#OctoTex - Texture Remastering Tool
1 messages · Page 5 of 1
Теней как ты и сказал не хватает
но выглядит всеравно лучше чем плоская текстурка с пбр
Попробуй сгенерировать глубину и шереховатости нейронкой, без анбейкера
А нормаль уже обычным алгоритмом
Это и так оно и было
.
Без анбейкера
Анбейкер у меня не работает)
Нормаль алгоритмом
Аааа
Ну тогда эф
Реально выглядит странно
Будто бы нормалей нет
Ты использовал legacy?
Блять я удалил старую нормаль от нейроник
Я говорю попробуй legacy нормаль сгенерировать
они вообще не видные
аа
ок
ток уже не буду соря
н
Лан

вот тебе смешно
а я даже удалить ничего не могу
только если работать на лабе на все 100%
но я хочу дебаг локально производить(
Country roads
take me home
to the place
i beloong
бл
а какой там порядок то был
west Virginia
mountain mamma
take me home
country roads
smol
было меньше
paper, metals, painted walls, bricks, tiles, asphalt, cardboards, carpets, chipboards, clays, concretes, grounds, fabric, gravel, granite, lether, stones, rocks
alot of things
planks, wood
snow, sand, signs
and every texture is rotated randomly
like so
with coordinates saving ofc
Как по мне знаки не должны быть с нормалями
?
уж извините
Before and after
Ок
ОК
А я новые нейронки хреначу
Да, Я понял
не точностей
Датасет маленький был
Но за POM спс
Щас побольше, намного
Ну посмотрим
Думаю будет шик
Also some other POM
Если хош я могу в твой соник несколько текстур от пбр ить
или ты все нейронкой хочешь?
Не спасибо
Хочу нейронкой
При этом я написал скрипт который их все попереворачивал, чтобы не было схожестей у тех же кирпичей
Я всего лишь чел из коммюнити
Норм
Да понимаю
Я весь ambientCG выкачал, и ещё сверху докинул
Не ну это классно не спорю
Может ты заметил, но нормали текущая нейронка генерирует странно
Они по оси Х размытые будто бы
Не детектит тупо
Это как раз таки из-за параллельных горизонтальных линий одинаковых на куче фоток
Теперь они все перевёртыши
И разные
И их в разы больше
Щас нормали на ночь оставил
Потолок поставил 500 эпох
А текущая нейронка с нормалями около 170 эпох только
Крч, во всех направлениях попытался улучшить
Посмотрим что будет
Посмотрим
Но будет лучше
Посмотрим крч
а более сложные б были
Но будет лучше однозначно
Верю
Чем текущая модель
На все 100
Если на все 100
Я уверен
Причем это не просто алгоритм как Legacy
Поэтому из одного лишь цвета можно вытянуть больше
Мне ещё нужно тренировать шереховатость, металлик, эммишен, РОМ
но порой выдает очень неплохие результаты
Шерох потрень норм
а то страдает все же
артефактами
500 эпох так же накину, не бойся
и порой не точностями
И ещё новые модели
А вот эммишен уже рил спорно
Металик нужен намного больше чем емишн кнч
Я честно хз какой результат будет у эммишена
но не настолько\
И будет ли он вообще работать
его как по мне проще самому сделать
Угу
Там по большей части фасады зданий
С включенным светом в окнах
Ну и потолки с лампами
Хз крч
Спорно
Но вот металлик нада
ну они единичны по сути
Ну я например металик далеко не везде использую
из 15 текстур всего 3 металика
Всё равно, если это метал, его нужно отрендерить как метал
Вот только
да
но а как быть с тем
что на текстуре есть как метал так и не метал?
Уже всё учтено
Вот тут лично у меня весьма спорно
Там в датасете даже есть текстуры дерева например, с гвоздями, и там метал только гвозди
Но никто не отменял текстуры с похожим цветом
Старые текстуры
вдруг там будет часть метал часть стены
Посмотрим что выдаст
Посмотрим
лан
и долго еще?
Поставил майнер на комп начальника 
Ну на комп который он тебе дал хз
У меня 3080
Понял
По сути он дал i5 10400f, rx580
Ну и сверху 36 гигов озу лан
Хорошо работать программистом 👍🏼
🫠
Хуета крч
3080 моя по сути, но мне её тупо ставить некуда кроме офисного пк
Думаю купить Рязань какую-нибудь
Ryzen 7 5800x3d например
эт из новых?
Нет
не жирно ли?
Новый не буду покупать
А почему нет
Потому
Когда перееду в немецию, там уже буду собирать топ из топов
Но это уже потом
немецию
По максимуму
Ich verstehe Sie nicht
А должен
Хехехе
Шучу, никому ничего не должен
Жена немка, вот и хотим уехать
21 💀
Oder in deutsch, ich bin einundzwanzig Jahre alt
Всм?
Про свадьбу?
ну типо
Да я сам не думал что женюсь так быстро
Но жена ахуенная, готовит невероятно, любит, всегда выслушает, и ещё такая мягкая, хаха
Я помню ты по моему говорил что из Казахстана
Казахстан
хах
Ну да
Ещё и живём отдельно
Хаты нет своей правда, никто из родственников не оставил даже надежды на квартиру, ни с моей стороны, ни с её
Приезжай, выпьем чаю
А ты откуда?
Украина
А у нас в КЗ либо родственники, либо никак
Та если б
Ого
Тут п.здец творится
В каком плане?
Каждые два месяца цены поднимаются всё больше и больше
И ещё куча наитупейших законов появляется
Бедная страна
По сути много драгоценных ископаемых, но вот только все залежи распродали
А щас вообще безвизовый режим китайцам нахуярили
по моему так же
Думаешь у нас её нет?
У нас всё очень зупущено
Большая часть зп у человека уходит на еду, слава богу если у него хата своя есть, иначе он будет тупо бомжом
а украли 100
Средняя за в месте где я живу, 90к тенге
это сколько в руб?
А у какой-нибудь уборщицы или сторожа вообще 60к
или в долларах
Это на месяц, на еду, коммуналку, и другие траты
Бумагу туалетную купить, мыло, шампунь
Снять хату за эти деньги нереально здесь
Хата которую мы снимаем стоит больше
А это однушка 💀
Это пиздец конечно
оказываеться тенге еще дешевле чем рубль 💀
Моих деньгах это где то 8к грн
А ты кем работаешь?
Я уже сказал
Программист
Системы виртуализации
Фулл стек веб программист, иногда девопёс
По меркам рашки конечно не сверх, но по нашим, нормас
Но у нас тут железо пиздец дорогое
Даже в Германии дешевле
С их зарплатами блять
Понятно
@left orchidanyway commands are not required he added batch script
run.bat doesn't include downloading imagemagick or the models
README just tells you to run webui.py
Models link is there below👇...imagemagick is an application required to be installed seperately...
yep. all solvable things, and all asked about in this channel already. But if you have a github already for the project, keeping the setup instructions on there up to date and complete is a good idea
I've only ever heard issues of people using the Windows Store python, it's better to just download python from their site IMO
this was more along the lines of me not knowing that I'd need to run pip install using requirements.txt, or that I would need to get imageMagick separately. I was just trying to follow the github readme
I didn't created this tool...I collected every half information and made this beginners guide...I already told about this to Alex...and there is more you need to know...first I have to check if he updated the requirements.txt
So requirements are updated yeah...now command is not needed running batch file does that all...plus imagemagick is required...
Then #1146726031022108752 message
oh I've got it all up and running, as far as I can tell. waiting on it to finish processing a batch of textures right now
but my main feedback about the overall process is that it's best to just keep the github readme up to date in terms of setup and running instructions, instead of relying on discord pinned messages
Exactly my concern is also same...he forgets plus he is constantly working on it not making guide...
I had a nightmare finding all the hocrux
@surreal anvil Sorry for the ping, 1 last tip
- AI creates better Roughness and Displacement .
- Legacy creates better Normals and Metalness
Alex is more focused towards tuning AI models while getting the whole app features together.
I'm working on a house cleanup branch to make setup easier, especially for ppl with no experience with python at all. Think I'll take some time to work on a professional python project readme doc as well so more ppl can setup and contribute 
If you're only ever using the same models all the time, I'd recommend converting them to onnx and using onnruntime-gpu instead of pytorch. It's a much smaller dependency. And if you wanna make the setup even easier, you could use one of the many python bundlers that can package everything into an exe. I wouldn't recommend doing that if you're still gonna use pytorch tho
Thats is pretty good. Cutting down dependencies surely help. Though packaging as exe might become a pain when dealing with other complex libs like openUSD. Kim had some issues afaik.
I'm working on a method to package everything pretty much like blender or omniverse apps do, including everything installed ready to run without requiring users to install or even know about python's existence 
What's the intended usage when applying Octotext to multiple games? Is it supposed to be installed next to the rtx-remix folder in each game?
In the windows store version the PATH is not set, u have to manually set the PATH in some "Variable" Place but I did try that and I was still getting errors
Yeah that sounds about right. Do you mean you tried to install python from the site and still got issues?
Or that you added it to the path and still isues
Yea I tried to add PATH to windows store version of Python 10 and it still gave errors saying this module is not set in a PATH or sumthin similair in cmd
Downloading python from it's official site gives u a option to tick on called PATH inside the installer so it's better to download from there then the windows store
Just wanted to make sure, yeah I avoid the windows store apps like a plague
Is it up now 😮
nope
im training roughness now
but goosh
it's so much better that the current model
the normals was trained 360 epoch
with much better dataset
now im interested how will emission and metalness maps will look
Newer versions should import just fine. The omniverse version of blender works better, has more options, and is on 4.0 already
even newer?
Hmm weird. Maybe the linux version doesn't ships with usd libs
ok my exporter is broken
normals are dying
kinda works
sheeesh there's a black hole
and rtx-remix does not captured every texture
idk why
now i need to translate them
according to the positions in the capture.usd
make sorta blend file or something
that supports layers
like that
That still doesn't 100% seem right
This looks like the old model but it at least added some vertical lines
It's closer to your legacy stuff but seemingly not near the sharpness I'd expect for the components between the bricks
I guess that's where the slider comes in though
Basically what i'm saying is that the vertical lines should match the depth of the horizontal lines
yes
yeah ik
but the model don't like the red color
Can you try this texture and see how it looks?
yup gimme a sec
This is from Dark Messiah, the normal should be close to this
It's a good reference to know where it's weak. Last time it could not see any vertical aspects of the normal at all which is a great improvement, so hopefully it helps you refine more
do you have any other textures?
Oh wait
i've applied unbaker
and situation is better now
still not perfect
but better
with unbaker
Just trying to give you some odder textures to help your training
THAT looks good!
okay
Но все же эта вышла неплохо хотя алгоритм сделал бы так же
именно на этой текстуре
шарю
но нам такое нинадо)
я знаю шо можно просто взять и в пеинте или фотошопе нажать
на инверт цветов
у нвидэ при переводе в octohedral в классе для питухона перевод идёт из dx стиля
поэтому сё
цыц
а для ogl есть функция для перевода в dx
Вот тебе еще задачка
ну хз
не знал
да и похуй)
ну, а говоришь что знаешь
meh
BRUH

Is ai producing better normals now ? and what about Metalness ?
I'm training roughness now
And normals are weird
Sometimes they're amazing
And sometimes they're not
First one is good
But the second, naah
It's the same ai model, but the results are weird
hmm
keeep it up
roughness was alredy good...more better hell yeah
They'll be even better
And pom
And new models for metallic and emission
👀
will this still be limited to legacy
I'm talking about ai
Not legacy
Legacy does not have ai models
oh nice so ai will now make metal too great
And emission too
impressive...just make the normals better
I'm already working on it
Don't worry
and update the readme in repo
I'll update
Wow I've found screenshots of my old ray tracer for unity
Yeeeah
That was a gtx 950
I also had a gtx 950 💀
gui wher
GUI n gihab
hah, same. loved it when i had it though
There is a weird image generation artifact where the image texture is coarse with criss crossing patterns
yeah i see
i did not create the architecture so
It affects the roughness model negatively as there's a distinct line where roughness values are not correct
yeah i see
Like along the edges of the blocks
can you send me a texture for the test please?
Sorry I don't have a texture lying around
oh ok
Hi 🙂 I got this error and I don't know how to fix it
the texture is too fat, or stretched
👍
You need to remove comments on line 23,24 In OctoTex/ai/PBR on the eval files
Can you send a picture of what it should look like? because i think i misunderstood
Should look like this, in all of these files
it worked! thank you 🙂
training metalness AI
off/on denoising
Хмм выглядит неплохо но деталей не хватает
bro whens the update coming ?
nice
немного странные отражения сбоку
ты лучше сверху покажи
там вроде нагляднее
хммм
даже незнаю
вроде бы и неплохо но что то не так
мб и алгоритм не справился
Я имею ввиду
и если бы я через алгоритм делал
типо и алгоритм не оч справился
idk if that's how it should look
hmmm
nah not legit
my first mesh replacement
yaaay
i mean
noooo
it... works?...
💚
ai/PBR/eval_ files
There's is a commented lines that will crop the texture, and center them
You need to uncomment them
я уже сделал это, текстуры создаются
но прямоугольные текстуры всё ещё не работают
Как раз таки по этой причине
Потому что ты разкомментировал
Нейронка не принимает такие текстуры
Будет ошибками срать
печально, в мафии большинство текстур прямоугольные, надо прям искать ракурс для скрина, чтобы было незаметно, что всё сломано
Земля пуховиком
Щас особо нет времени
pip install Python-IO
in the requirements.txt wrtten smaller "io"
i installed...i think i already ahad installed but still i have error
octotex is opening
Ok I will pause the development for a bit, because I need to work harder now, to pay for the rent, food, etc, cuz I live in a shitty country with extra large prices and low salary, I rly have no time, but I will accept the commits, thanks ❤️
sure take care
now octotex working now... also had to uncomment from the eval files
where do i find mod.usda
it will be in rtx-remix folder after you generate with octotex
im struggling to even install octotex haha
commands are not necesary there is a batch file included that will do the rest
and download new models from github repo
yup
how do you update it
Same as you installed first time...down from the repo and models...then run the batch that's it
Damn that’s a pain
Any ideas???
My Python is the Windows Store version 3.11 as I'm assuming it's supposed to be
I did what I could too install the dependencies one by one too see if that'd fix this and I got closer on loading the webui itself but it's saying I'm missing a cv2 dependency I can't seem to install.
i think it was said before that the windows store version of Python causes issues
Oh okay, the guide specified too use that version so that was my mistake. I'll install the original version and see if that fixes it, thank you.
pip install Python-IO
I tried that and I'm still getting the errors
Ran it again and it's confirming they're installed
same ahppened to me after i re opened it worked
Maybe a lucky reboot will fix it, running the .bat one more time and if that doesn't work I'll restart my PC and try.
im on 3.10 which is recomended for usd core
Okay good too know, it does seem like it's a Python version issue
Still getting the same issue, downgraded and it seems too run smoother but even after reinstalling the Python-IO it's still not running.
dont try to run octotex with batch having requirements...just remove first line
The furthest I got it was saying it was missing the cv2 dependencies.
Now we're back too this
i dont think you are using the lastest version of octotex
Damn I think I was on the wrong branch entirely. I didn't realize there were two branches on the github
and downlod the new models too
Same issue after downloading the main branch instead of master branch, replacing the checkpoint folder inside of PBR, and directly running the webui
Not sure where I'm going wrong. At the end of the Python-IO installation it is suggesting upgrading it but I'm assuming that's not correct.
Also double checked too make sure all the dependencies were still installed and everything is fine up untill the end with the IO dependency.
dont replace the with old version
I didn't, I rid of the master branch folder and redid the steps with the main branch folder but I don't think there's an actual difference.
I made sure the models I had were the latest on the GitHub aswell.
is web ui opening ?
Not at all, the same error as my previous screenshot
everything's is here go above in the chat and find out or check pins...u can also type the word in the search to check reelated chat
Nothings helped and I'm just getting pissed off at this point so I'll leave this here I need a break from this bs.
upgrade the pip
I have and it seems too break it more
u are doing something wrong why its saying requires python 3.7 or 3.10..check ur version
I've installed 3 different versions at this point and none of them worked.
3.7
3.10
3.11
None worked.
I uninstalled all but 3.10 since you had success with it
did u followed the instructiin th pin section
Yeah that's exactly what I followed.
i cant help its working for every1
download python 10, when installing it tick the PATH option and thats it, just install every thing in the requirements.txt using cmd(i installed all of them separately like pip install (module name) just in case, also remove this in the requirements.txt torch --index-url https://download.pytorch.org/whl/cu118 cause for some reason it doesnt install it correctly if u just pip install requirements.txt -r
cv2 is the wrong name real name is opencv-python so type pip install opencv-python in cmd
also pip install pillow
are you liking the new model ? because i dont
havent used the latest version mch just installed it today and only remastered 1 old cs 1.6 map, used ambient cg textures most of the time today for testing
u can put them on breakable glass addon by Mee and i just made breakable gold block with reflections dang
dont delete the old models before you regret...and try it out then i would like to know your veiws on it...amiant cgi is good addon
i deleted it 💀
i have it
I'm midway through installing them one by one but does the pip version matter? I reinstalled Python online 3.10 specifically and ticked PATH. Everything seems too be going well thus far just worried that it suggests upgrading pip version.
should wrk in any python 10 or 11 versionn i have Python 3.10.11, had python 11 before but downgraded for usd core
Alright, thanks for the help both of y'all I'll get back with the results when I've installed everything.
I'm excited too get this running because I wanna try and at least PBR these Cry of Fear textures for these GMOD ports, I think it'll look sweet.
It worked! I had too sort out some system issues after installing it through the Microsoft store but I got it running! Thanks so much for the help!
you didnot gave your review on new models
I used it for some Gmod Props, they look fine and good
i dont know why am not satisfied
Regarding Imagemagic, is the latest version here fine? https://imagemagick.org/script/download.php
yes
I know I have the CD of this somewhere... 👀
@mental wasp could you update the guide?
@left orchid ded
Actually
There is nothing to update...what's the issue you are having ?..btw i am sick 😷 from past week so not much using phone and pc
Oh those instructions are already pinned
not really the intended purpose but I have been remastering ps2 assets using this tool
just editing the textures so I dont get in trouble lol
so is the upscaled texture supposed to show up in the webgui?
I ran the upscaler but it was weirdly fast
Yeah that's a bug
ahh gotcha
The texture will not show up as upscaled
I've broken something in the latest updates 😛
there were a bunch of my textures that were just invisible as well
I dont think its a bug with your app but I think it might give me a better understanding as to why they are not showing up in remix
if you got a sec to help @left orchid that would be great
ok, I have found something weird/wild
Lemme guess, you have a lot of the textures right?
You can use the PBR stuff if you want, my models are free to use/sell
They have no restrictions lol
Oh that's a ps2
Then there's no point to use PBR I think
oh thats a seperate thing
my laptop is dying one mo
here we go
@left orchid would bit depth affect this?
Idk really
Hm
Idk
all good man
You've got it running?
Oh
ah they are jpgs
ok, this is making more sense
since an image is made of R G B A
8bits of each. So a 24bit image is an image without the transparency
just RGB
Oh, makes more sense now
I still dont know why the image is showing as a black square though
do you know of a way to batch convert jpg to png? or more specifically png32?
paint does this pretty nicely
Jpg have no transparency isn't it?
yup
If it was my task, I would write a Python script or something
does jpg work in your tool?
Nope
coolio
It checks only the PNG images
You can edit a script a bit
I think it will be able to process the jpg images
gotcha! thnks
is it the main.py?
ok, I am assuming I just switch out ('dds','png') with ('dds','png','jpg') in load?
welp I think it worked
lol nope it did not.
getting srt cannot be interpreted
replaced the .py files and it worked /shrug
getting frequent cuda out of memory errors, i have a 4080 upscaling by 8x. this normal? should i upscale to 4x or try to change to fp16? if so, how do i do that?
Do this your issue will be fixed...#1146726031022108752 message
Yeah i fixed it, kept getting issues with writing the mod tho. Some missing material
Went back to a september version
anyone know what to do with the tensor error?
oh it needs to be the power of 2
weird, my upscale with AI seems really quick
I've fixed that already
But I did not commited to GitHub
Do you recall which version you went back to? I have been plagued with that issue as well. Thanks!
Any chance we could get a pause and resume feature for the mass processing functions? It takes a really long time to upscale hundreds of textures to 8x on my 12GB 3060.
I picked the last september 27th or 25th build
Error handling request Traceback (most recent call last): File "C:\Python310\lib\site-packages\aiohttp\web_protocol.py", line 433, in _handle_request resp = await request_handler(request) File "C:\Python310\lib\site-packages\aiohttp\web_app.py", line 504, in _handle resp = await handler(request) File "E:\Games\Mods and Files\OctoTex\OctoTex-main\OctoTex-main\webui.py", line 195, in processing_routing reqType = req_types[ fileType[len(fileType)-1].split("?")[0] ] KeyError: 'ini' anyone get this error?
Now with the luminance on the road, I think I need to add a feature to pack the mods for uploading
hey - running into an issue installed Octotex, downloaded the github repo last night:
File "E:\Odessa\ai_testing\octotex\webui.py", line 6, in <module>
import functions
File "E:\Odessa\ai_testing\octotex\functions.py", line 3, in <module>
import load as loader
File "E:\Odessa\ai_testing\octotex\load.py", line 3, in <module>
from tqdm import tqdm
ModuleNotFoundError: No module named 'tqdm'```
have followed the instruction to install from the requirements.txt and run python webui.py, but no luck
I did have a pip update I needed to do so I'll try running that and then running the requirements.txt install again
pip3 install tqdm
hello there btw
that fixed that one, but I'm now also missing cv2 - I assume I can run it the same way, but it looks like some things might be missing from the requirements.txt for initial installation
thanks for the quick response though 🙂
putting pip3 install cv2 trhwos back a ERROR: Could not find a version that satisfies the requirement cv2 (from versions: none) ERROR: No matching distribution found for cv2
pip3 install opencv-python
yeah, i was lucky to check the discord messages
wow, that's a deep error
i think you have a .ini files and tried to access it via browser, but the webui server doesn't know about ini type of file
strange error
i'll try to fix it
was also missing torch, but have installed that one with pip3 install torch
yeah, the req file has a torch, but it installs cpu one
missing basicsr too
for some reason
yeeeah, there's alot of the missing packages
btw i've broke something in the latest updates, so the textures will not be updated in real time after upscaling
realesrgan also didn't come in through requirements
I'll make a list when this is all done so it's all in one place
missing packages:
tqdm, opencv-python, torch, basicsr, realesrgan, wand, clipboard, aiohttp
other missing dependencies:
imagemagick
but I've now got it working 👍
is it possible to run the tool without having created a capture? i.e. just basically treat it as an upscaler that generates PBR textures, and then write the changes into a Remix mod manually instead of the tool having to be hooked in first?
yup, but only for png files
that's fine - I've put some PNG files in the "diffuse" folder in textures/processing, but getting this error on the website:
it will try to create an OctoTexGUI folder inside of mods folder whe you press a write button
two dots at the name
oh
not sure where the webgui is getting this name from
it shows the error when you trying to upscale it?
yeah
yea, its real odd
oh I think I know what happened
you know how in google drive it likes to make a hidden desktop file? I must have moved one in there by mistake
can you open a dev menu in the browser, open the network tab and try again, it will record from what function the error comes out
oh, i don't use the google drive that often
sorry 😦
how do you open the developer menu?
ctrl+shift+i
ah, the inspect element tab, right
it will look like this
you'll need a network tab
this is what happened from the attempt to upscale again
will be fixed in the new commit
Thanks Bro
@left orchid it must have been one of these desktop.ini
oh
[.ShellClassInfo] ConfirmFileOp=0 IconResource=C:\Program Files\Google\Drive File Stream\82.0.1.0\GoogleDriveFS.exe,22
not much going on but it must have borked it somehow
Try the latest version now please
fixed for ini files
alright, I'll try a redownload
btw strangely all the dependencies are already in req file
did you tried pip3 install -r requirements.txt?
yep, that's the first thing I ran
did you downloaded the main branch right?
yep, top-level result on github, downloaded the zip
huh, i dont have any idea then 😅
wait i check something
yup
prev commits didn't changed anything
awesome thanks
except for the --index-url lines
does it works now?
when I installed this myself I had to go through one by one and download the requirements. so not sure if I did something wrong or if there is an issue with the text files
cant check until I get home I am afraid
yes, I will check as soon as I am home
god I forgot I left the washing in the machine fuu
I think I saw this error earlier
oh nevermind then haha
Why your folder named like Odessa? 💀
not really relevant
just replace this folder
if there is one suggestion I would have for this. It would be to have some sort of progress bar in the gui. I noticed that it has one in the CMD window but only for some functions but I would love to have one in webgui or even a console function. some textures I have are pretty complex and take an age
just helps me narrow down if I broke something or not haha
the changes are on github also
ok, now I don't get an error, but nothing seems to happen. this is the texture I put into the "diffuse" folder and then click "upscale 4x"
this is the texture that gets exported into "upscaled"
and no errors?
no errors, no
can you send me the file?
ah, thanks
keeping in mind I'm not doing this with an RTX Remix mod installation, I'm running it totally on its own. no captures, not in an RTX directory, I just want to process these textures on their own and separately
wait wait
you placed the texture, and then pressed upscale right away?
yeah
let me guess, you have no the texture in the right menu yeah?
put pngs in the diffuse folder, they were picked up by the webgui, tried clicking upscale on the webgui
ah
I do, they're all in there
interesting
what I did to test the upscale function was to make a bat file using just upscale.py
then I found it worked and from there I used the webgui
try using the materials tab and upscale from there
see if its any different
now the button works, but nothing gets placed into the "upscaled" folder
@left orchid I also noticed that I had issues using the batch function on the first tab
give me all the textures please
attempting to click the "generate normal map" button comes up with this:
same with roughness
ah yeah, you will need to remove the torch
the textures are all the same format, size, and from the same origin point as the one I sent you, so one working should mean they all work, and vice versa
and reinstall it with cuda support
i mean, maybe the combination of the names is messing something uo
idk
oh I and dont forget that if the uninstall command doesnt work to try using pip uninstall torch rather than pytorch
clearing hashes still produces nothing
the legacy "generate map" options are working, but the upscaler is not
if you run the requirements script does it still say you are missing some?
does your path to the octotex contains some unicode? utf-8 or other symbols?
in my enviroment I am using python10, and had to manually grab the requirements
ai model also needs to be ai/PBR/checkpoints
