Earngenix Logo
Skip to main content

Glossary · Generation Modes

What Is Text-to-Image in ComfyUI?

By Earngenix Team ·

⚡ Quick Answer

Text-to-image (often shortened to txt2img) is the most basic ComfyUI generation mode. You write a text prompt describing what you want, and the sampler builds an image entirely from random noise guided by that prompt — no starting picture is involved anywhere in the process.

It's the default mode most beginners learn first, and the foundation every other generation mode — image-to-image, inpainting, video — builds on top of.

Where You'll See It

The tell-tale sign of a text-to-image setup is an Empty Latent Image node feeding directly into the KSampler. That node generates a blank canvas of random noise at whatever resolution you set — there's no Load Image node anywhere upstream.

Quick Example

Type "a red fox sitting in snow, photograph" into your positive prompt, connect an Empty Latent Image node at 1024×1024 into the KSampler, and hit Queue Prompt. The model has nothing to start from but that prompt and pure noise.

Generating far outside the resolution your checkpoint was trained on — for example, 512×512 on an SDXL-class model — often causes duplicated subjects or warped anatomy. Stick close to the model's native resolution.

Frequently Asked Questions

No — that’s the defining feature. Text-to-image begins from an Empty Latent Image node filled with random noise, not from any photo or picture you provide.

Match the resolution your checkpoint was trained at — typically 1024×1024 for SDXL-class models and 512×512 for older SD 1.5 models. Generating far outside that range often produces duplicated subjects or distorted anatomy.

Yes. ControlNet plugs into a text-to-image workflow to add structural guidance — a pose or depth map — on top of the prompt, without needing a starting photo of your own.

See It In Action

Ready to generate your first image?

Our text-to-image tutorial covers the full setup, step by step.

Published: 2026-09-17 · Last updated: 2026-09-17

Discussion

Join the discussion

Sign in to leave a comment or reply

💬

No comments yet

Be the first to share your thoughts!