If you’re exploring realistic image generation with cutting-edge models, the HiDream-I1 GGUF version is a fantastic option—especially for users running on lower-spec hardware or preferring a more lightweight setup. In this post, I’ll walk you through a simple and native ComfyUI workflow to get high-quality photorealistic results from HiDream-I1 without any extra extensions or complicated configs. Whether you’re a beginner or just want a clean, minimal pipeline, this guide will help you get started quickly and effectively.
Models
- GGUF – There are three versions of HiDream-I1: full, dev, and fast. You can download the gguf models here: full gguf, dev gguf, fast gguf. Put the gguf file under ComfyUI\models\unet folder. I have a mobile 4090 GPU with 16GB VRAM. I used the hidream-i1-full-Q8_0.gguf for all the samples in this post.
- Text encoders – You need 4 text encoders. They can be found here. Download all 4 of them and put them in ComfyUI\models\text_encoders folder. Note that you might already have the t5xxl text encoder.
- VAE – The Flux VAE works well. If you don’t have it, you can download it here.
Installation
- Update ComfyUI to the latest version. If updating through ComfyUI manager doesn’t work, try running update\update_comfyui.bat script to update it.
- Drag this picture to ComfyUI’s canvas.
- Use ComfyUI manager to install missing nodes.
- Restart ComfyUI
Nodes
This node loads the GGUF model.
This is the new node to load the 4 text encoders.
Positive prompt and negative prompt. Try to use only 128 tokens or lower for positive prompt. I have included an image of a long prompt, and the result is not good. Note that negative prompt is only for full model. I had troubles initially to generate realistic images. After I used this negative prompt, the generations for realistic images are much better. Don’t use this negative prompt, if you want to generate anime or other illustration based images.
This is the sampler.
Here are official sampling settings
### HiDream Full
* hidream_i1_full_fp16.safetensors
* shift: 3.0
* steps: 50
* sampler: uni_pc
* scheduler: simple
* cfg: 5.0
### HiDream Dev
* hidream_i1_dev_bf16.safetensors
* shift: 6.0
* steps: 28
* sampler: lcm
* scheduler: normal
* cfg: 1.0 (no negative prompt)
### HiDream Fast
* hidream_i1_fast_bf16.safetensors
* shift: 3.0
* steps: 16
* sampler: lcm
* scheduler: normal
* cfg: 1.0 (no negative prompt)
This loads the VAE.
Samples
Prompt: A young woman standing on the beach at golden hour, facing the viewer with a gentle smile, wind blowing her dress slightly. The warm sun backlights her hair while her face is softly illuminated, capturing natural shadows and skin texture. Realistic water and sky in the background, photorealistic portrait with serene mood and lifelike detail. realistic, 8k hd
Prompt: Photo of a traveler stands on a grassy cliff beside ancient stone ruins, overlooking a mist-covered valley. Realistic textures in the stones and moss, soft morning light casting gentle shadows, distant mountains, high-resolution details, subtle mystical ambiance with a photorealistic finish.
Prompt: A woman in a casual yet stylish outfit—a cozy knit sweater, tailored jeans, and ankle boots—standing in a modern apartment with large windows letting in soft afternoon light. Photorealistic details of the fabric textures, natural light highlighting her look, high-resolution shot with clear, lifelike details of her relaxed yet fashionable appearance.
Prompt: Realistic photo of afemale tech operative standing indoors under cool-toned lighting, facing directly toward the viewer, wearing a futuristic visor and armored suit with subtle neon accents. Her expression is focused and serious, with photorealistic facial details, reflections on her gear, and visible UI holograms floating around her. Realistic sci-fi environment with sharp detail and cinematic clarity. realistic, 8k hd
Prompt: A woman in a flowing, midnight blue evening gown standing on a grand staircase inside a luxurious ballroom, soft chandelier light illuminating her figure, photorealistic fabric details, elegant pose, realistic textures of the gown’s intricate design, and the polished marble floors. The high-resolution image captures the beauty of high-end fashion in a sophisticated, cinematic manner.
The following is an image generated using a long prompt.
Prompt: A graceful Chinese woman in her early twenties, wearing a flowing, pastel-colored hanfu with intricate floral embroidery and sheer silk sleeves, stands on a stone bridge overlooking a tranquil lotus pond. Her long, dark hair is styled in a traditional updo adorned with delicate jade hairpins and silk ribbons that flutter gently in the breeze. She faces the camera with a serene, composed expression, hands softly resting on the wooden railing. Behind her, ancient willow trees drape their branches over the water, and the tiled roof of a traditional pagoda is visible through the early morning mist. The surface of the pond reflects soft light from the cloudy sky, dotted with floating lotus flowers. Photorealistic details are captured in the texture of her hanfu fabric, natural shadows on her face, and the worn stone under her feet. The image is sharp and cinematic, with high-resolution clarity, lifelike lighting, and subtle color grading that enhances the natural, peaceful atmosphere of the scene.
Conclusion
HiDream-I1 proves that photorealism doesn’t have to come at the cost of complexity. With a basic native ComfyUI setup, you can generate beautiful, cinematic images with minimal overhead and full control. This workflow is perfect for creative exploration, testing prompts, or building toward more advanced pipelines. If you’re looking for a smooth, no-fuss way to tap into HiDream-I1’s power, this is a great place to start.
Leave a Reply