Open samliu315 opened 4 months ago
The batch upload is for batch processing. It is not equivalent to previous Multi-Inputs
. Multi-Inputs
feature is removed. See https://github.com/lllyasviel/stable-diffusion-webui-forge/discussions/40 .
You need to turn on multiple units each with an input to achieve the same result as previous Multi-Inputs
. Please file an feature request and we will see if it is necessary to add that feature back.
photomaker has never worked, I've tried it, either with a single image or multiple controlnets.
And the vertical alignment of controlnet is also rather inconvenient, is it because my UI preset is not selected correctly?
FWIW I am unable to get PhotoMaker to generate an image using the image provided in the CN Single Image console either.
Could be user error? But I'm pretty sure I have the proper model and there are no errors in the console. The generated image just isn't altered by the model.
100%|████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████| 29/29 [00:05<00:00, 5.26it/s] Total progress: 100%|████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████| 29/29 [00:05<00:00, 4.92it/s] 2024-02-06 21:15:29,613 - ControlNet - INFO - ControlNet Input Mode: InputMode.SIMPLE████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████| 29/29 [00:05<00:00, 5.29it/s] 2024-02-06 21:15:29,617 - ControlNet - INFO - Using preprocessor: ClipVision (Photomaker) 2024-02-06 21:15:29,618 - ControlNet - INFO - preprocessor resolution = 0.5 Warning torch.load doesn't support weights_only on this pytorch version, loading unsafely. 2024-02-06 21:15:32,094 - ControlNet - INFO - Current ControlNet PhotomakerPatcher: D:\stable-diffusion-webui-forge\models\ControlNet\photomaker-v1.bin 2024-02-06 21:15:32,148 - ControlNet - INFO - ControlNet Method ClipVision (Photomaker) patched. To load target model SDXL Begin to load 1 model unload clone 3 Moving model(s) has taken 1.64 seconds 100%|████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████| 29/29 [00:03<00:00, 7.48it/s] Total progress: 100%|████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████| 29/29 [00:04<00:00, 6.46it/s] 2024-02-06 21:17:31,613 - ControlNet - INFO - ControlNet Input Mode: InputMode.SIMPLE████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████| 29/29 [00:04<00:00, 7.62it/s] 2024-02-06 21:17:31,617 - ControlNet - INFO - Using preprocessor: ClipVision (Photomaker) 2024-02-06 21:17:31,617 - ControlNet - INFO - preprocessor resolution = 0.5 2024-02-06 21:17:31,738 - ControlNet - INFO - Current ControlNet PhotomakerPatcher: D:\stable-diffusion-webui-forge\models\ControlNet\photomaker-v1.bin 2024-02-06 21:17:31,791 - ControlNet - INFO - ControlNet Method ClipVision (Photomaker) patched. To load target model SDXL Begin to load 1 model unload clone 2 Moving model(s) has taken 2.53 seconds 100%|████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████| 29/29 [00:03<00:00, 7.42it/s] Total progress: 100%|████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████| 29/29 [00:04<00:00, 6.05it/s] 2024-02-06 21:23:23,992 - ControlNet - INFO - ControlNet Input Mode: InputMode.SIMPLE████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████| 29/29 [00:04<00:00, 7.63it/s] 2024-02-06 21:23:23,996 - ControlNet - INFO - Using preprocessor: ClipVision (Photomaker) 2024-02-06 21:23:23,996 - ControlNet - INFO - preprocessor resolution = 0.5 2024-02-06 21:23:24,118 - ControlNet - INFO - Current ControlNet PhotomakerPatcher: D:\stable-diffusion-webui-forge\models\ControlNet\photomaker-v1.bin 2024-02-06 21:23:24,174 - ControlNet - INFO - ControlNet Method ClipVision (Photomaker) patched. To load target model SDXL Begin to load 1 model unload clone 2 Moving model(s) has taken 2.55 seconds 100%|████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████| 29/29 [00:04<00:00, 7.23it/s] Total progress: 100%|████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████| 29/29 [00:04<00:00, 6.00it/s] Total progress: 100%|████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████| 29/29 [00:04<00:00, 7.49it/s]
System Info:
```python
{
"Platform": "Windows-10-10.0.19045-SP0",
"Python": "3.10.6",
"Version": "f0.0.10-latest-58-ge6263135",
"Commit": "e62631350a408edf3b1f0a9dd45a43f9f0e95ead",
"Script path": "D:\\stable-diffusion-webui-forge",
"Data path": "D:\\stable-diffusion-webui-forge",
"Extensions dir": "D:\\stable-diffusion-webui-forge\\extensions",
"Checksum": "f1b5e4261ccd6eed11591f91ee11455510852ccc6649d03c744d021fac250d93",
"Commandline": [
"launch.py",
"--ckpt-dir",
"e:\\stable Diffusion Checkpoints"
],
"Torch env info": {
"torch_version": "2.1.2+cu121",
"is_debug_build": "False",
"cuda_compiled_version": "12.1",
"gcc_version": null,
"clang_version": null,
"cmake_version": null,
"os": "Microsoft Windows 10 Pro",
"libc_version": "N/A",
"python_version": "3.10.6 (tags/v3.10.6:9c7b4bd, Aug 1 2022, 21:53:49) [MSC v.1932 64 bit (AMD64)] (64-bit runtime)",
"python_platform": "Windows-10-10.0.19045-SP0",
"is_cuda_available": "True",
"cuda_runtime_version": "12.1.66\r",
"cuda_module_loading": "LAZY",
"nvidia_driver_version": "551.23",
"nvidia_gpu_models": "GPU 0: NVIDIA GeForce RTX 4090",
"cudnn_version": null,
"pip_version": "pip3",
"pip_packages": [
"numpy==1.26.2",
"open-clip-torch==2.20.0",
"pytorch-lightning==1.9.4",
"torch==2.1.2+cu121",
"torchdiffeq==0.2.3",
"torchmetrics==1.3.0.post0",
"torchsde==0.2.6",
"torchvision==0.16.2+cu121"
],
"conda_packages": null,
"hip_compiled_version": "N/A",
"hip_runtime_version": "N/A",
"miopen_runtime_version": "N/A",
"caching_allocator_config": "",
"is_xnnpack_available": "True",
"cpu_info": [
"Architecture=9",
"CurrentClockSpeed=3200",
"DeviceID=CPU0",
"Family=207",
"L2CacheSize=16384",
"L2CacheSpeed=",
"Manufacturer=GenuineIntel",
"MaxClockSpeed=3200",
"Name=Intel(R) Core(TM) i9-14900K",
"ProcessorType=3",
"Revision="
]
},
"Exceptions": [
{
"exception": "'dict' object has no attribute 'shape'",
"traceback": [
[
"D:\\stable-diffusion-webui-forge\\modules\\call_queue.py, line 57, f",
"res = list(func(*args, **kwargs))"
],
[
"D:\\stable-diffusion-webui-forge\\modules\\call_queue.py, line 36, f",
"res = func(*args, **kwargs)"
],
[
"D:\\stable-diffusion-webui-forge\\modules\\txt2img.py, line 110, txt2img",
"processed = processing.process_images(p)"
],
[
"D:\\stable-diffusion-webui-forge\\modules\\processing.py, line 749, process_images",
"res = process_images_inner(p)"
],
[
"D:\\stable-diffusion-webui-forge\\modules\\processing.py, line 920, process_images_inner",
"samples_ddim = p.sample(conditioning=p.c, unconditional_conditioning=p.uc, seeds=p.seeds, subseeds=p.subseeds, subseed_strength=p.subseed_strength, prompts=p.prompts)"
],
[
"D:\\stable-diffusion-webui-forge\\modules\\processing.py, line 1275, sample",
"samples = self.sampler.sample(self, x, conditioning, unconditional_conditioning, image_conditioning=self.txt2img_image_conditioning(x))"
],
[
"D:\\stable-diffusion-webui-forge\\modules\\sd_samplers_kdiffusion.py, line 251, sample",
"samples = self.launch_sampling(steps, lambda: self.func(self.model_wrap_cfg, x, extra_args=self.sampler_extra_args, disable=False, callback=self.callback_state, **extra_params_kwargs))"
],
[
"D:\\stable-diffusion-webui-forge\\modules\\sd_samplers_common.py, line 260, launch_sampling",
"return func()"
],
[
"D:\\stable-diffusion-webui-forge\\modules\\sd_samplers_kdiffusion.py, line 251,
Note that photomaker is a special control that need you to add the trigger word "photomaker"
Your prompt should be like "a photo of photomaker"
clipvision(photomaker) this preprocessor didn't work. I couldn't find the model's filename.
clipvision(photomaker) this preprocessor didn't work. I couldn't find the model's filename.
Have you downloaded Photomaker-v1.bin from here? https://huggingface.co/TencentARC/PhotoMaker/tree/main
clipvision(photomaker) this preprocessor didn't work. I couldn't find the model's filename.
Have you downloaded Photomaker-v1.bin from here? https://huggingface.co/TencentARC/PhotoMaker/tree/main
yes. It is a cn model,not a preprocessor.
I have the same issue. Photomaker CN doesn't seem to work at all. I used "a photo of photomaker" prompt and Photomaker-v1.bin CN. I have this error on cmd when using it, doesn't happen with other CNs:
"Warning torch.load doesn't support weights_only on this pytorch version, loading unsafely."
Use SDXL checkpoints, not 1.5
Also not working for me, the image does not change after enabling photomaker with photomaker bin model plus Photomaker in the prompt, same output
I think there is a problem, when using the required "a photo of photomaker" if a "," is just next to it for example: "a photo of photomaker, a woman in astronaut suit" that coma is ruining the activation.
I also noticed that photomaker does work when the prompt is only "a photo of photomaker". Additional words disable the effect.
Photomaker github recommends a different trigger word. I'm AFK but maybe someone can test to see if changes something.
(From their github)
img
must follow the class word for personalizationprompt = "a half-body portrait of a man img wearing the sunglasses in Iron man suit, best quality" negative_prompt = "(asymmetry, worst quality, low quality, illustration, 3d, 2d, painting, cartoons, sketch), open mouth, grayscale" generator = torch.Generator(device=device).manual_seed(42) images = pipe( prompt=prompt, input_id_images=input_id_images, negative_prompt=negative_prompt, num_images_per_prompt=1, num_inference_steps=num_steps, start_merge_step=10, generator=generator, ).images[0] gen_images.save('out_photomaker.png')
clipvision(photomaker) this preprocessor didn't work. I couldn't find the model's filename.
Have you downloaded Photomaker-v1.bin from here? https://huggingface.co/TencentARC/PhotoMaker/tree/main
Thanks.... and where do we put this file?
clipvision(photomaker) this preprocessor didn't work. I couldn't find the model's filename.
Have you downloaded Photomaker-v1.bin from here? https://huggingface.co/TencentARC/PhotoMaker/tree/main
Thanks.... and where do we put this file?
Inside the models/controlnet folder
I also noticed that photomaker does work when the prompt is only "a photo of photomaker". Additional words disable the effect.
Photomaker github recommends a different trigger word. I'm AFK but maybe someone can test to see if changes something.
(From their github)
Note that the trigger word
img
must follow the class word for personalizationprompt = "a half-body portrait of a man img wearing the sunglasses in Iron man suit, best quality" negative_prompt = "(asymmetry, worst quality, low quality, illustration, 3d, 2d, painting, cartoons, sketch), open mouth, grayscale" generator = torch.Generator(device=device).manual_seed(42) images = pipe( prompt=prompt, input_id_images=input_id_images, negative_prompt=negative_prompt, num_images_per_prompt=1, num_inference_steps=num_steps, start_merge_step=10, generator=generator, ).images[0] gen_images.save('out_photomaker.png')
The "img" alone didn't work, "photomaker" alone didn't work, only "a photo of photomaker" works, at least in my test.
So, after some experimentation, I found that:
Example: These work:
a photo of mad scientist photomaker
a photo of mad scientist photomaker wearing a red jacket
a photo of mad scientist photomaker , wearing a green jacket
a photo of mad scientist photomaker , masterpiece, best best quality, high quality
These don't work:
a photo of mad scientist photomaker, wearing a red jacket
a photo of mad scientist photomaker, masterpiece, best best quality, high quality
@lllyasviel FYI
So, after some experimentation, I found that:
- It only works with SDXL models (as it's meant to, but it's not clearly labeled as such in the program).
- Adjusting the control net weight does very little (this is fine).
- You need to use the trigger word "photomaker" in the prompt.
- You can't have a comma right after the photomaker word. Sometimes I wasn't allowed to have a space then comma after. But sometimes I was. Not quite sure.
Example: These work:
a photo of mad scientist photomaker a photo of mad scientist photomaker wearing a red jacket a photo of mad scientist photomaker , wearing a green jacket a photo of mad scientist photomaker , masterpiece, best best quality, high quality
These don't work:
a photo of mad scientist photomaker, wearing a red jacket a photo of mad scientist photomaker, masterpiece, best best quality, high quality
@lllyasviel FYI
Still doesn't work very well probably better in comfyui
So, after some experimentation, I found that:
- It only works with SDXL models (as it's meant to, but it's not clearly labeled as such in the program).
- Adjusting the control net weight does very little (this is fine).
- You need to use the trigger word "photomaker" in the prompt.
- You can't have a comma right after the photomaker word. Sometimes I wasn't allowed to have a space then comma after. But sometimes I was. Not quite sure.
Example: These work:
a photo of mad scientist photomaker a photo of mad scientist photomaker wearing a red jacket a photo of mad scientist photomaker , wearing a green jacket a photo of mad scientist photomaker , masterpiece, best best quality, high quality
These don't work:
a photo of mad scientist photomaker, wearing a red jacket a photo of mad scientist photomaker, masterpiece, best best quality, high quality
@lllyasviel FYI
Still doesn't work very well probably better in comfyui
"A photo of a woman wearing a jacket img photomaker img", I used these prompt words, and the total number and total length of all prompt words cannot be too long. You can try to reduce the number or start from the length.
a photo of woman img photomaker wearing a jacket img,photo inset,solo,brown hair,bag,smile,looking at viewer,long hair,suit,skirt,formal,brown eyes,photo background,outdoors,jacket,ground vehicle,shirt,office lady,Cinematic Lighting,
Also working:
the photo was shot on Kodak Ektachrome 64 film with a Nikon F SLR camera and a Nikkor f/1.4 50mm lens, shutter speed 1/250s, ISO 64. It shows a man photomaker wearing a red jacket, in the forest
It seems to me that the trigger word "photomaker" just needs a space behind it, no matter what follows afterwards. "img" is not needed as a trigger (at least in my tries).
I've noticed that it does not work when you hit "enter" after trigger word "photomaker" (if case you want to continue typing prompts in the new line). Even when there is no more prompts in new line.
So, it seem like only "space" can be used after trigger word "photomaker".
Did some quick testing as well, findings:
I finally got it to work. I took a screengrab from the movie "A Scanner Darkly" (rotoscoped real actors).
Photos of params used are attached. I didn't have any luck until I set cfg scale low (3.5), no independent control image (I'm using img-to-img), and the only denoise strength that would work is 0.65.
Checklist
What happened?
controlnet -photomaker didn't work
![2](https://github.com/lllyasviel/stable-diffusion-webui-forge/assets/77907388/4fe0c736-5bcd-4b9b-9e91-ecfb78c4f8f1)
Steps to reproduce the problem
What should have happened?
Generate a graphic with the faces of the people I uploaded
What browsers do you use to access the UI ?
Google Chrome
Sysinfo
sysinfo-2024-02-06-16-36.json
Console logs
Additional information
No response