CTI - Camera To Image | Live generation from camera capture | OpenPose, Lineart, Tagger
5.0
0 reviewsDescription
What this workflow does
In this workflow you can generate almost live (based on your graphic card) images from your webcam or if you have another source then u can use it too.
How to use this workflow
First you will need extension named: "ComfyUI_toyxyz_test_nodes" created by toyxyz.
Install it by using ComfyUI Manager.
After installation (with ComfyUI restart) u must go and find folder "./ComfyUi/custom_nodes/ComfyUI_toyxyz_test_nodes/CaptureCam".
Then using CDM/PowerShell/GitBash write commend "pip install -r requirements.txt" ofcourse if you have pip installed.
Now you can double click on "run.bat" and do not close cmd window.
In new window u need to:
- select your webcam from dropdown list in "Webcam" section
- set width of your cam captured image
- set haight on 0 (that will scale it by self)
- set FPS on your preferences
- toogle Webcam switch - to see your webcam
- browse folder "./ComfyUI/custom_nodes/ComfyUI_toyxyz_test_nodes/CaptureCam/captured_frames"
- browse image (to oversave) "./ComfyUI/custom_nodes/ComfyUI_toyxyz_test_nodes/CaptureCam/captured_frames/capture.png" (if u want u can use existing image capture.jpg, but i preffer *.png format)
- choose format of browsed image from dropdown list
- on the right select "Constant" from dropdown list
- now click START
Now in ComfyUI in section/group "BASE SETINGS" find node named "Load Webcam Image" and paste full path to your captured image ex. "G:\AI\ComfyUI\custom_nodes\ComfyUI_toyxyz_test_nodes\CaptureCam\captured_frames\capture.png"
Next only u need to do is to load checkpoint, lora for LCM sampler, custom VAE, and controlnets for openpose and lineart/canny.
Now selest "AutoQueue" in Extra Option.
Important: To not overide your output image you must change one option in last node named "Image Save". Set "overwrite_mode" on false if u want to save every generated image otherwise set this second option.
If u have found nice lora and u have good graphic card ( I have only GTX 1060 6GB mobile ;.; ) then u can use generated image insted of camera.
Find the golden ratio and have fun.
Foxy
VAE SWITCH:
1 - BAKED (FROM CHECKPOINT)
2 - CUSTOM
OPENPOSE SWITCH:
1 - NOT USING OPENPOSE
2 - USING OPENPOSE
LINEART SWITCH:
1 - NOT USING LINEART
2 - USING LINEART
Load Webcam Image - Path on your harddrive to image from webcam (read how to use "additional software")
Set groups to Never option ( "Set Group Nodes to Never" ) if u dont want use it and remember about switches otherwise u will get error. Tagger dont need switch i hack it XD.
Tips about this workflow
!!! Remember every step in KSampler, controlnets and tagger will slow down generation. !!!
!!! Remember before you change some parameters in this program (CameraCapture) u must click STOP first !!!
PS: You can also generate openpose na lineart from your poses in live for your other projects. We are only limited by the extent to which we let our imagination lead us - we have to look out of the box.
Discussion
(No comments yet)
Loading...
Reviews
No reviews yet
Versions (1)
- latest (2 years ago)
Node Details
Primitive Nodes (15)
Integer (1)
Note (1)
PrimitiveNode (13)
Custom Nodes (64)
- CR Apply LoRA Stack (1)
- CR Integer To String (4)
- CR LoRA Stack (2)
ComfyUI
- ModelSamplingDiscrete (1)
- CLIPTextEncode (3)
- PreviewImage (5)
- ControlNetLoader (2)
- ConditioningCombine (1)
- VAEEncode (1)
- ImageScaleBy (1)
- CheckpointLoaderSimple (1)
- VAELoader (1)
- CLIPSetLastLayer (1)
- SaveImage (2)
- ControlNetApplyAdvanced (2)
- ImpactSwitch (5)
- ToBasicPipe (1)
- ImpactImageInfo (3)
- FromBasicPipe (1)
- ImpactInt (2)
- OpenposePreprocessor (1)
- LineArtPreprocessor (1)
- PixelPerfectResolution (1)
- WD14Tagger|pysssss (1)
- LoadWebcamImage (1)
- Float (1)
- KSampler (Efficient) (1)
- StringFunction|pysssss (6)
- ReroutePrimitive|pysssss (7)
- ShowText|pysssss (2)
- Text Concatenate (1)
- Image Save (1)
Model Details
Checkpoints (1)
anyloraCheckpoint_bakedvaeBlessedFp16.safetensors
LoRAs (2)
LCM_LORA_15_FAST.safetensors
kimetsu_no_yaiba_style_offset.safetensors