xmyx / ainodes-engine Goto Github PK

From what I've seen I'm super excited about this and eager to try it out. I'm stuck using a Mac, so I was happy to see the mac install.

Following the instructions, I am running into an issue on the step:
pip install -r requirements_mac.txt

There doesn't seem to be any file by that name. When I run from the non "_mac" version it installs fine, but I run into the error:
ModuleNotFoundError: No module named ‘PyQt6.Qsci’

A disclaimer. This is my first time posting anything on github and I'm a designer first, so I'm not the most fluent in the terminal. I'm guess this is most likely user error somewhere along the line.

Great work. Are there going to be any other docs and/or example "networks?"

I hope to draw canny edge directly in node such as canvas.
Also pose.

The one click installer batch runs, shows "AiNodes - Engine installation complete, starting.", shows the splash screen for a split second and that's all.

Double-clicking the desktop icon does the same.

Running from a command line ends with
import psutil
ModuleNotFoundError: No module named 'psutil'

Also, at the end of the install the screen is cleared which stops anyone scrolling up to see any stats and potential causes of the problem, so I do not have any log to share. That screen clear should be removed. Line 144 of setup_ainodes.bat

I can't add an image to the input, the only way is to reload and there it asks me for the dialog box, win10, 3080ti

I followed the instructions for installing in Linux. The first error was:

Package gobject-introspection-1.0 was not found in the pkg-config search path.
Perhaps you should add the directory containing `gobject-introspection-1.0.pc'
to the PKG_CONFIG_PATH environment variable
No package 'gobject-introspection-1.0' found
Command '('pkg-config', '--print-errors', '--exists', 'gobject-introspection-1.0 >= 1.56.0')' returned non-zero exit status 1.

  Try installing it with: 'sudo apt install libgirepository1.0-dev'
  [end of output]

note: This error originates from a subprocess, and is likely not a problem with pip.
ERROR: Failed building wheel for pygobject
Failed to build pygobject
ERROR: Could not build wheels for pygobject, which is required to install pyproject.toml-based projects

So I installed the libgirepository1.0-dev package. Now the program throws this error:

PyGIWarning: Gtk was imported without specifying a version first. Use gi.require_version('Gtk', '3.0') before import to ensure that the right version gets loaded.
from gi.repository import Gtk
Using HF Cache in app dir
qt.qpa.plugin: Could not load the Qt platform plugin "xcb" in "" even though it was found.
This application failed to start because no Qt platform plugin could be initialized. Reinstalling the application may fix this problem.

Available platform plugins are: vkkhrdisplay, wayland, eglfs, minimal, wayland-egl, offscreen, xcb, vnc, linuxfb, minimalegl.

Aborted (core dumped)

I have GTK 3 installed. Is there some other library missing?

I see that this frontend has a lot of movement and the developer seems interested in people trying it. But I see it very complicated because of the usability of the program.

• Nowhere is it explained how it works. There is no wiki, not even an example project at hand, or I have not come across it. What is the order of the nodes, how to set up a simple image generation?
• Even if it is nodal, it is not different from other programs. If it is nodal but you can't even blur the images, you can't do a face restoration and blur it...

I mean, actually the fact that the program is nodal only leads to a messy interface. I'm very attracted to the project but it needs a better UX and a clearer objective than "it does the same as others but in nodes".

I used the Deforum example graph and then removed the connection for Deforum Cadence and connected Deforum anim params to the color params and then clicked run. This produced no output in the console.

What I was hoping would happen, was that for nodes that I leave default/disabled, that it would not require that node to be present. At a minimum it should error out in the console to inform the user something is wrong.

ControlNet Loader model selection box is too short, I organized my model names with slightly longer names, but the box is too short and summarizes them and they all look the same, the node seems to support a bigger box, tanx, win10, 3080 ti, fullhd screen

I followed the Linux instructions from https://github.com/XmYx/ainodes-engine, but the instructions for Linux failed. I switched to conda and tried again, but I had to change...

source nodes_env/bin/activate
pip install -r requirements.txt

to

source nodes_env/bin/activate
pip install -r requirements-linux.txt

I removed the line

cd ai_nodes

as that directory does not exist, and I am already in ainodes-engine

At

cd ainodes_engine_base_nodes
pip install -r requirements.txt

it crashes with:

INFO: pip is looking at multiple versions of realesrgan to determine which version is compatible with other requirements. This could take a while.
ERROR: Cannot install -r requirements.txt (line 38) and basicsr>=1.3.4.9 because these package versions have conflicting dependencies.

The conflict is caused by:
    The user requested basicsr>=1.3.4.9
    realesrgan 0.3.0 depends on basicsr>=1.4.2

I fiddled about with versions in the requirements-linux.txt file but the version conflicts were endless, so I just deleted all version requirements... and everything installed.

But python main.py returned:

Traceback (most recent call last):
  File "/home/jw/store/src/ainodes-engine/main.py", line 211, in <module>
    from deforum.generators.comfy_utils import ensure_comfy
ModuleNotFoundError: No module named 'deforum.generators'

I then tried ./ainodes.sh, which seemed to do a lot of what I just did, but created more subdirs, and apparently created yet another venv inside a venv (conda)! but hey, it works!

Can you provide updated instructions as to how to install on Linux properly?

drag and drop image to input? Would it be possible? Now with the net control models one places many control images in the interface and it becomes very slow to have to go looking for them one by one, and even more so if one tries many, if possible it would be good to drag directly to the node input to change the image and also drag and an input node is created with the image, so one can make different combinations in a fluid way, thanks, win 10 3080ti.

Looking forward to trying this but I'm getting the following error:

Already up to date.
Using HF Cache in app dir
Traceback (most recent call last):
File "D:\ainodes-engine\main.py", line 35, in
from ainodes_frontend.nodes.base.node_window import CalculatorWindow
File "D:\ainodes-engine\ainodes_frontend\nodes\base\node_window.py", line 8, in
from ainodes_frontend.nodes.base.node_sub_window import CalculatorSubWindow
File "D:\ainodes-engine\ainodes_frontend\nodes\base\node_sub_window.py", line 8, in
from ainodes_frontend.nodes.base.node_config import CALC_NODES, get_class_from_opcode, LISTBOX_MIMETYPE, node_categories
File "D:\ainodes-engine\ainodes_frontend\nodes\base\node_config.py", line 77, in
from ainodes_frontend.nodes.image_nodes import image_op_node, input, output, video_input, video_save_node, image_blend_node, matte_node
File "D:\ainodes-engine\ainodes_frontend\nodes\image_nodes\matte_node.py", line 12, in
from ainodes_backend.matte.matte import MatteInference
File "D:\ainodes-engine\ainodes_backend\matte\matte.py", line 13, in
result = subprocess.run(['wget', '--version'], capture_output=True, check=True)
File "C:\Users\PC\AppData\Local\Programs\Python\Python310\lib\subprocess.py", line 501, in run
with Popen(*popenargs, **kwargs) as process:
File "C:\Users\PC\AppData\Local\Programs\Python\Python310\lib\subprocess.py", line 969, in init
self._execute_child(args, executable, preexec_fn, close_fds,
File "C:\Users\PC\AppData\Local\Programs\Python\Python310\lib\subprocess.py", line 1438, in _execute_child
hp, ht, pid, tid = _winapi.CreateProcess(executable, args,
FileNotFoundError: [WinError 2] The system cannot find the file specified

And I don't know if it's planned but it would be great to be able to use this as an API in the future.

in the conditioning node the color of the background is white and that of the text is a very light gray, it can only be read by selecting it, windows 10, 3080ti, fullhd screen

Looks like a very interesting and promising project!

Here's a few things I'd like to see (as a technical artist), so that I can begin using this to make animations:

• A Basic Working Graph, with all the nodes necessary to run an animation, kind of like the basic settings that are automatically loaded with the Auto1111 WebUI version of Deforum. This will give users a place to work from. As things are, with no documentation and no examples, I'm not sure where to begin and indeed have no way of knowing if there is a setup that will function and create an animation at this time. I know it's super early to ask for this, but you have a public download, so I'm assuming there's a way to get output, so including a basic graph that works and loads when you start the program would be great.

• A settings feature, where you can specify locations of the various models/checkpoints. I edited the yaml file in the config directory to try to specify some of the file locations, but not all of them were there and when I tried to use the File>Models Dir, none of my changes took, because it opened the same directories as before, even when I tried several kinds of formatting, such as G://models//checkpoint, etc. I have 100GB of models in my SD directories, so I probably have all the files required, but I'd rather not move every kind of file over or set up Symlinks every time I wish to use the software. Maybe have the program automatically download all the smaller files might be easier than hoping users will know where to find these things and where to put them all.

• Ability to make networks you can go inside, like in TouchDesigner would be very cool. Obviously, basic functionality is more important, but just making an observation, with the amount of settings required to make Deforum work, it would be nice if you could hide the settings that weren't being changed, perhaps in this way. Speaking of TouchDesigner, there are a number of projects that use TD to control SD and you might want to check them out for ideas, because there's no point in re-inventing the wheel.

• Sometimes, the panning function (middle mouse button) stopped working until I zoomed in/out.

I'm okay with experimenting and trial and error, but only if I know there's a way to make it work. If the software works, but needs users to request stuff and find bugs, no problem. But if you want people to put time and effort into the repo, the burden is on you to first provide settings files or instructions we can follow to get some kind of functionality, so we know we're not wasting our time setting up something that's not going to work.

Bear in mind, I'm just technical artist, so my views may not reflect those of your ideal users. I hope something here is useful to you and I wish you luck on your project!

I look forward to seeing this repo become the ultimate node-based animation software! Keep up the good work!

1. Create a new directory
2. Download the latest one click installer from https://github.com/XmYx/ainodes-engine/releases/tag/v0.4.2 and save into the new directory
3. Run the one click installer.
4. Once it finishes double-click run_ainodes.bat

The logo flashes for less than a second and then nothing. GUI no longer starts.

No command line stats.

(nodes_env) D:\ainodes\ainodes-engine>run_ainodes.bat

(nodes_env) D:\ainodes\ainodes-engine>

I have followed the instructions for installing on windows.
I have python 3.10, nvidia drivers with CUDA, and GIT.
I have downloaded the 1 click installer and run it.
I have the shortcut generated on my desktop.
I double click it, console pops up and closes straight away.

I am on windows 11.

and I am now sad

Hi @XmYx, in AI-Nodes Engine, the positive and negative prompts are not visible, although the cursor appears at the point of writing and moves, it is as if writing in white color. They are also not visible if you copy and paste from notepad. This is my first foray into this new application. I have only found one video in Hungarian with no subtitles available, also the AI-Nodes interface of the author of the video is a bit different from mine.
I have configured the settings according to the video, so that it loads the models and extensions from the webui folder of stable diffusion automatic 1111, but it seems that it does not load from there. Should I put the models inside AI_nodes engine?
When running both my test and the example nodes that come gives errors. I don't know programming. I'm trying to understand the application logic but I'm lost.
The url of the video:[(https://www.youtube.com/watch?v=QD7aMacE2f4)]
I attach screenshots.
Please if anyone can help me I appreciate it.

Thanks!!!!

I attach screenshots

For me, zbar failed to build a wheel on Ubuntu. I fixed it by doing pip install zbar-py.

Edited by @XmYx
This was addressed in today's update. Not closing the issue as I'm not yet fully happy with navigation, and I'd like to track progress of that here:

• Keyboard Navigation
• Space to initiate Hand Drag mode
• Better Middle Mouse Drag
• Fix Middle Mouse Drag
• Use MiniMap for navigation
• Add settings panel
• Load User settings at startup
• Python Code Editor / Runner node
• MultiControlnet with start/stop

ORIGINAL ISSUE by @KewkLW BELOW (fixed)

Video of the issue here https://cdn.discordapp.com/attachments/1003154198562623549/1117938162812072068/2023-06-12_15-04-36.mp4

Essentially dragging the canvas sometimes works sometimes doesn't. I think it's something to do being close to a node group.

looks really neat! I like comfyUI too.

Once a workflow is designed, can we run this headless easily?

The front-end I guess is a QT app, but I really just want to run complex workflows with LoRAs directly.
The main.py seems to be to start the client app?

I was looking for something like a server.start() but with QT, is it true there's no line between client and server, as we're used to for web apps? It's all just function calls in the same app?

Is there anyway I could run this on runpod or other cloud service? I'm a mac user so nvidia isn't really an option.

i just try 'basic_sampler', but i got error below

torch 2.0.0+cu118, cuda 11.8, cudnn 8700

  0%|                                                                                           | 0/20 [00:00<?, ?it/s]
  5%|####1                                                                              | 1/20 [00:00<00:05,  3.45it/s]
 15%|############4                                                                      | 3/20 [00:00<00:02,  7.01it/s]
 25%|####################7                                                              | 5/20 [00:00<00:01,  8.97it/s]
 35%|#############################                                                      | 7/20 [00:00<00:01, 10.21it/s]
 45%|#####################################3                                             | 9/20 [00:00<00:01, 10.86it/s]
 55%|#############################################1                                    | 11/20 [00:01<00:00, 11.31it/s]
 65%|#####################################################3                            | 13/20 [00:01<00:00, 11.89it/s]
 75%|#############################################################5                    | 15/20 [00:01<00:00, 12.06it/s]
 85%|#####################################################################7            | 17/20 [00:01<00:00, 12.41it/s]
 95%|#############################################################################8    | 19/20 [00:01<00:00, 12.34it/s]
100%|##################################################################################| 20/20 [00:01<00:00, 10.90it/s]
XSAMPLE: (512, 512, 3)
K SAMPLER: 20 steps, euler  seed:  98336451659587041
Skipping execution: eval_signal(int) needs 1 argument(s), 0 given!

i have no idea what is problem
here is my environment:
windows 10
Ryzen 5800X
RTX3080 12GB
RAM 64GB