Thanks for that!
I actually do have my system temp directory on a separate NVMe SSD for the reasons you mention.
DE-censoring JAV with Free Open source LADA application
- Thread starter avatarthe
- Start date
-
Akiba-Online is sponsored by FileJoker.
FileJoker is a required filehost for all new posts and content replies in the Direct Downloads subforums.
Failure to include FileJoker links for Direct Download posts will result in deletion of your posts or worse.
For more information see this thread.
I actually do have my system temp directory on a separate NVMe SSD for the reasons you mention.
I gave it a try with my RTX 4090. It took 1 hour 12 minutes to do a 1 hour 53 minute 1080P movie (APNS-395).
Can't say I'm super impressed. It seemed to do ok if the censored bits were near the middle of the screen, but it had quite a few problems when the cencored bits were near the edge of the screen. See for example:
Not sure whether that is something you can fix with the settings, I used the GUI version with the same settings as mycl500 mentioned a few posts ago.
Can't say I'm super impressed. It seemed to do ok if the censored bits were near the middle of the screen, but it had quite a few problems when the cencored bits were near the edge of the screen. See for example:
Not sure whether that is something you can fix with the settings, I used the GUI version with the same settings as mycl500 mentioned a few posts ago.
UI, just drag and drop in multiple files. Don't use the pause button it's still buggy on restart.
I was uncertain but decided to give it a try and I must admit that the results are insane. It took around 1h30 with an RTX 4080 exported to an
PCIe 5.0 NVMe SSDs. Old movies are way faster as they are not 1080p or 4k obviously.
Using @mei2 collab to add subtitle an AI to remove the censored and this is a game changer.
some example of the result acheived:
-Besh
PCIe 5.0 NVMe SSDs. Old movies are way faster as they are not 1080p or 4k obviously.
Using @mei2 collab to add subtitle an AI to remove the censored and this is a game changer.
some example of the result acheived:
-Besh
92 movies down... batch is working 24/7, I hope this won't fry my pc lol
bummer is I can't watch my old RM movies anymore, I've been redownloading a lot of my favs actresses censored version starting with Sakura Kirishima. Not sure why but the censored versions are getting harder to find. Lada is still weak on pussies though... I've done AQMB-031 and it's not very impressive...
bummer is I can't watch my old RM movies anymore, I've been redownloading a lot of my favs actresses censored version starting with Sakura Kirishima. Not sure why but the censored versions are getting harder to find. Lada is still weak on pussies though... I've done AQMB-031 and it's not very impressive...
Weak pussies is better than no pussies as the saying goes.
Here is an alternative gui for lada that has a lot of features but is in Japanese (also a chinese translation exists of it). so it needs to be translated with deepseek or any other LLM.
github.com
Mac gui for lada:
codeberg.org
A VR de-mosaic model for lada:
This model can also be used for non VR movies and is very good for single mosaic scenes.
An lada alternative which is very promising (a lot faster and already seems on paar quality wise with lada's mosaic detection but still in early development with a long way to go):
github.com
An English gui for it:
github.com
There is also an English lada gui but costs I think $10 and I don't want to recommand it because everyone works on lada as an community and this one guy wants to make money with it.
It also seems to be possible to convert images to videos, de-mosaic the videos with lada and convert them back to images. There is also another chinese gui for Lada that I heard about with many features but to use it also needs to be tranlated from Chinese to English first..
GitHub - UnderGate-lab/LADA_GUI: simple gui for lada.
simple gui for lada. Contribute to UnderGate-lab/LADA_GUI development by creating an account on GitHub.
github.com
Mac gui for lada:
ladamac
A native macOS GUI application for LADA video restoration with Apple Silicon MPS acceleration.
codeberg.org
A VR de-mosaic model for lada:
An lada alternative which is very promising (a lot faster and already seems on paar quality wise with lada's mosaic detection but still in early development with a long way to go):
GitHub - Kruk2/jasna: JAV video restoration tool
JAV video restoration tool. Contribute to Kruk2/jasna development by creating an account on GitHub.
github.com
An English gui for it:
GitHub - suphamster/jasna_GUI at jasna_v0.2
simple gui for jasna. Contribute to suphamster/jasna_GUI development by creating an account on GitHub.
github.com
There is also an English lada gui but costs I think $10 and I don't want to recommand it because everyone works on lada as an community and this one guy wants to make money with it.
It also seems to be possible to convert images to videos, de-mosaic the videos with lada and convert them back to images. There is also another chinese gui for Lada that I heard about with many features but to use it also needs to be tranlated from Chinese to English first..
Last edited:
wow thx a lot. I'm definitely going to try jasnaHere is an alternative gui for lada that has a lot of features but is in Japanese (also a chinese translation exists of it). so it needs to be translated with deepseek or any other LLM.
GitHub - UnderGate-lab/LADA_GUI: simple gui for lada.
simple gui for lada. Contribute to UnderGate-lab/LADA_GUI development by creating an account on GitHub.
Mac gui for lada:
ladamac
A native macOS GUI application for LADA video restoration with Apple Silicon MPS acceleration.
A VR de-mosaic model for lada:
This model can also be used for non VR movies and is very good for single mosaic scenes.
An lada alternative which is very promising (a lot faster and already seems on paar quality wise with lada):
GitHub - Kruk2/jasna: JAV video restoration tool
JAV video restoration tool. Contribute to Kruk2/jasna development by creating an account on GitHub.
An English gui for it:
GitHub - suphamster/jasna_GUI at jasna_v0.2
simple gui for jasna. Contribute to suphamster/jasna_GUI development by creating an account on GitHub.
There is also an English lada gui but costs I think $10 and I don't want to recommand it because everyone works on lada as an community and this one guy wants to make money with it.
It also seems to be possible to convert images to videos, de-mosaic the videos with lada and convert them back to images. There is also another chinese gui for Lada that I heard about with many features but to use it also needs to be tranlated from Chinese to English first..
edit: man that first compile is nasty... let's hope it pays off...
Last edited:
So for those interested I've tested jasna on my end. I don't have the hardware to even come close to the results advertized by the developper but I can compare to the results I've been getting with lada in the same environment.
First of all, it looks like he's using the same mosaic restoration model as lada. The detection model is different and doesn't support anything other than rfdetr-v2 while lada offers 3 diiferent approaches one of which beeing v4-accurate which I've been using in this test.
Basicaly, the only variables you can play with as of now, are max-clip-size and temporal-overlap. You do need to fine tune the encoder-settings for hevc on your own (no presets like in lada and hevc being the only codec supported) for the output to look closer to the original.
jasna will build a new engine from scratch based on every new set of parameters you throw at it. It's long. But it is that engine that will speed things up once you've found your sweet spot and intend to reuse these params.
I dialed down to max-clip-size 40 and overlap 4 for the model to be able to run at a constant 8.4GB/12 VRAM which was ok.
The dev used 180 with a RTX 5090 and a shitload of vram (>40) to get the best results. I don't think a lot of us have that at home.
so for jasna I used:
and this for lada:
I used UMSO-415 for comparison with lada. The movie is basically all mosaic from start to end. Length is 2h:01. Jasna completed the job in 2h:06. Lada took 40mn more. So safe to say Jasna is significantly faster. As for the results, I'm not really sure what to make of them. I think overall lada takes it home although some cases where handled better with jasna. The project is still a bit young but very promissing. The idea of pre-compiling an engine for a specific set of params you'll always be using is great.
It's worth keeping a close look on how both evolve. On my end, I think I'll be keeping lada for the time being...
First of all, it looks like he's using the same mosaic restoration model as lada. The detection model is different and doesn't support anything other than rfdetr-v2 while lada offers 3 diiferent approaches one of which beeing v4-accurate which I've been using in this test.
Basicaly, the only variables you can play with as of now, are max-clip-size and temporal-overlap. You do need to fine tune the encoder-settings for hevc on your own (no presets like in lada and hevc being the only codec supported) for the output to look closer to the original.
jasna will build a new engine from scratch based on every new set of parameters you throw at it. It's long. But it is that engine that will speed things up once you've found your sweet spot and intend to reuse these params.
I dialed down to max-clip-size 40 and overlap 4 for the model to be able to run at a constant 8.4GB/12 VRAM which was ok.
The dev used 180 with a RTX 5090 and a shitload of vram (>40) to get the best results. I don't think a lot of us have that at home.
so for jasna I used:
Code:
jasna --input "G:\UMSO-415.mp4" --output "G:\UMSO-415 [RL].mp4" --fp16 --max-clip-size 40 --temporal-overlap 4
Code:
lada-cli
--input "G:\UMSO-415.mp4"
--output "G:\UMSO-415 [RL] lada.mp4"
--temporary-directory "G:\tmp"
--device "cuda"
--encoding-preset hevc-nvidia-gpu-hq
--mosaic-restoration-model basicvsrpp-v1.2
--max-clip-length 400
--mosaic-detection-model v4-accurateI used UMSO-415 for comparison with lada. The movie is basically all mosaic from start to end. Length is 2h:01. Jasna completed the job in 2h:06. Lada took 40mn more. So safe to say Jasna is significantly faster. As for the results, I'm not really sure what to make of them. I think overall lada takes it home although some cases where handled better with jasna. The project is still a bit young but very promissing. The idea of pre-compiling an engine for a specific set of params you'll always be using is great.
It's worth keeping a close look on how both evolve. On my end, I think I'll be keeping lada for the time being...
Last edited:
Can someone explain how to install the English GUI?Here is an alternative gui for lada that has a lot of features but is in Japanese (also a chinese translation exists of it). so it needs to be translated with deepseek or any other LLM.
GitHub - UnderGate-lab/LADA_GUI: simple gui for lada.
simple gui for lada. Contribute to UnderGate-lab/LADA_GUI development by creating an account on GitHub.
Mac gui for lada:
ladamac
A native macOS GUI application for LADA video restoration with Apple Silicon MPS acceleration.
A VR de-mosaic model for lada:
This model can also be used for non VR movies and is very good for single mosaic scenes.
An lada alternative which is very promising (a lot faster and already seems on paar quality wise with lada's mosaic detection but still in early development with a long way to go):
GitHub - Kruk2/jasna: JAV video restoration tool
JAV video restoration tool. Contribute to Kruk2/jasna development by creating an account on GitHub.
An English gui for it:
GitHub - suphamster/jasna_GUI at jasna_v0.2
simple gui for jasna. Contribute to suphamster/jasna_GUI development by creating an account on GitHub.
There is also an English lada gui but costs I think $10 and I don't want to recommand it because everyone works on lada as an community and this one guy wants to make money with it.
It also seems to be possible to convert images to videos, de-mosaic the videos with lada and convert them back to images. There is also another chinese gui for Lada that I heard about with many features but to use it also needs to be tranlated from Chinese to English first..
I´m not that good in working with command lines.
The lada gui from Undergate is in Japanese and not English.Can someone explain how to install the English GUI?
I´m not that good in working with command lines.
Installation (point 3): (Release lada_gui.exe) Copy the above to the LADA installation folder (the folder where lada-cli.exe is located).
Translate it to English or use the normal Lada gui.
That's written in the repo:
Download jasna.py from the repo and put it near jasna.exe
Then press Windows Key + R, type cmd and run the command "python jasna.py" or you can just use the !GUI.bat from the repo.
Also make sure to have python installed.
Similar threads
