<h1 id="ernie-image-aio-for-comfyui">🧠 ERNIE-Image AIO for ComfyUI</h1>All-in-One Checkpoints • BF16 &amp; FP8 • Text Encoder + VAE Integrated • Simple ComfyUI Setup<hr /><h2 id="what-is-this">✨ What is this?</h2>This release provides AIO (All-in-One) checkpoint versions of ERNIE-Image and ERNIE-Image-Turbo for ComfyUI.Instead of loading separate model parts, these AIO versions already include the image model, text encoder, and VAE inside a single checkpoint file. Just place the model into your checkpoint folder, load it with the regular Checkpoint Loader node, and build a very small workflow around it.The original ERNIE-Image family from Baidu focuses on strong instruction following, good text rendering, structured compositions, and support for a broad range of styles, from design-oriented images to more realistic or stylized outputs.<hr /><img src="https://image.civitai.com/xG1nkqKTMzGDvpLrqFT7WA/1e3fae5c-260c-4db1-9824-0d7f629b5721/original=true/1e3fae5c-260c-4db1-9824-0d7f629b5721.jpeg" /><hr /><img src="https://image.civitai.com/xG1nkqKTMzGDvpLrqFT7WA/0350d893-c281-4e9a-b5d4-007f2b8c0002/original=true/0350d893-c281-4e9a-b5d4-007f2b8c0002.jpeg" /><hr /><h3 id="huggingface.co-releases"><a target="_blank" rel="ugc" href="http://huggingface.co">huggingface.co</a> Releases</h3><ul><li><a target="_blank" rel="ugc" href="https://huggingface.co/SeeSee21/ERNIE-AIO">https://huggingface.co/SeeSee21/ERNIE-AIO</a></li></ul><h3 id="workflow-releases">Workflow Releases</h3><ul><li><a target="_blank" rel="ugc" href="https://civarchive.com/models/2553748/ernie-aio">https://civarchive.com/models/2553748/ernie-aio</a></li></ul><hr /><h2 id="included-versions">📦 Included Versions</h2>This release contains AIO checkpoints for:<ul><li>ERNIE-Base-AIO</li><li>ERNIE-Turbo-AIO</li></ul>Each version is provided in:<ul><li>BF16</li><li>FP8</li></ul><hr /><h2 id="aio-format">⚙️ AIO Format</h2>These checkpoints are made for a simple ComfyUI workflow.<h3 id="install-path">Install path</h3>Place the <code>.safetensors</code> files here:<code>\ComfyUI\models\checkpoints</code>Then load them with:<ul><li>Checkpoint Loader</li><li>CLIP Text Encode (Prompt)</li><li>Empty Latent Image</li><li>KSampler</li><li>VAE Decode</li><li>Save Image</li></ul>That is all you theoretically need to run the model.<hr /><h2 id="quantization-notes">🧪 Quantization Notes</h2>The FP8 versions in this release were quantized by me.For these FP8 AIO builds:<ul><li>Image model: FP8</li><li>Text encoder: FP8</li><li>VAE: FP32</li></ul>This setup was chosen to reduce size and improve practicality while keeping decoding stable and reliable.<hr /><h2 id="recommended-usage">🖼️ Recommended Usage</h2><h3 id="ernie-base-aio">ERNIE-Base-AIO</h3><ul><li>Best for: general-purpose generation, stronger prompt fidelity, more deliberate results</li><li>Recommended CFG: 4.0</li><li>Recommended steps: 50</li></ul><h3 id="ernie-turbo-aio">ERNIE-Turbo-AIO</h3><ul><li>Best for: fast generation, quick iteration, efficient workflows</li><li>Recommended CFG: 1.0</li><li>Recommended steps: 8</li></ul>The official ERNIE-Image documentation recommends the same core resolutions for both releases, including:<ul><li>1024×1024</li><li>848×1264</li><li>1264×848</li><li>768×1376</li><li>896×1200</li><li>1376×768</li><li>1200×896</li></ul><hr /><h2 id="prompting">🧾 Prompting</h2>ERNIE models work best with natural language prompts.Describe:<ul><li>subject</li><li>scene</li><li>composition</li><li>lighting</li><li>mood</li><li>camera framing</li><li>details you actually want in the image</li></ul>These models are especially interesting for:<ul><li>posters</li><li>text-heavy designs</li><li>comics</li><li>structured layouts</li><li>prompt-sensitive image generation</li></ul><hr /><h2 id="comfyui-notes">🔧 ComfyUI Notes</h2>These AIO checkpoints are designed for a checkpoint-style workflow, so you do not need to manually load separate text encoder or VAE files.A workflow for this release will also be uploaded separately.<hr /><h2 id="credits">🙏 Credits</h2>Huge thanks to the original ERNIE-Image team at Baidu for releasing these models and their research to the community. The original official releases are available here:<ul><li><a target="_new" rel="ugc" href="https://huggingface.co/baidu/ERNIE-Image?utm_source=chatgpt.com">ERNIE-Image on Hugging Face</a></li><li><a target="_new" rel="ugc" href="https://huggingface.co/baidu/ERNIE-Image-Turbo?utm_source=chatgpt.com">ERNIE-Image-Turbo on Hugging Face</a></li><li>Official GitHub Repository</li></ul>Both official model pages list the release under the Apache-2.0 license.<hr /><h2 id="final-note">❤️ Final Note</h2>I made these AIO versions to make ERNIE-Image easier to use in ComfyUI — especially for people who prefer the classic single-checkpoint workflow.Load checkpoint, prompt, sample, decode, save — done.

ERNIE-AIO

⚡ Ernie-Turbo-AIOThis is the Turbo AIO version of ERNIE-Image for ComfyUI.It is designed for fast image generation and quick iteration while still keeping strong prompt adherence and clean visual output. Recommended settings are around 8 steps and CFG 1.0. Like the Base release, this AIO version already includes the text encoder and VAE, so you can load it directly as a standard checkpoint from <code>ComfyUI\models\checkpoints</code>.

Files with SHA256: 9f6d0a1d77c39347d847abf20a869847c76b3e17860e20254548be1489be6fbb