StreamDiffusionTD v0.3.1 🚀🖼️

Hi! I'm following your “iPad drawing into StreamDiffusion – TouchDesigner Tutorial”, but I'm having an issue with the image input. In your tutorial, as soon as you connect the Circle TOP → Displace TOP → StreamDiffusionTD, I can see that StreamDiffusion immediately reacts to the circle and continuously updates as the Noise/Displace moves. In my case, StreamDiffusionTD stays stuck on the initial prompt-generated image. The stream itself is running at around 37 FPS, but changes in the input TOP are not reflected in the Stable Diffusion output at all. I'm using StreamDiffusionTD v0.4.0, so I know the version is newer than the one used in the tutorial. I already verified that: The input is connected to StreamDiffusionTD correctly. The input reaches the internal in1 correctly. The input is 512×512. The Displace/Noise animation is working. CUDA-Link Sender is actively sending new frames. python -m sd_installer verify returns 20 passed, 0 failed. StreamDiffusion is running normally at ~37 FPS with SDXL-Turbo in img2img mode. Is there something different in v0.4.0 that I need to enable for StreamDiffusionTD to continuously use the incoming TOP as the img2img input? Right now it seems to keep generating from the initial prompt/image instead of reading the changing input frames.

I had the tox v0.2.6 running in an installation last year all smooth. But I'm running in a lot of issues with this version. 1st hiccup: CUDA SDK 12.8 needs Microsoft Visual Studio 2019 to install all it's features. There's no installer available from Microsoft, and the one thing that is available won't install because of newer versions (Visual C++ Redistributable for Visual Studio 2019). So I continued without all the components installed from the CUDA 12.8 installer. (with the expectation for the tox to fail, because the intstall of CUDA / VS-2019 is already incomplete). 2nd hiccup: From the tox install [venv + req] it stalls and gives errors at step [4/8] ERROR: No matching distribution found for onnxruntime-gpu==1.24.3; extra == "tensorrt" Obviously installing TensorRT also fails. When I do the installation verify I get 2 out of 11 passed: OK: PyTorch with CUDA: 2.7.0+cu128+cu12.8 OK: numpy (<2.0.0 required): 1.26.4 All others are failed. Where do I go from here? Since the documentation doesn't mention any of my problems during install. I need the installation up and running again since it's booked by a client in October. Hope someone can point me in the right direction getting this tox working.

It doesnt work on my 5060

hi it will work on 3060 ? :-(

(venv) E:\geggg\StreamDiffusion\StreamDiffusion-installer>python -m sd_installer repair StreamDiffusionTD Auto-Repair ======================================== Found 3 issue(s) to fix: StreamDiffusion: Cause: huggingface_hub version incompatible Fix: pip install huggingface_hub==0.35.0 diffusers fork: Cause: huggingface_hub version incompatible Fix: pip install huggingface_hub==0.35.0 peft (USE_PEFT_BACKEND): Cause: huggingface_hub version incompatible Fix: pip install huggingface_hub==0.35.0 Apply fixes? [y/N]: y Applying fix for StreamDiffusion... OK Applying fix for diffusers fork... OK Applying fix for peft (USE_PEFT_BACKEND)... OK Re-verifying... OK: PyTorch with CUDA: 2.7.0+cu128+cu12.8 | NVIDIA GeForce RTX 5090 FAIL: StreamDiffusion core: FAILED ImportError: cannot import name 'cached_download' from 'huggingface_hub' (E:\geggg\StreamDiffusion\venv\Lib\site-packages\huggingface_hub\__init__.py) OK: timm (>=1.0.24 required): OK OK: mediapipe solutions: OK OK: transformers (MT5Tokenizer): OK OK: huggingface_hub: OK OK: numpy (<2.0.0 required): 1.26.4 FAIL: diffusers (varshith15 fork with kvo_cache): FAILED ImportError: cannot import name 'cached_download' from 'huggingface_hub' (E:\geggg\StreamDiffusion\venv\Lib\site-packages\huggingface_hub\__init__.py) OK: accelerate: OK OK: controlnet_aux: OK FAIL: peft (required for Cached Attention/StreamV2V): FAILED ImportError: cannot import name 'cached_download' from 'huggingface_hub' (E:\geggg\StreamDiffusion\venv\Lib\site-packages\huggingface_hub\__init__.py) OK: protobuf (<5.0 required for TRT): 3.20.3 OK: onnx (<1.20 required for TRT): 1.19.0 Results: 10 passed, 3 failed「HELP ME! T_T It can‘t work」Thank you~

The environment has been installed, but after clicking to start, the plug-in will always be stuck in loading the StreamDifftionTD model. Is this normal?

I keep getting the error : Traceback (most recent call last): File "C:\Users\jainy\Desktop\ZKM\TD\stream_diffusion_Latest\StreamDiffusion\StreamDiffusion\src\streamdiffusion\wrapper.py", line 1334, in _load_model from streamdiffusion.acceleration.tensorrt import TorchVAEEncoder ModuleNotFoundError: No module named 'streamdiffusion.acceleration.tensorrt' 12:01:53 - root - ERROR - Model loading failed: Acceleration has failed. Traceback (most recent call last): File "C:\Users\jainy\Desktop\ZKM\TD\stream_diffusion_Latest\StreamDiffusion\StreamDiffusion\src\streamdiffusion\wrapper.py", line 1334, in _load_model from streamdiffusion.acceleration.tensorrt import TorchVAEEncoder ModuleNotFoundError: No module named 'streamdiffusion.acceleration.tensorrt' During handling of the above exception, another exception occurred: Traceback (most recent call last): File "C:\Users\jainy\Desktop\ZKM\TD\stream_diffusion_Latest\StreamDiffusion\StreamDiffusion\streamdiffusionTD\td_main.py", line 408, in __init__ self.manager = TouchDesignerManager( ^^^^^^^^^^^^^^^^^^^^^ File "C:\Users\jainy\Desktop\ZKM\TD\stream_diffusion_Latest\StreamDiffusion\StreamDiffusion\streamdiffusionTD\td_manager.py", line 72, in __init__ self.wrapper: StreamDiffusionWrapper = create_wrapper_from_config(self.config) ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ File "C:\Users\jainy\Desktop\ZKM\TD\stream_diffusion_Latest\StreamDiffusion\StreamDiffusion\src\streamdiffusion\config.py", line 55, in create_wrapper_from_config wrapper = StreamDiffusionWrapper(**wrapper_params) ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ File "C:\Users\jainy\Desktop\ZKM\TD\stream_diffusion_Latest\StreamDiffusion\StreamDiffusion\src\streamdiffusion\wrapper.py", line 300, in __init__ self.stream: StreamDiffusion = self._load_model( ^^^^^^^^^^^^^^^^^ File "C:\Users\jainy\Desktop\ZKM\TD\stream_diffusion_Latest\StreamDiffusion\StreamDiffusion\src\streamdiffusion\wrapper.py", line 1898, in _load_model raise Exception("Acceleration has failed.") Exception: Acceleration has failed. Traceback (most recent call last): File "C:\Users\jainy\Desktop\ZKM\TD\stream_diffusion_Latest\StreamDiffusion\StreamDiffusion\src\streamdiffusion\wrapper.py", line 1334, in _load_model from streamdiffusion.acceleration.tensorrt import TorchVAEEncoder ModuleNotFoundError: No module named 'streamdiffusion.acceleration.tensorrt' During handling of the above exception, another exception occurred: Traceback (most recent call last): File "C:\Users\jainy\Desktop\ZKM\TD\stream_diffusion_Latest\StreamDiffusion\StreamDiffusion\streamdiffusionTD\td_main.py", line 526, in main() File "C:\Users\jainy\Desktop\ZKM\TD\stream_diffusion_Latest\StreamDiffusion\StreamDiffusion\streamdiffusionTD\td_main.py", line 521, in main app = StreamDiffusionTD(osc_reporter) ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ File "C:\Users\jainy\Desktop\ZKM\TD\stream_diffusion_Latest\StreamDiffusion\StreamDiffusion\streamdiffusionTD\td_main.py", line 408, in __init__ self.manager = TouchDesignerManager( ^^^^^^^^^^^^^^^^^^^^^ File "C:\Users\jainy\Desktop\ZKM\TD\stream_diffusion_Latest\StreamDiffusion\StreamDiffusion\streamdiffusionTD\td_manager.py", line 72, in __init__ self.wrapper: StreamDiffusionWrapper = create_wrapper_from_config(self.config) ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ File "C:\Users\jainy\Desktop\ZKM\TD\stream_diffusion_Latest\StreamDiffusion\StreamDiffusion\src\streamdiffusion\config.py", line 55, in create_wrapper_from_config wrapper = StreamDiffusionWrapper(**wrapper_params) ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ File "C:\Users\jainy\Desktop\ZKM\TD\stream_diffusion_Latest\StreamDiffusion\StreamDiffusion\src\streamdiffusion\wrapper.py", line 300, in __init__ self.stream: StreamDiffusion = self._load_model( ^^^^^^^^^^^^^^^^^ File "C:\Users\jainy\Desktop\ZKM\TD\stream_diffusion_Latest\StreamDiffusion\StreamDiffusion\src\streamdiffusion\wrapper.py", line 1898, in _load_model raise Exception("Acceleration has failed.") Exception: Acceleration has failed. Press any key to continue . . .

@dotsimulate Hi! I’m using StreamDiffusionTD v0.3.1 with an RTX 4060 Laptop GPU (8GB VRAM). When I press Start Stream, the command window starts loading/building TensorRT, but then it suddenly closes and the stream stops with no clear error. Do you know what might be causing this, and what settings I should try?

Can't run it with acceleration: engines\td\stabilityai\sd-turbo--tiny_vae-True--min_batch-1--max_batch-1--use_cached_attn-False--mode-img2im Stuck on building engine... 3 hours already GPU: NVIDIA GeForce RTX 5070

can i use this with mac os

Hey, since the 2th of May the daydream stream is not starting. With the version 0.3.0 everything is working fine! Do you have any issues?

Hi, I have the old version of TouchDesigner, is it compatible? I also use a Mac.

after seeing LOPs 0.2.0 in the changelog, I've been searching for a download link. finaly found this post, but can't find "dev-zone channel". I'm kind of chomping at the bit to get my hands on this update. Please help!!!

Great ! The pervious version did not work with me on my 5070 Ti , now this version works PERFECT and smooth installation

Hi there, is there plan to add Local Mac support? (Sorry if this is a dupe question)

You rock! Always better

amazing, it runs so smoothly and fast on my 5090 !! thanks <3

Smooth install and running well with a 5080. you rock! thanks!!

❤️❤️❤️