Credits for the workflows and custom nodes:HiDiffusion (<a target="_blank" rel="ugc" href="https://github.com/blepping/comfyui_jankhidiffusion">https://github.com/blepping/comfyui_jankhidiffusion</a> - here you will find two example workflows without ELLA, for SD1.5 or SDXL)ELLA ( <a target="_blank" rel="ugc" href="https://github.com/TencentQQGYLab/ComfyUI-ELLA/tree/main/examples">https://github.com/TencentQQGYLab/ComfyUI-ELLA/tree/main/examples</a> ).HiDiffusion can be used to generate high resolution images without duplicated artifacts that can normally appear, and it should be also faster. The ComfyUI custom nodes are experimental according to the author.ELLA should help with prompt adherence, it's available for SD1.5 only. You'll need to download several files from HuggingFace: <a target="_blank" rel="ugc" href="https://huggingface.co/QQGYLab/ELLA">ELLA Models</a> in models/ella/ and <a target="_blank" rel="ugc" href="https://huggingface.co/QQGYLab/ELLA/tree/main/models--google--flan-t5-xl--text_encoder">FLAN-T5 XL Text Encoder</a> in the models/ella_encoder/ folder (note that you will have to download multiple files here).<img src="https://image.civitai.com/xG1nkqKTMzGDvpLrqFT7WA/246f9730-b40a-4827-ae54-9c31445d695c/width=525/246f9730-b40a-4827-ae54-9c31445d695c.jpeg" />

Files with SHA256: 371b215b70403c96949b0ba15c6f948ee32873b2e8ace8631044c29ed0d7fe76