Ideogram-v4-workflow + Json Prompts — v2.0

Ideogram-v4-workflow + Json Prompts — v2.0 is for turning a structured JSON description of a scene into an image, using high-level description, style, composition, colors, and rendered text as inputs.

The JSON is organized into high_level_description, style_description, and compositional_deconstruction.

The style section can specify aesthetics, lighting, photo, medium, and hexadecimal colors; the composition section can specify the background and elements.

Provide high_level_description, style_description, and compositional_deconstruction as JSON fields.

Each element can be an object or text with a bounding box, description, and hexadecimal color palette; text elements also include the text to render.

It is listed as a Workflows entry, version v2.0, with base model Other.

Required model files listed: flux2-vae.safetensors, ideogram_4_turbotime_v1.safetensors, ideogram4_fp8_scaled.safetensors, and qwen3vl_8b_fp8_scaled.safetensors.

Required node packs listed: cg-use-everywhere, comfyui-detail-daemon, comfyui-impact-pack, comfyui-kjnodes, RES4LYF, and was-ns.

The listed model files are labeled as VAE, LoRA, U-Net, and text-encoder components.

Use Euler instead of res_multistep; the description warns that res_multistep can create ugly noise artifacts.

Vorschlag · nicht geprüft

To control the image description, fill in the scene description, style fields, background, elements, text, bounding boxes, and color palettes.

Vorschlag · nicht geprüft

Before setup, verify that every listed model filename and node-pack name is available in your environment.

Vorschlag · nicht geprüft

No license is recorded, so check usage rights before publishing outputs.

Vorschlag