The HiDream-I1 model has quickly gained attention for its impressive image generation, offering remarkable detail. If you’re working in ComfyUI, you can enhance your results by integrating LoRA (Low-Rank Adaptation) models — a powerful way to fine-tune outputs toward specific characters, styles, or aesthetics without retraining the entire model. In this guide, we’ll walk through how to properly load and use LoRA models with HiDream-I1 in ComfyUI to boost your workflow and achieve customized visual results with minimal setup.
Models
- If you haven’t used HiDream-i1 before, please see this post and download necessary models.
- LoRA model – I am using this Porsch Targe LoRA. Download the model file and put it under ComfyUI\models\loras.
Installation
- Make sure ComfyUI is update to date. I tested this on v0.3.30. Please update your ComfyUI to the lastest if yours is lower than v0.3.30.
- Drag the following image to your ComfyUI canvas.
- Use ComfyUI Manager to install all missing nodes if there are any.
- Restart ComfyUI.
Node
There is only one new node in this workflow. You specify the LoRA you downloaded earlier here. Note that some LoRAs have trigger words, you will have to specify the trigger words in your positive prompt. In this case, the trigger words are Porsche Targa. You can find out the trigger words from where you download the model from.
If you want to use multiple LoRAs, just chain multiple LoraLoaderModelOnly together.
Examples
Prompt: realistic photo of a black Porsche 911 Targa driving through a neon-lit city street at night, cinematic angle, glossy reflections on wet pavement, motion blur on background, high detail headlights, sleek body design, luxury car commercial style, soft glow from street lights, vibrant atmosphere
Prompt: realistic photo of a silver Porsche 911 Targa on a winding mountain road at sunrise, front three-quarter view, golden light hitting the car, misty valley below, clear sky, dramatic scenic background, crisp details on the car, luxury performance vibe, commercial photography style
The prompt for the feature image is: realistic photo of a blue Porsche 911 Targa on a vast reflective salt flat at sunset, dramatic lighting, calm mirror-like surface, mountains in the background, soft pastel sky, cinematic composition, front-side view, ultra-detailed car body, shadows and reflections, high resolution
The negative prompt for all the images is: anime, cartoon, sketch, painting, illustration, blurry, unrealistic, CGI, render, doll-like, smooth skin, plastic texture, low quality, oversaturated, low resolution, 3d, artstation, concept art, stylized, fantasy face, glowing eyes
Conclusion
Using LoRA models with HiDream-I1 in ComfyUI opens the door to efficient style adaptation and subject-specific tuning, making it easier to generate high-quality, personalized images. Whether you’re aiming for subtle enhancements or distinct transformations, LoRA models provide flexibility while maintaining the base model’s photorealistic strength. With just a few adjustments in ComfyUI, you can integrate LoRAs into your pipeline and unlock a new level of creative control.
Leave a Reply