CGFlowStudioGUIDES
Back to All Tutorials
Intermediate21 min read

Complete Ideogram 4.0 ComfyUI Workflow Guide: Accurate Text Rendering & Structured JSON Prompts

Step-by-step tutorial on running Ideogram 4.0 (V4) natively in ComfyUI — covering Qwen3-VL-8B + Gemma 4 dual text encoding, in-image typography, and structured JSON layout control.

By CGFlowStudio
•
2026-08-20

Complete Ideogram 4.0 ComfyUI Workflow Guide: Accurate Text Rendering & Structured JSON Prompts

Ideogram 4.0 (Ideogram V4) marks a massive advancement in open-weights image generation. Renowned for its world-class in-image typography rendering, photorealistic detail, and precise spatial layout adherence, Ideogram 4.0 allows local creators to generate complex graphic design assets, poster art, logos, and cinematic photography directly on their own hardware.

What makes Ideogram 4.0 truly unique is its dual prompt architecture: it natively supports both intuitive Natural Language Prompts and highly controllable Structured JSON Prompts.

With official native support integrated into ComfyUI, setting up Ideogram 4.0 locally is straightforward. In this tutorial guide, we cover model installation, directory hierarchy, prompt formatting options, safety filter management, and step-by-step workflow execution.

1. Introduction & Core Capabilities

Before running your first generation, let’s understand why Ideogram 4.0 excels at graphic design and typography:

💡
Why Ideogram 4.0 Leads in Text Rendering: Ideogram 4.0 utilizes a dual text encoder system combining Qwen3-VL-8B and Gemma 4. This enables exact spelling alignment, font style compliance, and spatial placement of rendered text inside images without letter distortion.

Key Advantages

  • Flawless In-Image Typography: Renders long text strings, signs, poster titles, and apparel quotes accurately.
  • Dual Prompt Support: Accept standard natural language or fine-grained structured JSON captions.
  • Hex Color Palette Control: Define exact brand colors via CSS hex codes inside structured prompts.
  • Spatial Bounding Box Layout: Specify bounding box positions for individual subjects and text elements.

2. Getting Started – Loading Official Workflows

ComfyUI includes an official native template for Ideogram 4.0.

1 Update ComfyUI

Update ComfyUI to the latest release to register core Ideogram 4.0 nodes and subgraphs.

2 Open Template Library

Click Templates in the ComfyUI top menu and search for Ideogram v4: Text to Image.

3 Inspect Workflow Setup

The template routes the Ideogram4 Subgraph through the ResolutionSelector node and into SaveImage.

3. Model Installation & Directory Structure

Place all required Ideogram 4.0 checkpoints, dual text encoders, and VAE weights into their corresponding ComfyUI folders:

📂 Ideogram 4.0 Model Component Directory Map ComfyUI/models/
📂 ComfyUI/models/
📁 diffusion_models/
📄 ideogram4_fp8_scaled.safetensors (Primary Diffusion Model)
📄 ideogram4_unconditional_fp8_scaled.safetensors (Unconditional Negative Guidance)
📁 text_encoders/
📄 qwen3vl_8b_fp8_scaled.safetensors (Primary Qwen3-VL-8B Vision Encoder)
📄 gemma4_e4b_it_fp8_scaled.safetensors (Gemma 4 Instruction Encoder)
📁 vae/
📄 flux2-vae.safetensors (High-Resolution VAE)

(Official weights are available from Comfy-Org/Ideogram-4 on Hugging Face).

4. Prompt Formats Explained: Natural Language vs. Structured JSON

Ideogram 4.0 supports two distinct prompting paradigms depending on the level of precision required:

Option A Plain Text

Natural Language Prompting

Standard descriptive paragraph. Ideal for quick artistic exploration, cinematic photography, and general concept renders.

Option B Structured JSON

Structured JSON Captions

Precise JSON block specifying scene summary, style tags, hex color palettes, and bounding box placement for text and subjects.

Structured JSON Prompt Blueprint Example

{
  "scene_summary": "A vintage coffee shop logo poster.",
  "style": ["vintage graphic design", "retrospect typography", "flat illustration"],
  "color_palette": ["#2C1A1D", "#E8C89A", "#D9534F"],
  "text_elements": [
    {
      "text": "ARTISAN COFFEE",
      "font_style": "bold serif vintage",
      "position": "top center"
    }
  ]
}

5. Step-by-Step Execution Guide

Follow these five steps to run Ideogram 4.0 in ComfyUI:

Step 1: Download & Organize Model Weights

Download the FP8 scaled checkpoints from Hugging Face and place them into ComfyUI/models/ as shown in Section 3.

Step 2: Load Official Template Workflow

Open Templates → Image → Ideogram v4: Text to Image in ComfyUI.

Step 3: Enter Your Prompt

Double-click the Ideogram4 Subgraph:

  • Enter a natural language prompt or paste a structured JSON block.
  • For in-image text, wrap quotes clearly: e.g. "A neon sign glowing on a brick wall reading 'NIGHT OWL' in bright cyan letters."

Step 4: Configure Output Resolution

Use the ResolutionSelector node to pick target dimensions (e.g. 1024x1024 for 1:1 square, 1216x832 for landscape).

Step 5: Run & View Output

Click Queue Prompt (Ctrl+Enter). Ideogram 4.0 will encode dual text conditioning and save final high-resolution images to ComfyUI/output/.

6. Managing the Safety Filter

Ideogram 4.0 includes a local safety evaluation filter.

⚠️ Safety Filter Message: "Image blocked by safety filter"

If your render triggers a false positive safety block, note that plain-text natural language prompts have a higher false-positive rate due to keyword ambiguity.

💡 Solution: Switch your prompt to a structured JSON format. Explicitly defining scene summary and object tags eliminates keyword ambiguity and significantly reduces false-positive blocks.

7. Practical Tips & Best Practices

🔤 Quote Formatting for In-Image Text

Always enclose exact words to be rendered inside single quotes or double quotes (e.g., text reading 'BEAST MODE'). This flags the dual text encoder to treat the string as literal typography.

🎨 Brand Hex Color Matching

When designing posters or logos, supply exact CSS hex codes inside the color_palette block of structured JSON prompts for brand color accuracy.

💾 VRAM Considerations (16GB GPUs)

Using FP8 scaled weights (ideogram4_fp8_scaled.safetensors) allows the dual Qwen3-VL-8B + Gemma 4 encoders to run on 16GB GPUs with native memory offloading.

8. Conclusion & Next Steps

Ideogram 4.0 delivers unmatched text rendering accuracy and layout control inside ComfyUI. Whether you use natural language for rapid concept generation or structured JSON for pixel-precise graphic design, Ideogram 4.0 brings professional typography directly to your workstation.

Next Steps to Explore

  • Test generating commercial poster designs using Structured JSON with custom hex color palettes.
  • Combine Ideogram 4.0 typography renders with Krea-2 or FLUX.1 control networks.
  • Browse more image guides on our ComfyHub Tutorials Directory.