adirik/prompt-to-prompt-realvisxl-3.0 🔢🖼️📝❓ → 🖼️
About
Image editing with Prompt-to-Prompt for RealVisXL-v3.0
Example Output
Output
Performance Metrics
21.55s
Prediction Time
123.67s
Total Time
All Input Parameters
{
"seed": 864,
"scheduler": "DPM++_SDE_Karras",
"edited_prompt": "a pink dragon riding a bicycle on the beach",
"guidance_scale": 2,
"original_prompt": "a pink bear riding a bicycle on the beach",
"prompt_edit_type": "Replacement",
"self_replace_steps": 0.4,
"cross_replace_steps": 0.8,
"num_inference_steps": 25
}
Input Parameters
- seed
- Random seed. Leave blank to randomize the seed for original output.
- image
- Optional input image. If provided, DDIM inversion is performed to retrieve initial latents for image generation.
- local_edit
- Comma seperated words to determine which area should be changed. If None, then the whole image can be changed.
- edited_prompt
- Prompt used for editing the original sdxl output image. If prompt_edit_type above is `Re-weight`, you can leave this empty.
- guidance_scale
- Text guidance scale, use higher values for better alignment with input prompt.
- equalizer_words
- Words to be re-weighted (either enhancement or diminishment). Provide the words in the format of 'word1, word2, word3'. If you are not using reweight, leave this empty.
- original_prompt (required)
- Prompt to generate image with SDXL
- prompt_edit_type (required)
- Choose the type of the prompt editing, can be either of `Replacement`, `Refinement`, `Re-weight`
- self_replace_steps
- Number of diffusion steps in which self attention should be replaced
- cross_replace_steps
- Number of diffusion steps in which cross attention should be replaced
- equalizer_strengths
- Strengths for the words to be re-weighted. It can be positive or negative values. Provide the strengths in the format of 'strength1, strength2, strength3', for respective equalizer_words. If you are not using reweight, leave this empty.
- num_inference_steps
- Number of diffusion denoising steps for image generation.
- num_inversion_steps
- Number of diffusion denoising steps for inversion.
Output Schema
Output
Example Execution Logs
Using seed: 864
Changing scheduler {self.config} to have `lower_order_final` set to True to handle uneven amount of inference steps. Please make sure to always use an even number of `num_inference steps when using `lower_order_final=False`.
0%| | 0/25 [00:00<?, ?it/s]
4%|▍ | 1/25 [00:01<00:45, 1.90s/it]
8%|▊ | 2/25 [00:02<00:22, 1.02it/s]
12%|█▏ | 3/25 [00:02<00:17, 1.27it/s]
16%|█▌ | 4/25 [00:03<00:14, 1.43it/s]
20%|██ | 5/25 [00:03<00:12, 1.55it/s]
24%|██▍ | 6/25 [00:04<00:11, 1.62it/s]
28%|██▊ | 7/25 [00:05<00:10, 1.67it/s]
32%|███▏ | 8/25 [00:05<00:09, 1.71it/s]
36%|███▌ | 9/25 [00:06<00:09, 1.74it/s]
40%|████ | 10/25 [00:06<00:08, 1.75it/s]
44%|████▍ | 11/25 [00:07<00:07, 1.77it/s]
48%|████▊ | 12/25 [00:07<00:07, 1.78it/s]
52%|█████▏ | 13/25 [00:08<00:06, 1.79it/s]
56%|█████▌ | 14/25 [00:08<00:06, 1.80it/s]
60%|██████ | 15/25 [00:09<00:05, 1.80it/s]
64%|██████▍ | 16/25 [00:10<00:04, 1.80it/s]
68%|██████▊ | 17/25 [00:10<00:04, 1.81it/s]
72%|███████▏ | 18/25 [00:11<00:03, 1.81it/s]
76%|███████▌ | 19/25 [00:11<00:03, 1.81it/s]
80%|████████ | 20/25 [00:12<00:02, 1.81it/s]
84%|████████▍ | 21/25 [00:12<00:02, 1.81it/s]
88%|████████▊ | 22/25 [00:13<00:01, 1.81it/s]
92%|█████████▏| 23/25 [00:13<00:01, 1.81it/s]
96%|█████████▌| 24/25 [00:14<00:00, 1.81it/s]
100%|██████████| 25/25 [00:14<00:00, 1.81it/s]
100%|██████████| 25/25 [00:14<00:00, 1.67it/s]
Number of output images: 2
Version Details
- Version ID
f5de4e355ac44bca7fbdcf1518c6f74c1716e3883ed54d408c636ae716549c26- Version Created
- March 16, 2024