetensors, and qwen3vl_8b_fp8_scaled.safetensors.
The workflow also names the comfyui-kjnodes node pack.
qwen_3_8b.safetensors has a verified checksum, and comfyui-kjnodes is pinned to 3df15ed824045ca45ba668869965a72c7d9d9825.
Settings are already provided for a first generation; adjust the scheduler, sampler, and steps afterward.
Flash attention nodes are bypassed by default; if Flash Attention is installed, you can enable them and run.
Before use, confirm that all five named model files are available.
Suggestion · not verified
The supplied record does not list a license.
Does this need to be edited?
Sign in to send an edit request.
Sources
1 sourceSource excerpts
3 excerptsSource context: 356 downloads · Type Workflows · Base model Other
VAE · Hugging Face · Comfy-Org/flux2-dev
ideogram4_fp8_scaled.safetensors
Not resolvedCheckpoint · Hugging Face · Comfy-Org/Ideogram-4
ideogram4_unconditional_fp8_scaled.safetensors
Not resolvedCheckpoint · Hugging Face · Comfy-Org/Ideogram-4
Text encoder · 15.3 GB · SAFETENSORS · Unknown
qwen3vl_8b_fp8_scaled.safetensors
Not resolvedText encoder · Hugging Face · Comfy-Org/Qwen3-VL
Node pack · Registry
Source context: 355 downloads · Type Workflows · Base model Other
Source context: 354 downloads · Type Workflows · Base model Other