Not 100% sure since I've long since then messed around more with the paramteers, but it was either Gonzales_v8_newCocktail or Gonzales-NSFW-PonyV1/V2-DMD they're both very fast Pony models, it would've been around 8-12 steps, cfg scale of around 1.3-1.5 and 0.3-0.35 noise.
The same problem is still present
Can you tell us what model you are using here?Thank you so much for this! it worked amazingly, using img2img and a low noise value really does wonders for keeping things appearances consistent and the poses "sane" while still getting some nice improvements.
I figured I'd do my part and help out a bit as well by trying to remove the square background, which proved to be both easier and more annoying than I thought, mostly because the way I started doing it (using the rembg extension) just refused to work, but after I finally switched to another extension instead, it was incredibly simple so now I'm beating myself up over all the hours I wasted working on the wrong solution
View attachment 4895044
I've attached my modified sdcclient_ren.py file, but all I did was literally to add this line to the API-request:
'"script_args": [true, false, false, "#000000", false], "script_name": "abg remover", ' + \
There's no settings to turn it off or anything, so if anyone tries to use this, you have to make sure that the ABG extension is installed.
View attachment 4895045
Ok, I got it to work, but had some issue in SD Forge - turns out "Write infotext to metadata of the generated image" needs to be turned off in SD Forge settings, as it messes with the alpha channel of the output image. Also the "File format for images" has to be set to "png"Sadly this version isn't without issues either, if the person is wearing black, they can sometimes become transparent. I've experimented with changing the saved image to have a purple background instead, seems less likely to get mixed up with the color of clothing, but that leaves people with a purple halo instead. Honestly not sure which version is better
for purple halo-version, it's easy enough to change from
toPython:image = person.build_person_displayable(position, emotion, special_modifier, lighting) surface = renpy.render_to_surface(image).convert()
in the SDClient_ren.py file. Or you could just change #ff00ff to what ever color code you prefer, guess you could even use the persons designated color to give all of them a "personalized" halo similar to the original aura-thingyPython:image = person.build_person_displayable(position, emotion, special_modifier, lighting) newimage = Composite(position_size_dict.get(position), (0,0), Solid("#ff00ff"), (0,0),image) surface = renpy.render_to_surface(newimage).convert()Then it'd be something like this instead:
But I guess, depending on what display color they have, you'd once again increase the risk of the background color melting into the color of their clothing, causing the background remover to get confused.Python:image = person.build_person_displayable(position, emotion, special_modifier, lighting) newimage = Composite(position_size_dict.get(position), (0,0), Solid(person.get_display_colour_code()), (0,0),image) surface = renpy.render_to_surface(newimage).convert()
I really wish I could've gotten rembg to work since that seems to be a more flexible option for background-removal than the what I'm using now, but while rembg worked on it's own, combining it with ADetailer made it break (rembg would always be called first and remove the background and then adetailer would run after that and mess things up by removing the alpha channel and creating a new, wonky-looking, background instead).
I've tried it with several different models and it's worked with all the ones I've tried, the background remover script doesn't remove any specific color, it can even remove the background if it's a "real" background image, it's been removing everything fine(ish) for everything I've tried. Are you using Automatic1111 or forge or something else?
i did see something about Layer Diffusion but all examples I found on it was only in forge, I didn't find anything for automatic1111. Since I already have ComfyUI and Automatic1111 installed I had hoped to find something that works with what I have installed and not have to install yet another way of running SD, I'm starting to run low on disk space
This is joiplay issue, or whatever you use to run the game on ur mobile, not the game.
It turns out this was all user error on my part. Disregard.I've tried it with several different models and it's worked with all the ones I've tried, the background remover script doesn't remove any specific color, it can even remove the background if it's a "real" background image, it's been removing everything fine(ish) for everything I've tried. Are you using Automatic1111 or forge or something else?
I'm running this with an automatic1111 installation so can't really comment on other options but If you're using forge, it might be worth having a look a ycbelane's latest response:
i did see something about Layer Diffusion but all examples I found on it was only in forge, I didn't find anything for automatic1111. Since I already have ComfyUI and Automatic1111 installed I had hoped to find something that works with what I have installed and not have to install yet another way of running SD, I'm starting to run low on disk space
You'll need to downgrade your Renpy plugin to 8.3.2. Make sure to delete 8.3.3 plugin from your APPS and reinstall 8.3.2.
I'm sorry, but an uncaught exception occurred.
While running game code:
TypeError: __init__() takes from 3 to 4 positional arguments but 6 were given
-- Full Traceback ------------------------------------------------------------
Full traceback:
File "game/people/Ashley/ashley_story_ren.py", line 28, in script
init 5 python:
File "F:\MEGA\LabRats2-Reformulate-2025.04-market\renpy\ast.py", line 834, in execute
renpy.python.py_exec_bytecode(self.code.bytecode, self.hide, store=self.store)
File "F:\MEGA\LabRats2-Reformulate-2025.04-market\renpy\python.py", line 1184, in py_exec_bytecode
exec(bytecode, globals, locals)
File "game/people/Ashley/ashley_story_ren.py", line 253, in <module>
ashley_story_path_submission = Story_Path("Submission",
TypeError: __init__() takes from 3 to 4 positional arguments but 6 were given
Ren'Py 8.3.6.25022803
Lab Rats 2 - Down to Business 2025.07-betaVTMod4.0.31Did you use Unren to decompile the AIO version? Only time I've ever seen that kind of error. Unren doesn't work on this game. (At least all the versions of unren I have seen). Best to get the source code directly from git if thats what you want.
Why do you need save? You'll miss the events that are one time and non-repeatable.
