SamuKata
Lyell Hintz
Lyell Hintz

patreon


StreamDiffusionTD 0.2.6 [ Changelog ]

Get the TOX here >>> https://www.patreon.com/posts/122151912/

Installation Updates:

  • NOTE: If StreamDiffusionTD was previously working, you do NOT need to run any part of the Installer again with the new operator. Just grab the new TOX and set the Basefolder to your previous location.

  • PC: Greatly updated / improved installation process for windows + Fixed installation error - cannot import name 'cached_download' - Note: for anyone installing for the first time, I recommend using 3.11.9 on windows now. This is helpful so that this python version can be used inside TouchDesigner for other things.

  • MAC: (apple silicon) supported and builtin installation. Install python 3.11 using homebrew then run the normal process. Note: inference speeds are not as fast as with dedicated gpu and tensorrt is not supported on mac

New Features / Fixes

  • Easy OP Loading - sets Basefolder parameter based on last install location. Starting in this version, the operator should detect the location of your installation. allowing the operator to automatically be configured when dropped into project. Note: The first time you load this version, you will need to set the Basefolder parameter to your previous install

  • Textual Inversion / Embeddings supported (see Models page)

  • Fixed loading multiple (2+) LoRA models + adjusting multiple LoRA weights + added clear LoRA loading error message

  • Fixed cook loops when using loopback and refined timing using the pause toggle. (Tutorial covering this exciting new approach coming soon)

Other / Menu Updates

  • Added settings 3 page (includes Op configurations, presets, tensorrt settings loader, and huggingface model downloader)

  • Model page cleaned and now only includes model selection parameters.

  • Several minor bug fixes and optimizations throughout.

Upgrade Options:

  • Drag the new version on to active DotUpdater in your working project to keep current settings (save first)

Also - if you are experiencing your TouchDesigner project crashing when the stream launches (due to failure in numpy_share op), then it is likely that TouchDesigner's python environment's numpy version has been upgraded to >=2.0.0 somehow. TD comes with numpy version 1.24.1 so you will need to uninstall and install that version manually.

If you have any issues, please report your issues in my discord server in the #🔧troubleshoot-central and tag me @dotsimulate so I see it ! Link to the server in the About / About Me in my TD operators!

Enjoy and thanks for the support! Ive been working on a bunch and have lots to share.

Tutorial on using loopback in StreamDiffusion coming soon! 

Comments

Hi! I managed to install succesfully Stream Diffusion but after some minutes TD is shutting down. I am running on laptop with 4070 RTX. Ty!

Alexandra Niaka

any chance of a Mac installation guide? Thanks in advance :)

skinny r

+1

Kenneth O'Regan

Stream Diffusion is it compatible with 5080 rtx nvidia video card?

Jacopo Bianchi

I am getting an error that says python not found, although it is already exisiting and included in the path environment

selcuk artut

acivate your env... cmd. exe ..go to your steamdiffuse folder then type venv\Scripts\activate afterwards one by one pip uninstall torch torchvision torchaudio pip install --pre torch torchvision torchaudio --index-url https://download.pytorch.org/whl/nightly/cu128 pip uninstall xformers do not use v2v yet

Koray Birand

does not work with 5090. Need to use phyton 3.1 Cuda 12.8 And use pytorch cu 128 nightly but I could not get the v2v working since the xfromer is not compatible I have to rebuild for sm120 but could not succeed. Any help ?

Koray Birand

I am using the 5090 graphics card, cuda version 128, pytorch version 2.7. I want to know how to install it correctly.

君 千澄

Hi there! i never succeed to run a lora. i tried loras in 2.1 as I think it's what I should be but apparently I'm wrong. any tips please?

Frédéric Bussière

Hey, playing around with the TOX. Is it possible to have 2 instances running? I´m separating foreground from background and want to generate each layer individually.

Nils Gallist

Hi, I just had a major problem. Touchdesigner suddenly closes after I start streamdiffusion and the following messages come out. TouchDesigner has encountered an error and must close. It will now attempt to save your current file to C:\Users\lucam\Desktop\CrashAutoSave.Prova stream.toe. Failed inside operator: /project/StreamDiffusionTD/numpy_share_out_cn Saved diagnostic file to 'C:/Users/lucam/Desktop/TouchDesignerCrash.2023.12230_1.dmp'. Please emall this file to support@derivative.ca so we can fix the problem.

Luca Magenta

Hello DOT! managed to run it ! But i'm having issues loading LORAS, I loaded them into the folder and selected them on the TD drop down menu but I'm not seeing any changes on the visuals, is there something more I need to considerate? when loading Loras?

Adrian Baqueiro

Does not run on my 5090 - is there a solution?

Daniel Gary

Hello, I’m currently experiencing an issue with my desktop Mac running macOS. I can install and launch image generation with Stream Diffusion without any problems. However, on my MacBook, although the installation completes successfully, nothing appears in the Stream Diffusion container. The terminal indicates that the stream has started successfully, but no images are displayed. Configuration: Mac M1 Max (Apple Silicon) Python 3.11.9 https://drive.google.com/drive/folders/156iTS5c99TNNm_c-gwMME3dcYJpVHqhp?usp=sharing

Neto Pompeu

i run a Nvidia 5080 ...

Romain Renoux

probleme with Pytorch ... torchvision ... when the Batch file instal the env, it dont find the good pytorch : Looking in indexes: https://download.pytorch.org/whl/Detected, https://pypi.ngc.nvidia.com ERROR: Could not find a version that satisfies the requirement torch==2.1.0 (from versions: none) ERROR: No matching distribution found for torch==2.1.0 do you have solutions ???

Romain Renoux

I NEED HELP!, I've been forever trying to install it on Mac and it seems to be impossible so far, I managed to get the initial pulse running to get the server connected, but the stream is not working, it says: ~~~~~~~~~~ OverflowError: cannot fit 'int' into an index-sized integer

Angel Hurtado

Dear dotsimulate, I am unfortunately having difficulties to run streamdiffusion with rtx5090. I guess it is due to the pytorch version? At the moment it runs on 7fps with the rtx5090, while it used to run on 35 fps with my rtx4090.

Phil Jungschlaeger

Hi, very grateful if you can give us some work on the Blackwell 50XX GPU's and it's compatibility with StreamDiffusion? I'm stuck in idle. Thanks so much!

Nicholas D

is there a new update in progress / coming soon?

GK Artist

I can't install pip install -r streamdiffusionTD/requirements_mac.txt

Fundamentos Motion

Are there any instructions for reliably installing this on a mac? Keep running into various issues when installing step 2 or starting stream. ` print(f"\Controlnet Stream '{input_mem_name}' not found. Try changing the Stream Out Name (to SD) parameter in Stream Settings 2 of TD operator.")`

Rishabh Aggarwal

Thank-you. Can you say if this latest version will work with the latest 5xxx series Nvidia GPU's? I have an RTX 5070 ti and I have been able to get standard stable diffuson working by upgrading pytorch wheels as per this discussion (Pytorch nightly build) https://github.com/comfyanonymous/ComfyUI/discussions/6643

Nicholas D

Hi, when I try to install venv + all req, it says python not found. I've ticked in the option to add it to path when I installed python 3.11.9. Any idea what should I do?

Marko

Hi, I have a question. When I try to launch TOX (after installation), it crashes and gives the following error: ModuleNotFoundError: No module named ‘torchvision’ However, torchvision is correctly installed inside the Python 3.10 folder. I noticed that TouchDesigner 2023 uses Python 3.11 by default, so I added the Python 3.10 path in TouchDesigner’s preferences and also in the system environment variables on Windows 10. But the error is still there… What else can I try?

Valerio

Maybe I solved it myself? syphon_utils except about commandBuffer. CommandQueue should be initialized once. By moving the line, It did not stop updating. That line is command_queue = self.device.newCommandQueue() To after self.device = Metal.MTLCreateSystemDefaultDevice() in __init__

Narumi Inada

Hello, Nice to comment to you. Appreciating. I'm using macOS M-series. v2.6.0 stops every few minutes. v1.9.0 works well but has a shared memory issue with "1740356461432 - Shared memory 'StreamDiffusion-TD' not found". Would you like to check this live stream about v2.6.0? -> https://www.youtube.com/watch?v=eQ4e_l5OFzo

Narumi Inada

Ok so that means there are no crucial features in 0.2.6 that 0.2.3 doesn't have? good to know, I am very interested in multiple controlnets as well for future updates. Thank You!

Victor E Morales

Not at all. I will add a note on that. There is absolutely NO need to change or update your installation if it is working. The operator contains new code that will be used when you Start Stream pulse. Install page is only a 1 time thing unless I very specifically mention it in the release notes.

dotsimulate

Is updating python strictly necessary? My set up is working nice and I am afraid to break it with update ? also multiple controlnets would be awesome :)

Victor E Morales

SEND!!::)))))))))

KOI BOI

SEND BRO

KOI BOI

??? semd msg instgre

KOI BOI

Thank you! How can we use downloaded SD3 Large model?

Atakan Kızıltan

Getting the same error as @Pascal Kulcsar. Followed the steps above and all looked good until getting that error when starting the stream.

Bry Torres

Noted ! I will look back into this soon.

dotsimulate

there is an invite link in the StreamDIffusionTD operrators About page (hit the about me pulse) - see you there ! Hopefully going to get mac installation process figured out today and have an update posted.

dotsimulate

For anyone hitting this issue, @ me on discord. Eager to figure this out and will provide update once we've got an operator that is confirmed. The issue is the installation process. Once the environment is setup, as far as I know there shouldn't be any other issues.

dotsimulate

For anyone hitting this issue, @ me on discord. Eager to figure this out and will provide update once we've got an operator that is confirmed. The issue is the installation process. Once the environment is setup, as far as I know there shouldn't be any other issues.

dotsimulate

Thanks Dariusz ! I will definitely see if I can improve the mac install process from these steps. It seems like the issue may have been the order which I was installing things for mac.

dotsimulate

Hi, I followed your steps to install [venv+req]. After installing [venv+req], I went back to the main interface and clicked ‘Start Stream’ but nothing happened. How can I fix it?

Denny Teng

One problem though.. in the "Start_StreamDiffusion.sh"-terminal I get the following error: "objc.nosuchclass_error: SyphonServerDirectory" Any ideas about that?

Pascal Kulcsar

Amazing! Thank you!

Pascal Kulcsar

I had the same errors as you, and after a couple of hours with the gpt chat, I managed to get everything running o my mac. Here is the final instruction that helped me. Do it in Terminal # 1. Navigate to your project directory # Replace "/path/to/your/project" with the actual path to your project cd /path/to/your/project # 2. Remove the old virtual environment (if it exists) rm -rf venv # 3. Create a new virtual environment with Python 3.11 python3.11 -m venv venv # 4. Activate the virtual environment source venv/bin/activate # 5. Upgrade pip to the latest version pip install --upgrade pip # 6. Install wheel (for building packages) pip install wheel # 7. Install numpy (compatible version for Python 3.11) pip install numpy==1.23.5 # 8. Install PyTorch with MPS support (for Apple Silicon) pip install torch torchvision --index-url https://download.pytorch.org/whl/cpu # 9. Install huggingface_hub (specific version for compatibility) pip install huggingface_hub==0.24.6 # 10. Install additional dependencies from requirements_mac.txt pip install -r streamdiffusionTD/requirements_mac.txt # 11. Install missing packages you mentioned earlier pip install opencv-python python-osc einops syphon # 12. Deactivate the environment (when you're done) deactivate

Dariusz Jablonski

Could you make it running?

Pascal Kulcsar

Possible to use multiple GPUs ?

Daniel Gary

same here :'(

xandro Vandewalle

hello, i get an error during the installation on mac saying "ModuleNotFoundError: No module named ‘cv2’“ - any ideas? :)

adrionaut

Hey there, could you please send me a discord invite? I am facing issues installing streamdiffusion on an apple silicon Mac.

Pascal Kulcsar

OMG thank you so much for making it Mac friendly, been trying so many things and couldn't figure it out

boni

Hoped for multiple controllnets 😔

Kuba Kub

Ty Lyell 🙏🙏

V P

Amazing!!!

V P


More Creators