A free, Apache-licensed image model for posters and infographics that fits on a 24GB consumer card, so hobbyists with decent hardware can self-host it.
Kenning's take
What it does
ERNIE-Image is a free, open text-to-image model from Baidu with 8 billion parameters, small enough to run on a 24GB consumer graphics card. It is aimed at images that need accurate text and layout, such as posters, infographics and comic panels. A faster Turbo version is also published, both under the Apache 2.0 licence on Hugging Face.
- 8B diffusion transformer
- Accurate long text rendering
- Posters, comics and multi-panel layouts
- Runs on 24GB consumer GPUs
- Turbo version for faster generation
- Apache 2.0 licence
Who it suits. Developers and designers who want a self-hosted image model that handles text well.
Key facts
- Category
- AI models
- Made by
- Baidu
- Based in
- China
- Released
- 15 April 2026
- How to get it
- Download weights from Hugging Face; demos on Hugging Face and Baidu AI Studio
- Works on
- open weights, Hugging Face
- In the UK
- Yes
- Free plan
- open weights, Apache 2.0
FAQ
Is ERNIE-Image free?
Yes. The weights are free on Hugging Face under the Apache 2.0 licence.
When was ERNIE-Image released?
Baidu introduced ERNIE-Image on 15 April 2026.
What hardware does ERNIE-Image need?
Baidu says it runs on consumer graphics cards with 24GB of memory.
Facts from the maker's own pages and the sources listed. Checked 2026-10-11.
Similar tools
ACE-Step 1.5
Open-source music generation model that writes full songs with lyrics and runs on ordinary consumer GPUs.
AlphaGenome Atlas
Google DeepMind database of AI-predicted effects for every possible single-letter change in human DNA.
Amazon Nova 2
Amazon's second family of Nova AI models, covering reasoning, speech conversation and combined text-and-image generation.
Cisco Antares
Cisco's small open-weight models that find where known security vulnerabilities sit in a codebase, running locally.