Home Creators Posts Import Register Favorites Logout

Downloads

Some files exceed the archive size limit and weren't saved: 1 file. Please note these limits are in place to keep this site running long term, without costing a fortune. You can favorite the creator though, certain milestones increase the limit.

Content

The second part of this series is all about advanced AI content creation using comfyUI on Google Colab. The Colab Notebook is set up with everything you need for this workflow. Note that you need ~12GB space on your Google Drive.Attached you find the full comfyUI workflow file, some masks and style references for you to play with and the Google Colab notebook (just upload it on your Google Drive and doubleclick to open it).Also here are lots of interpolated and upsclaed comfyUI renders I did: https://www.dropbox.com/scl/fo/1ou9sl6oos8ov2svdu1w7/AFUe1CiQb-EEqDgVXR55MCM?rlkey=844nkqqhwzefe14nlzdu59zql&st=59ilew5m&dl=0Drive Mirror: https://drive.google.com/drive/folders/1fL7jx4Z5gTjPNtI2Hw1TutAhiwpPQmMa?usp=sharingMore masks: https://drive.google.com/file/d/1Lh2daCqEeDPMxyDPAyqS9u_kR6apx0U8/view?usp=sharing

Files

Content Creation for Immersive Spaces - comfyUI & Colab Tutorial 2/2

The second part of this series is all about advanced AI content creation using comfyUI on Google Colab. The Colab Notebook is set up with everything you need for this workflow. Note that you need ~12GB space on your Google Drive. For the workflow file, extra masks and style references check out my Patreon https://patreon.com/reflekkt and Instagram https://www.instagram.com/reflekkt_/ Colab Notebook link: https://drive.google.com/file/d/1IrixrlWpMvNnfhqtJBYSHyVIMvo-wcix/view?usp=sharing / Mirror: https://www.mediafire.com/file/61hxlj87w8wtht7/comfyUI_colab.ipynb/file AbsoluteReality checkpoint model: https://huggingface.co/digiplay/AbsoluteReality_v1.8.1/resolve/main/absolutereality_v181.safetensors FlowFrames: https://nmkd.itch.io/flowframes 0:00 Intro 1:04: comfyUI general Information 2:12 Google Colab general information 3:22 Workflow & nodes explanation 12:40 Install & start comfyUI on Google Colab 19:08 Build the comfyUI workflow 31:57 Test render in 512x512 34:21 Longer test render 36:06 Make videos loopable 37:08 Change to more powerful L4 GPU for high resolution generation 43:49 Frame Interpolation with FlowFrames 46:03 Topaz for upscaling (if you don't have it, in my Patreon there is an intergrated upscaler included in the workflow) 49:05 Immersive room simulation 52:15 Outro & general information

Comments

Retone

Hi - I am having an issue using the HighRes-Fix Script node. A red box appears around the node when i un-bypass it and the script output in colab says 'output will be ignored' and the job does not get started. I tried changing the latent-upscaler selection to ttl_nn.SDXL to see if a different selection would help but I get the same issue. A few things seem to fail to validate but i think they are in other bypassed nodes ie the pixel upscaler and that doesn't usually stop the output. I can't see exactly what fails to validate on the HighRes-Fix Script

Retone

Failed to validate prompt for output 3: * HighRes-Fix Script 16: - Value not in list: preprocessor: 'none' not in ['_'] - Value not in list: pixel_upscaler: '2x-sudo-UltraCompact.pth' not in ['RealESRGAN_x2plus.pth', 'RealESRGAN_x4plus.pth'] - Value not in list: control_net_name: 'SD15\control_v11e_sd15_ip2p.pth' not in ['control_v11f1p_sd15_depth_fp16.safetensors', 'control_v11p_sd15_lineart_fp16.safetensors', 'control_v1p_sd15_qrcode_monster_v2.safetensors', 'controlnet_checkpoint.ckpt'] Output will be ignored Failed to validate prompt for output 18: Output will be ignored Prompt executed in 0.02 seconds

Retone

Hi thanks for the reply. I tried that with no luck, an re-ran the download of models / nodes with some of the extra models enabled too. It seems that the HighRes script is looking for 'control_net_name: 'SD15\control_v11e_sd15_ip2p.pth' but im not sure how to look inside the script to check what its doing. The error seems to be that 'control_net_name: 'SD15\control_v11e_sd15_ip2p.pth'' is not in the list of controlNet models which it is not so i'm unable to select one that matches what the script is looking for in side controlNet models. The models available in latent_upscaler do not match control_net_name: 'SD15\control_v11e_sd15_ip2p.pth', nor do any of the models in pixel_upscaler if i select pixel as the upscale_type. The models selected in the upscalers do not seem to be an issue, more that the script is looking for control_net_name: 'SD15\control_v11e_sd15_ip2p.pth' somewhere? * HighRes-Fix Script 16: - Value not in list: control_net_name: 'SD15\control_v11e_sd15_ip2p.pth' not in ['control_v11f1p_sd15_depth_fp16.safetensors', 'control_v11p_sd15_canny_fp16.safetensors', 'control_v11p_sd15_lineart_fp16.safetensors', 'control_v11p_sd15_scribble_fp16.safetensors', 'control_v1p_sd15_qrcode_monster_v2.safetensors', 'controlnet_checkpoint.ckpt'] - Value not in list: preprocessor: 'none' not in ['_'] Output will be ignored I was able to use the Pixel Upscaler successfully, but i have this disabled when trying to use the HighRes-Fix Script I just noticed you said to deactivate controlNet toggle - this was always enabled, should i deactivate it when using the HighRes script?

Retone

Hi - Did you ever encounter any issues with Ksampler running out of memory? This is using the Highres FIx and a similar res to the large screens you used torch.cuda.OutOfMemoryError: Allocation on device 0 would exceed allowed memory. (out of memory) Currently allocated : 20.67 GiB Requested : 1.96 GiB Device limit : 22.17 GiB Free (according to CUDA): 26.88 MiB PyTorch limit (set by user-supplied memory fraction) : 17179869184.00 GiB

432 1 Music

Hi, thank you for the content you share. I am incorporating a floor projection in addition to wall projections. I was wondering if it would be possible to arrange a remote session to discuss how I can expand the pipeline to seamlessly integrate both floor and wall projections.