justingirard/draft-ui-designer 🖼️🔢❓📝✓ → 🖼️
About
An experiment: a fine-tuned FLUX model for UI design generation
Example Output
Prompt:
"A screen that shows many items to buy from a shop, including pricing information. The design should look like THE_SLEEK_UI."
Output
Performance Metrics
29.94s
Prediction Time
29.94s
Total Time
All Input Parameters
{
"image": "https://replicate.delivery/pbxt/Lw7tTNZgCZfNZJovklt6rYqnUiCZ9XUOs89V2CSCsn9OyXjJ/Untitled.png",
"model": "dev",
"width": 1024,
"height": 1024,
"prompt": "A screen that shows many items to buy from a shop, including pricing information. The design should look like THE_SLEEK_UI.",
"lora_scale": 1,
"num_outputs": 1,
"aspect_ratio": "1:1",
"output_format": "webp",
"guidance_scale": 3.24,
"output_quality": 90,
"prompt_strength": 0.75,
"extra_lora_scale": 1,
"num_inference_steps": 28
}
Input Parameters
- mask
- Image mask for image inpainting mode. If provided, aspect_ratio, width, and height inputs are ignored.
- seed
- Random seed. Set for reproducible generation
- image
- Input image for image to image or inpainting mode. If provided, aspect_ratio, width, and height inputs are ignored.
- model
- Which model to run inference with. The dev model performs best with around 28 inference steps but the schnell model only needs 4 steps.
- width
- Width of generated image. Only works if `aspect_ratio` is set to custom. Will be rounded to nearest multiple of 16. Incompatible with fast generation
- height
- Height of generated image. Only works if `aspect_ratio` is set to custom. Will be rounded to nearest multiple of 16. Incompatible with fast generation
- prompt (required)
- Prompt for generated image. If you include the `trigger_word` used in the training process you are more likely to activate the trained object, style, or concept in the resulting image.
- go_fast
- Run faster predictions with model optimized for speed (currently fp8 quantized); disable to run in original bf16
- extra_lora
- Load LoRA weights. Supports Replicate models in the format <owner>/<username> or <owner>/<username>/<version>, HuggingFace URLs in the format huggingface.co/<owner>/<model-name>, CivitAI URLs in the format civitai.com/models/<id>[/<model-name>], or arbitrary .safetensors URLs from the Internet. For example, 'fofr/flux-pixar-cars'
- lora_scale
- Determines how strongly the main LoRA should be applied. Sane results between 0 and 1 for base inference. For go_fast we apply a 1.5x multiplier to this value; we've generally seen good performance when scaling the base value by that amount. You may still need to experiment to find the best value for your particular lora.
- megapixels
- Approximate number of megapixels for generated image
- num_outputs
- Number of outputs to generate
- aspect_ratio
- Aspect ratio for the generated image. If custom is selected, uses height and width below & will run in bf16 mode
- output_format
- Format of the output images
- guidance_scale
- Guidance scale for the diffusion process. Lower values can give more realistic images. Good values to try are 2, 2.5, 3 and 3.5
- output_quality
- Quality when saving the output images, from 0 to 100. 100 is best quality, 0 is lowest quality. Not relevant for .png outputs
- prompt_strength
- Prompt strength when using img2img. 1.0 corresponds to full destruction of information in image
- extra_lora_scale
- Determines how strongly the extra LoRA should be applied. Sane results between 0 and 1 for base inference. For go_fast we apply a 1.5x multiplier to this value; we've generally seen good performance when scaling the base value by that amount. You may still need to experiment to find the best value for your particular lora.
- replicate_weights
- Load LoRA weights. Supports Replicate models in the format <owner>/<username> or <owner>/<username>/<version>, HuggingFace URLs in the format huggingface.co/<owner>/<model-name>, CivitAI URLs in the format civitai.com/models/<id>[/<model-name>], or arbitrary .safetensors URLs from the Internet. For example, 'fofr/flux-pixar-cars'
- num_inference_steps
- Number of denoising steps. More steps can give more detailed images, but take longer.
- disable_safety_checker
- Disable safety checker for generated images.
Output Schema
Output
Example Execution Logs
Using seed: 15837 Prompt: A screen that shows many items to buy from a shop, including pricing information. The design should look like THE_SLEEK_UI. [!] Resizing input image from 2824x1346 to 2832x1360 [!] img2img mode [!] Using dev model for img2img Using dev model Weights already loaded Loaded LoRAs in 0.02s 0%| | 0/21 [00:00<?, ?it/s] 5%|▍ | 1/21 [00:01<00:22, 1.13s/it] 10%|▉ | 2/21 [00:02<00:24, 1.26s/it] 14%|█▍ | 3/21 [00:03<00:23, 1.31s/it] 19%|█▉ | 4/21 [00:05<00:22, 1.33s/it] 24%|██▍ | 5/21 [00:06<00:21, 1.34s/it] 29%|██▊ | 6/21 [00:07<00:20, 1.35s/it] 33%|███▎ | 7/21 [00:09<00:18, 1.35s/it] 38%|███▊ | 8/21 [00:10<00:17, 1.36s/it] 43%|████▎ | 9/21 [00:12<00:16, 1.36s/it] 48%|████▊ | 10/21 [00:13<00:14, 1.36s/it] 52%|█████▏ | 11/21 [00:14<00:13, 1.36s/it] 57%|█████▋ | 12/21 [00:16<00:12, 1.36s/it] 62%|██████▏ | 13/21 [00:17<00:10, 1.36s/it] 67%|██████▋ | 14/21 [00:18<00:09, 1.36s/it] 71%|███████▏ | 15/21 [00:20<00:08, 1.36s/it] 76%|███████▌ | 16/21 [00:21<00:06, 1.36s/it] 81%|████████ | 17/21 [00:22<00:05, 1.36s/it] 86%|████████▌ | 18/21 [00:24<00:04, 1.36s/it] 90%|█████████ | 19/21 [00:25<00:02, 1.36s/it] 95%|█████████▌| 20/21 [00:27<00:01, 1.36s/it] 100%|██████████| 21/21 [00:28<00:00, 1.37s/it] 100%|██████████| 21/21 [00:28<00:00, 1.35s/it]
Version Details
- Version ID
f1114b15c528a5d9e40b905a5138f6abdeecb91b57e33ddbb3a71ca10ac92b41- Version Created
- July 27, 2025