LongCat Image | Text to Image

What is the LongCat Image text to image ComfyUI workflow?#

LongCat Image text to image is a compact ComfyUI workflow for generating square 1024x1024 images from text prompts. It uses the LongCat-Image diffusion model together with the Qwen 2.5 VL text encoder and the AE VAE, giving you a straightforward prompt-to-image setup for portraits, product shots, and polished editorial-style visuals.

The graph is intentionally simple: choose a square resolution, write your prompt, run the workflow, and save the image. It works well for quick iteration in either English or Chinese prompts, and the included note suggests trying 50 steps if you want to compare the result against the original model setting.

Key features of LongCat Image text to image#

• Square-first generation: Default setup is tuned for 1:1 output at 1024x1024.
• Compact workflow design: The graph stays focused on prompt-to-image generation without extra routing complexity.
• Flexible prompting: Suitable for both English and Chinese text prompts.
• Easy quality tuning: Start from the default 20-step setup, then increase steps when you want slower but more deliberate sampling.

How to use LongCat Image in ComfyUI#

1. Choose the output size
   • Use the Resolution Selector node to keep the default square layout or adjust the target megapixels if needed.
2. Write your prompt
   • Open the Text to Image (LongCat Image) subgraph and replace the default prompt with your own subject, lighting, mood, and composition instructions.
3. Run the workflow
   • Queue the graph to generate a single image from your prompt.
4. Save the result
   • The Save Image node writes the final output once the run is complete.

Tips and settings#

• The current default setup runs at 20 steps with CFG 4.
• If you want to compare against the original recommendation from the source workflow, try 50 steps.
• Clear, concrete prompts tend to work better than broad or abstract prompt fragments in this compact graph.

Resources#

• Workflow source: Comfy.org workflow page
• Official model: meituan-longcat/LongCat-Image on Hugging Face