DAINAPP

Rife-App 3.36

Here we go, the new version of the Rife-App.

Pretty much the change is the new models, but keep in mind that is not simple "more training", each version I edit the architecture of the model, testing a lot of stuff to see what is sticks.

Right now the architeture have some interesting stuff that is unusual for a Unet Model. I'm gonna to do some extra tests on it (to see if it improve a little)  before releasing it as a open source.
Link:
2023-05-30 01:26:29 +0000 UTC View Post

Rife-App and Bark-Gui

First of all, I apologize for the lack of recent posts, but I have to say that I've been tinkering with the Rife-App model all this time. In the past few weeks, I've tested, tested, and tested some more various configurations and logics to improve the model, and I must say that all this work has paid off. I am currently finishing training the new models for it, and after releasing this version, I intend to work a bit more on them and then release them as open-source on GitHub. I can say witho...

Idea for a small project.

I have this idea for a small project for some time, but not sure if it's worth to pursue.
So I want to know what you all think.

I don't plan to profit from it, just think that is something that some people might enjoy, that why I don't mind if someone want to try to create it themselves.

The main focus of the project is to find buddies to talk to, there is quite a few applications for romance, but not so much to find friends. I do regularly see people sad that it's hard to fi...

New RIFE APP models in training.

Right now they are already working better than the last models, but still gonna take some training, maybe a week? In the mean time I'm working on the GUI and a new release.

Vram is starting to be a problem, been using a 16vram card, and is starting to run out of memory 😬

After this models is done, maybe I can think about renting a better card, we will see.

New Github project & more.

Hi everyone, currently I've just been keeping my head down and focusing on work, and some projects are almost ready. First of all, I've just released a project on GitHub: https://github.com/BurguerJohn/torch-felix

I hope it has some usefulness for developers who train their own AI models. It is based on functions used to train upscalers and (in my case) the Rife-App. It contains various improvements that ...

Rife-App 3.35

Here we go, the new models full trained:

There little modifications on the app itself, right now I'm working on improving some options on it for the next release.

Link:
https://drive.google.com/file/d/1QOoxKO8wNh39ckJxmrwSo_mtJJEnFSxC/view?usp=share_link

The main difference is the models itself.
I fixed the code to run it...

Close to complete the training for the new Rife-App models.

I can easily say that this will be the best models so far. Usually I had to choose for a better flow or less glitches, but this one seen to do both and better than the older ones.

Anime is almost complete, Life Action may take a few more days to be complete.

Here is a interpolation for the pokemon clip, at the start you can see some artifacts, but it probably will be fixed once the model finish training.

Both models should be ready by monday.

[Original]
2023-04-07 00:39:55 +0000 UTC View Post

Big changes for the Stable Diffusion GUI.

• Summary:
  At present, I am in the process of tidying up the code for the purpose of sharing it on Github as an open-source project. This means that anyone will be able to freely download it without any charge. Therefore, if you are only interested in accessing the latest versions of the project, there will no longer be a necessity to continue donating to Patreon, unless of course you wish to continue supporting me. I am releasing this information before the month ends...
  2023-03-22 18:43:04 +0000 UTC View Post

Quick App. Text2Video [Require 12Vram]

It was release yesterday a new project to make a prompt into a video.
So I made a quick GUI for it.
Here the project:
https://modelscope.cn/models/damo/text-to-video-synthesis/summary

Just like Stable Diffusion, will make this first build public (until is working if this one don't) and future updates here on Patreon.

I hope this first version work, I need to go to s...

Working On:

Hi all, just a post on what is being worked on:

For stable diffusion:

I'm working on improving the GUI overall, add some more options and fixing bug, like multiple loras, also making Control Net work with the GUI.

Whisper:
Implementing WhisperX, that is some cases can improve the timestamp of the output. Also trying to add an option to take the audio from the computer input for more of a Real Time Experience. Perhaps in the future I can...

SD 0.68

This is not a big update.
Mostly I was trying to redo the code that handle the Vram Memory, to see if I could fix it to work on 8Vram again.

If you had problem with the latest problem, please let me know if this fix it.

As a bonus, there is a new option. 

Prompt2Img+(Upscale)Img2Img

This will do a regular prompt2image, upscale the image and run img2img on it.

You can select the upscale % and img2img strength in this mode. 50% seen to w...

CUDA Memory Error on 0.67

Since there is a lot of reports on this one, decided to make a new post about it.

I'm doing some heavy debugging to find the source of the problem. There was some heavy changes to the .ckpt loader and Lora also do some changes to the weights, even if no Lora file is selected, so those are my best bets, but i'm still in the process of finding the problem. Thanks for all the feedback, it help me find the problem.

The main pipeline for image rendering is the same, so it should be a ...

SD 0.67

(Hope I did not mess anything in this build)

• Instruct Pix2Pix -> Is a new type of model, you can download it via the GUI

Here some examples:
https://www.timothybrooks.com/instruct-pix2pix

• Dreambooth:

There is two optimizers. AdamW is the defaul on older version. Lion is a optimizer released 10 hours ago. I liked it, so is on the GUI. Did a quick test and...

Information on the next SD Build

Hi all!
Spent most of yesterday testing all the changes to the GUI, will try to finish the build later today, as long everything work, it should be ready today.

Will make the changelog on the actual update post.

I want to thank the ones that used the Form to submit requests and bugs, currently there is 3 submissions, with good changes for the GUI. There is a lot of good suggestions for "Advanced prompt" and  automatic output-as-input. I want to add most of those on the u...

"Turn him into a cyborg"

Instruct Pix2Pix is working, is simiar to img2img, but it will take the original image a lot more in consideration.

Gonna try to finish the build tomorrow, the complex stuff is working, just need to fix now a lot of little stuff.

Update on the Lora code.

Its working horray. It should be available on the next build once the code is complete.

Info about SD development.

Hi all, following the results of the pool, I did not expend time (for now) on improving the seamless code.

I'm also giving a shot on using Google Forms to collect suggestions and bug reports:
https://forms.gle/Hq8e8jJNMo2CD3mZ6
This may help me to organize all suggestions and bugs that need fixing.

I'm currently very focused on making Lora work. Lora is more complex to implement th...

SD GUI Pool

Hi all, currently I'm working on fixing hypernetwork for SD, add Lora Support and Textural Inversion. Plus fixing some bugs.  Right now I'm doing some tests with advanced seamless textures. It would still take quite some work to implement this to the GUI, so I would like to know if you all think I should keep spending time on this code. Here are some examples (they all can be repeated seamless):

https:/...

SD GUI 0.66

Not a big update, just a few things.

• You can download inpainting model from the GUI, just like 1.5 on the last build.
• A fix on .ckpt model loader, it should work with some more models.
• Option to make the texture seamless (Tileable)

There been some requests to add the option to make the images tileable.

I'm not sure how other users handle this, but I gave it a shot on making the code from scratch, me thinks the result are pretty good.
There is...

Rife-App 3.34 [WIP]

Hi all, someone requested a update from Rife. I'm still figuring out the models, this is a WIP version with the current training from the models. They are still improving, so it might contain some interpolation bugs that will be fixed with a little more work.

AMD is not working on this version and the only change are the models, so if you are happy with the last update, no need to download this one.

Link:
2023-01-30 21:32:58 +0000 UTC View Post

Stable Diffusion GUI 0.65 + Real-ESRGAN App Update.

Hi all, been dealing with a few personal problems this last few days, so didn't managed to work as much I would like, but still got some new updates ready.

Real-ESRGAN:
There been some comments that the link was down. I accidentally had deleted it. Took this opportunity to clean the code a little and add an extra model to upscale anime videos. There still more updates I want to make, but with the code clean, it will be easier to update it soon:

2023-01-28 02:24:52 +0000 UTC View Post

Whisper GUI 0.1

Edit: Fixed the link with the missing license.txt, thanks Anthony for figuring it out.

Hi all, took a few slower days after the news eve, to try and relax my mind a little, still been working, but more limited and with more limited access to the internet.

Now I'm start going full steam again, to start I'm releasing a new GUI, this one is a lot simpler than Stable, it will not consume much of my time, in fact this first version have most of the stuff it will have ...

Happy holidays to all. 🍻

A little late, but I hope everyone is having a great end of the year.
Been a little busy with the festivities and family, but the next SD build is almost complete. It will add prior preservation for Dreambooth training, plus some bug fixes.

Prior preservation use renders from the SD model so it's don't overfit with the images you are training.
So let's say, right now if you train pictures of your dog, probably the model will override all dogs with your dog. Prior preservation le...

SD 0.64 [SD 0.63 have a old bug back]

There is a bug on one of the libraries from the project. I fixed this bug a few versions back, but now I had to update the library to fix the problem on Dreambooth training and the bug was back.

This time I did open an issue on the library source code, so the bug don't appear again in the future. 

The library does a test on xformer that can crash everything, even if you ar...

Sd GUI 0.63 (Dreambooth training fix)

This took some time in debug, because I had to train models multiple times to find and fix the code, in the end I had to rewrite quite a lot of the training code.

It probably still have some problems with float32 training, but float16 should work a lot better.

Link [With and without torch folder]:

https://drive.google.com/drive/fold...

A couple of notes on models on the last SD build.

• First, there been some users that put the 2_1 model inside the models/models_sd folder. That is wrong. V2 models are handled different, so they need to go inside models/models_V2 folder to render properly.
• A few users been trying to import old .PT files, older .PT files will not work on this new build, but from this point on, all .PT files should work in the future, unless something drastic change in the code.<...
  2022-12-15 22:51:33 +0000 UTC View Post

Riffusion?

Oh man, Stable Diffusion can generate spectogram with it and generate decent music?
https://github.com/hmartiro/riffusion-inference
https://www.riffusion.com/


Will take a look on the code and try to implement it.

Something funky with Dreambooth on the last build

It seen that there is a bug with some hardware + Accelerator Library + Float32 with Float16 calculations. It only happen once some numbers get to really low values.

I'm almost sure it something to do with the way Accelerator handle some variables.
I did some tests and it seen fixed, but will test a little more tomorrow to make sure, since it happen at random at some point in training, it can take some time between each test.

Thanks for all the reports on this and feedback, i...

SD GUI 0.62

While my search continue for a better host, here is a link for drive and a mirror for it.

Also made an extra build that is exactly like the regular build, but without the "torch" folder, so you just need to copy the torch folder from another build to this ne. It made the .rar a lot smaller.

SD GUI 0.62:
https://drive.google.com/file/d/1Nwl...

A little post about Stable Diffusion.

Developing a project on top of Stable Diffusion is a lot more chaotic than developing a regular project.
I don't have a stable base code to work on, instead it's a cascade of multiple unstable projects, one on top of another.

On top is Emad and the main source of SD:
They work hard on improving SD, releasing new versions of SD.
Each version have their own architecture, improvements and broken stuff.

Next week will be released a new SD method, a lot faste...