diff --git a/CHANGELOG.md b/CHANGELOG.md index 3a1df75..16af674 100644 --- a/CHANGELOG.md +++ b/CHANGELOG.md @@ -1,3 +1,6 @@ +### v1.4.6 - 2023 Sep.19 +- Add **HDR** Script + ### v1.4.5 - 2023 Sep.13 - Bug Fix for Color Wheel in `img2img` - Minor Formatting diff --git a/README.md b/README.md index 8a6852d..a365e22 100644 --- a/README.md +++ b/README.md @@ -1,23 +1,20 @@ # SD Webui Vectorscope CC -This is an Extension for the [Automatic1111 Webui](https://github.com/AUTOMATIC1111/stable-diffusion-webui), which performs *a kind of* **Offset Noise**[*](#offset-noise-tldr) natively. +This is an Extension for the [Automatic1111 Webui](https://github.com/AUTOMATIC1111/stable-diffusion-webui), which performs a kind of **Offset Noise**[*](#offset-noise-tldr) natively, +allowing you to adjust the brightness, contrast, and color of the generations. > [Sample Images](#sample-images) ## How to Use -After installing this Extension, you will see a new section in both **txt2img** and **img2img** tabs, -refer to the parameters and sample images below and play around with the values. +After installing this Extension, you will see a new section in both **txt2img** and **img2img** tabs. +Refer to the parameters and sample images below and play around with the values. -**Note:** Since this modifies the underlying latent noise, the composition may change drastically. +**Note:** Since this modifies the underlying latent noise, the composition may change drastically. Using the **Ones** scaling seems to reduce the variations. #### Parameters -- **Enable:** Turn on & off this Extension +- **Enable:** Turn on/off this Extension - **Alt:** Modify an alternative Tensor instead, causing the effects to be significantly stronger - **Skip:** Skip the last percentage of steps and only process the first few steps - - *Not as useful now that you can tune the [Scaling Settings](#scaling-settings)* - -

-

When Alt. is enabled, the image can get distorted at high value
Increase Skip to still achieve a stronger effect but without distortion

- + - *Not as useful now since you can tune the [Scaling Settings](#scaling-settings) instead* - **Brightness:** Adjust the overall brightness of the image - **Contrast:** Adjust the overall contrast of the image - **Saturation:** Adjust the overall saturation of the image @@ -65,10 +62,8 @@ refer to the parameters and sample images below and play around with the values. - Click **Refresh Style** to update the `Dropdown` if you edited the `styles.json` directly #### Advanced Settings - - **Process Hires. fix:** By default, this Extension only functions during the **txt2img** phase, so that **Hires. fix** may "fix" the artifacts introduced during **txt2img**. Enable this to process **Hires. fix** phase too. - This option does not affect **img2img** - - **Note:** Keep the **txt2img** base `steps` higher than **Hires. fix** `steps` if you enable this ##### Noise Settings > let `x` denote the Tensor ; let `y` denote the operations @@ -95,7 +90,7 @@ refer to the parameters and sample images below and play around with the values. ##### Scaling Settings Previously, this Extension offsets the noise by the same amount each step. -But due to the denoising process , this may produce undesired outcomes such as blurriness at high **Brightness** or noises at low **Brightness**. +But due to the denoising process, this may produce undesired outcomes such as blurriness at high **Brightness** or noises at low **Brightness**. Thus, I added a scaling option to modify the offset amount. > Essentially, the "magnitude" of the default Tensor gets smaller every step, so offsetting by the same amount will have stronger effects at later steps. This is reversed on the `Alt.` Tensor however. @@ -174,19 +169,20 @@ Thus, I added a scaling option to modify the offset amount. - [X] Fix the **Brightness** issues *~~kinda~~* - [X] Add API Docs - [X] Add Infotext Support *(by. [catboxanon](https://github.com/catboxanon))* +- [X] ADD **HDR** Script - [ ] Add Gradient features -

X/Y/Z Plot Support
(Outdated Contrast Value)

+

+

X/Y/Z Plot w/ Randomize

-

For Randomize in X/Y/Z Plot, the value is used as the random seed
-You can refer to the console to see the randomized values

+

The value is used as the random seed
You can refer to the console to see the randomized values

## API You can use this Extension via [API](https://github.com/AUTOMATIC1111/stable-diffusion-webui/wiki/API) by adding an entry in the `alwayson_scripts` of your payload. An [example](samples/api_example.json) is provided. -The `args` are the sent in the following order: +The `args` are sent in the following order: - **[Enable, Alt, Brightness, Contrast, Saturation, R, G, B, Skip, Process Hires. Fix, Noise Settings, Scaling Settings]** > `bool`, `bool`, `float`, `float`, `float`, `float`, `float`, `float`, `float`, `bool`, `str`, `str` @@ -195,9 +191,24 @@ The `args` are the sent in the following order: - Does not work with `DDIM`, `UniPC` samplers - Has little effect when used with certain **LoRA**s +## HDR +

BETA

+ +> [Discussion Thread](https://github.com/Haoming02/sd-webui-vectorscope-cc/issues/16) + +- In the **Script** `Dropdown` at the bottom, there is now a new option: **`High Dynamic Range`** +- This script will generate multiple images *("Brackets")* of varying brightness, then merge them into 1 HDR image +- *Do provide feedback in the thread!* + +#### Settings +- **Brackets:** The numer of images to generate +- **Gaps:** The brightness difference between each image +- **Automatically Merge:** When enabled, this will merge the images using a `OpenCV` algorithm and save to the `HDR` folder in the `outputs` folder; When disabled, this will return all images to the result section + - All the images are still saved to the `outputs` folder regardless +
-#### Offset Noise TL;DR +### Offset Noise TL;DR The most common *version* of **Offset Noise** you may have heard of is from this [blog post](https://www.crosslabs.org/blog/diffusion-with-offset-noise), where it was discovered that the noise functions used during **training** were flawed, causing `Stable Diffusion` to always generate images with an average of `0.5`. @@ -211,7 +222,7 @@ Though, the results may not be as good as using properly trained models.
-#### What is Under the Hood +### What is Under the Hood After reading through and messing around with the code, I found out that it is possible to directly modify the Tensors representing the latent noise used by the Stable Diffusion process. @@ -228,7 +239,7 @@ hence why you can control the brightness as well as the colors.
-#### Vectorscope? +### Vectorscope? The Extension is named this way because the color interactions remind me of the `Vectorscope` found in **Premiere Pro**'s **Lumetri Color**. Those who are experienced in Color Correction should be rather familiar with this Extension. diff --git a/javascript/vec_cc.js b/javascript/vec_cc.js index 53f4ed1..9a42a76 100644 --- a/javascript/vec_cc.js +++ b/javascript/vec_cc.js @@ -126,5 +126,11 @@ onUiLoaded(async () => { row2.style.alignItems = 'end' row2.style.gap = '1em' + // ----- HDR UIs ----- + const hdr_settings = document.getElementById('vec-hdr-' + mode) + const buttons = hdr_settings.getElementsByTagName('label') + + for (let i = 0; i < buttons.length; i++) + buttons[i].style.borderRadius = '0.5em' }) }) diff --git a/scripts/cc.py b/scripts/cc.py index 1952980..290089f 100644 --- a/scripts/cc.py +++ b/scripts/cc.py @@ -26,6 +26,7 @@ og_callback = KDiffusionSampler.callback_state class VectorscopeCC(scripts.Script): def __init__(self): clean_outdated('cc.py') + clean_outdated('cc_hdr.py') self.xyzCache = {} xyz_support(self.xyzCache) diff --git a/scripts/cc_hdr.py b/scripts/cc_hdr.py new file mode 100644 index 0000000..e1295c7 --- /dev/null +++ b/scripts/cc_hdr.py @@ -0,0 +1,109 @@ +import modules.scripts as scripts +import gradio as gr +import numpy as np +import cv2 as cv + +from modules.processing import process_images, get_fixed_seed +from copy import copy + +# https://docs.opencv.org/4.8.0/d2/df0/tutorial_py_hdr.html +def merge_HDR(imgs:list, path:str, depth:str, fmt:str, gamma:float): + import datetime + import math + import os + + output_folder = os.path.join(path, 'hdr') + if not os.path.exists(output_folder): + os.makedirs(output_folder) + + imgs_np = [np.array(img, dtype=np.uint8) for img in imgs] + + merge = cv.createMergeMertens() + hdr = merge.process(imgs_np) + hdr += math.ceil(0 - np.min(hdr) * 1000) / 1000 + + #print(f'{np.min(hdr)}, {np.max(hdr)}') + + target = (65535 if depth == '16bpc' else 255) + precision = ('uint16' if depth == '16bpc' else 'uint8') + + hdr = np.power(hdr, (1 / gamma)) + + ldr = np.clip(hdr * target, 0, target).astype(precision) + rgb = cv.cvtColor(ldr, cv.COLOR_BGR2RGB) + + cv.imwrite(os.path.join(output_folder, f'{datetime.datetime.now().strftime("%H-%M-%S")}{fmt}'), rgb) + +class VectorHDR(scripts.Script): + + def title(self): + return "High Dynamic Range" + + def show(self, is_img2img): + return True + + def ui(self, is_img2img): + with gr.Row(): + count = gr.Slider(label="Brackets", minimum=3, maximum=9, step=2, value=7) + gap = gr.Slider(label="Gaps", minimum=0.50, maximum=2.50, step=0.25, value=1.50) + + with gr.Accordion("Merge Options", elem_id='vec-hdr-' + ('img' if is_img2img else 'txt'), open=False): + auto = gr.Checkbox(label="Automatically Merge", value=True) + + with gr.Row(): + depth = gr.Radio(['16bpc', '8bpc'], label="Bit Depth", value="16bpc") + fmt = gr.Radio(['.tiff', '.png'], label="Image Format", value=".tiff") + + gamma = gr.Slider(label="Gamma", info='Lower: Darker | Higher: Brighter',minimum=0.2, maximum=2.2, step=0.2, value=1.2) + + return [count, gap, auto, depth, fmt, gamma] + + def run(self, p, count:int, gap:float, auto:bool, depth:str, fmt:str, gamma:float): + center = count // 2 + + p.seed = get_fixed_seed(p.seed) + p.scripts.script('vectorscope cc').xyzCache.update({ + 'Enable':'True', + 'Alt':'True', + 'Skip': 0, + 'Brightness': 0, + 'DoHR': 'False', + 'Method': 'Ones', + 'Scaling': '1 - Cos' + }) + + baseline = process_images(p) + pc = copy(p) + + imgs = [None] * count + imgs[center] = baseline.images[0] + + brackets = brightness_brackets(count, gap) + + for it in range(count): + if it == center: + continue + + pc.scripts.script('vectorscope cc').xyzCache.update({ + 'Enable':'True', + 'Alt':'True', + 'Skip': 0, + 'Brightness': brackets[it], + 'DoHR': 'False', + 'Method': 'Ones', + 'Scaling': '1 - Cos' + }) + + proc = process_images(pc) + imgs[it] = proc.images[0] + + if not auto: + baseline.images = imgs + return baseline + else: + merge_HDR(imgs, p.outpath_samples, depth, fmt, gamma) + return baseline + +def brightness_brackets(count, gap): + half = count // 2 + return [gap * (i - half) for i in range(count)] diff --git a/scripts/cc_version.py b/scripts/cc_version.py index 92788df..106e2f0 100644 --- a/scripts/cc_version.py +++ b/scripts/cc_version.py @@ -1,7 +1,7 @@ import modules.scripts as scripts import json -VERSION = 'v1.4.5' +VERSION = 'v1.4.6' def clean_outdated(EXT_NAME:str): with open(scripts.basedir() + '/' + 'ui-config.json', 'r') as json_file: