ERNIE-Image: 8B Open Source Text-to-Image Model for ComfyUI

ERNIE-Image is Baidu's 8B parameter open text-to-image model built on a single-stream DiT with a light Prompt Enhancer, excelling at text rendering and structured generation.

E

ERNIE-Image

Text-to-Image8B DiTText RenderingPrompt Enhancer

Open text-to-image generation model by Baidu's ERNIE team. Compact 8B DiT paired with a lightweight Prompt Enhancer that expands brief inputs into structured descriptions. Reaches SOTA among open-weight T2I models on complex instruction following, text rendering, and structured image generation for posters, comics, and multi-panel layouts.

Hugging Face (Upstream)HF Comfy-OrgGitHubBlog
DeveloperBaidu (ERNIE Team)
Release Date2025-11
Architecture8B Single-Stream Diffusion Transformer (DiT)
LicenseApache-2.0
Text EncoderMinistral 3B + Prompt Enhancer
VRAMRuns on 24GB consumer GPUs
VersionsERNIE-Image (50-step SFT) + ERNIE-Image-Turbo (8-step distilled)

Guides and workflows related to this model series.

No articles found.