[Fooocus 2.2.0] Metadata, Isolation, Array Support, Docker

[mashb1t]

TL;DR

• Isolate every image generation to truly allow multi-user usage
• Add array support, changes the main prompt when increasing the image number. Syntax: [[red, green, blue]] flower
• Add optional metadata to images, allowing you to regenerate and modify them later with the same parameters
• Now supports native PNG, JPG and WEBP image generation
• Add Docker support

What's Changed

• fix: sort with casefold, case insensitive by @mashb1t
• feat: add early return for prompt expansion when no new tokens should be added by @mashb1t
• feat: ignore DS_Store by @charliewilco in Ignore DS_Store #2313
• feat: advanced params refactoring + prevent users from skipping/stopping other users tasks in queue by @mashb1t in advanced params refactoring + prevent users from skipping/stopping other users tasks in queue #981
• feat: add list of 100 most popular animals to wildcards by @mashb1t in add list of 100 animals to wildcards #985
• feat: add advanced parameter for disable_intermediate_results (progress_gallery) by @mashb1t in add advanced parameter for disable_intermediate_results (progress_gallery) #1013
• feat: add ability to load checkpoints and loras from multiple locations by @dooglewoogle in Add ability to load checkpoints and loras from multiple locations #1256
• feat: allow users to specify the number of threads when running on CPU by @maxim-saplin in Allow users specify number of threads when running on CPU via the extension to --always-cpu arg #1601
• feat: improve bug report and feature request issue templates by @mashb1t in improve bug report and feature request issue templates #1631
• fix: correctly create directory for path_outputs if not existing by @mashb1t in correctly create directory for path_outputs if not existing #1668
• fix: allow path_outputs to be outside of root dir by @mashb1t in fix: allow path_outputs to be outside of root dir #2332
• feat: add button to enable LoRAs by @MindOfMatter in feat: add button to enable LoRAs #2210
• feat: make lora number editable in config by @MindOfMatter in Add max loras editable config #2215
• feat: make lora min max weight editable in config by @MindOfMatter in Add configuration for loras min max weight #2216
• feat: add array support on main prompt by @flannerybh in Add Array Support on main prompt #1503
• feat: use consistent file name in gradio by @mashb1t in feat: use consistent file name in gradio #1932
• feat: add metadata to images by @mashb1t in feat: add metadata to images #1940
• feat: add jpg and webp support, add exif data handling for metadata by @mashb1t and @eddyizm in Jpg config image extension #1863
• feat: add docker files by @whitehara and @mashb1t in Add docker files #1418
• docs: fix typo in readme by @gteti in Typo fix #2368

New Contributors

• @charliewilco made their first contribution in Ignore DS_Store #2313
• @dooglewoogle made their first contribution in Add ability to load checkpoints and loras from multiple locations #1256
• @maxim-saplin made their first contribution in Allow users specify number of threads when running on CPU via the extension to --always-cpu arg #1601
• @MindOfMatter made their first contribution in feat: add button to enable LoRAs #2210
• @flannerybh made their first contribution in Add Array Support on main prompt #1503
• @whitehara made their first contribution in Add docker files #1418
• @gteti made their first contribution in Typo fix #2368

Full Changelog: 2.1.865...2.2.0

[poor7]

Syntax: [[red, green, blue]] flower He always draws in the same sequence, first red, then green, and then a blue flower and again. Is this how it should be?

[flannerybh]

Yes, that is the intended result. It's like a wildcard, but without the randomness.

[mashb1t]

@poor7 this is why i wrote "when increasing the image number", which you have to do to get images according to all entries of your array. Please increase the image number and check the output. NVM, misunderstood your comment.
@flannerybh thanks again for the implementation!

[bsdesigner]

"Generate Image Grid for Each Batch" does not work (((

[mashb1t]

This went under the radar, thank you for reporting. Fixed in #2414, please restart Fooocus and check again.

[bsdesigner]

It's all working!!! I'm working in Fooocus via Stability Matrix, so I updated a little late. This is a very useful feature when testing! Thank you very much for your promptness!!!

[Brassee]

Images no longer saves in 'outputs'

UPD. I found new folder 'outputs' in user 'download' folder. All freshly generated images are there

[mashb1t]

Please check the path_outputs in the config.txt file and the terminal output for irregularities on application start or error messages. If you think there is something wrong, please gather all information and open an issue.

[Brassee]

Everything seems ok now Thanks!

[patrickchamp]

Love the update.

I’m trying to cycle through wildcards or arrays using the same seed. I thought disable_seed_increment might do it but the images are all identical. Can you clarify what this checkbox does? I think I just read that “Generate forever” on a batch of 1 might cycle through wildcards randomly but can’t use it atm because I’m on an iPad.

[patrickchamp]

Update: It was ignoring the array because one of the values had parentheses. I removed them and it’s properly working for arrays. Ideally, the seed increment would have the same effect on wildcards. In order to test values against the same seed, I had to take a wildcard list and convert it into a comma separated list.

[mashb1t]

@freshp124 wildcard read in order has not yet been merged, but is in this PR: #1761

[saadkhan955]

Is there a way to use multiple arrays? For example, could I create an array like A [[chair, table, lamp, bed]] and another array [[wood, iron, steel, plastic]] to generate the image? I tried a method like this, but it iterates through the first array and only selects "wood" from the second array.

[flannerybh]

That should work, can you post your full prompt? Remember to set the number of images to 16

[saadkhan955]

Thanks, it's working now after setting the image count to 16. I'm wondering if there's a way to select only one property from each array for each image. For example, on the first run, the prompt should be "A chair made out of wood," and on the second run, the prompt should be "A table made out of iron."

[flannerybh]

[[ A chair made out of wood, A table made out of iron ]]

[poor7]

How to load metadata from a saved image back into Fooocus?

[mashb1t]

Input Image > Metadata > Apply Metadata

[poor7]

If selected Output Format=jpg or webp then Save Metadata to Images=a1111 (plain text) does not work.

[eddyizm]

lol holy crap. I always use PNG and i thought for this feature the jpg would be ideal since I am very familiar with exif (photographer) so I literally shot myself in the foot and was totally wrong about it.

Ok so maybe I will put in a small pr to put that note in the label?

[mashb1t]

Sure, feel free, would do it tomorrow anyways!
Adjustments in 2 places, info and en.json

[eddyizm]

ok done.
#2434

[eddyizm]

Another question on the metadata, I added about 5 loras to test the civitia integration and although feeding it back to fooocus there is data and hashes regarding the loras used, civitia only sees the base checkpoint and no lora info at all.

Just want to verify if this is the expected outcome.

[mashb1t]

Yes, this is expected. Civitai can not read LoRAs from A1111 scheme, even when generated with A1111 (just checked with v1.8.0).

[eddyizm]

Interestingly this morning I was testing it out and found something a little odd.
@mashb1t I believe you have mentioned elsewhere to backup the config.txt elsewhere
I loaded the update to a new machine with an existing install and gave a few lora a try with some of the new features enabled.
The output was a little odd so when I checked the log, it seemed to have enabled two Fooocus styles, which in my preset are disabled.
"default_styles": [],
Nothing that I can see in the config though so still tracking down how this happened. Gonna spin up a fresh session on google colab to see if I can reproduce it.

[eddyizm]

ok, after a few experiments, It looks like these values get set when using the describe feature, unbeknownst to me,

Fooocus/webui.py

Line 659 in 4ea3baf

def trigger_describe(mode, img):

A little alarming when I wasn't expecting it but apparently not really a bug. I have not really ventured too deep into all the levers and buttons Fooocus has to offer.

[mashb1t]

2.2.1 is planned to be released tomorrow, content: https://github.com/lllyasviel/Fooocus/milestone/2?closed=1

[mashb1t]

=> https://github.com/lllyasviel/Fooocus/releases/tag/2.2.1 / #2440

[jaimeastorga2000]

The last two releases have not had binaries, only source code. I know existing installs auto-update if you have an internet connection, but can we please get the binaries as well? I know I would have never tried Fooocus if I had only seen source code on the release page.

[mashb1t]

The 7z file will be updated on package updates. As the last few releases didn't have any, there are no 7z files attached. FYI each 7z file is 1.7GB+ in size.

As you've mentioned correctly, you can simply download the code and use the auto-update or download the 7z file from the readme + extract and run it. After updating you can bundle the code again if you need it bundled in the latest version.

[eddyizm]

So not sure if this is already addressed some where but I have been seeing this error in my log a few times.
generate_image_grid = task.args.pop(0) IndexError: pop from empty list

[mashb1t]

This has not been reported and is also working fine in main and develop as well as my and your fork. Please create an issue if needed.

[andynormancx]

I've also started getting this, happened twice today so far. I've had to restart fooocus when it happens.

[poor7]

Is it possible to add additional metadata directly to the file properties, in particular in the “Comments” section, so that you can view them without using Fooocus? It would be much more convenient than now.

[mashb1t]

please open a new discussion or feature request with more information about what you're suggesting.