š¬ Cinematic Director Workflow (Qwen3VL + Flux)
This workflow is designed to transform simple images or illustrations into high-end cinematic key visuals (Film poster). It bridges the gap between AI vision and creative control, allowing you to not just "generate," but "direct" the final result.
š§ Core Components:
Custom Cinematic Node: A specialized logic controller that uses Qwen-VL to analyze your source image and generate complex, genre-specific transformation instructions.
Model Stack: Optimized for Flux.2 Klein 9B Distilled (Base) and Qwen3-VL-4B-Instruct (Vision Analysis).
Enrico's Nodes: This workflow requires the ComfyUI-enricos-nodes suite for advanced processing.
Precision Control: Powered by CompositorV3 for perfect subject placement.
š How to Use (Interactive Step):
This workflow features an interactive pause to give you full artistic control over the composition:
Initial Run: Start the queue. The process will automatically analyze your image and prepare the cinematic prompt.
The Pause: The workflow will stop at the CompositorV3 node.
Manual Composition: * Adjust the position, scale, or rotation of your subject on the canvas.
This is crucial for Outpainting ā you decide exactly where the character stands in the new cinematic environment.
Resume: Click the "Continue" button at the bottom of the CompositorV3 node to finish the generation.
š Key Features of the Custom Node:
Artistic Preservation: Toggle "Preserve Style" to keep the soul of your illustrations, paintings, or sketches without forcing them into realism.
Emotion Engine: Choose from 30+ detailed emotional states. The node translates simple emojis into professional micro-expression instructions for the LLM.
Canvas-Aware Outpainting: Specialized instructions to fill the empty areas of your composition with genre-consistent background details.
Identity & Pose Lock: Strictly maintains the character's face and original stance while upgrading everything else.
Ā
Ā
ā ļø IMPORTANT: First Run Note
When you start the workflow for the first time, it will appear as if nothing is happening.
Ā
Do not panic! The DownloadAndLoadQwen3_VLModel node is downloading the Qwen3-VL-4B-Instruct model (approx. 9GB).
Ā
Depending on your internet speed, this can take 15-20 minutes.
Ā
Check your terminal/console to see the download progress. Once finished, it will load into memory and proceed.
Ā
š¦ Installation Cinematic-Key-Visual node
Download the provided archive.
Ā Extract the folder named ComfyUI-Cinematic-Key-Visual into your ComfyUI/custom_nodes/ directory. Restart ComfyUI.
š Troubleshooting & Installation Fixes
If you see Red Nodes , follow these steps in order:
1. Install Missing Nodes (The Standard Way)
If you just downloaded the workflow and see red boxes:
Open ComfyUI Manager.
Click "Install Missing Custom Nodes".
Install all required packages and Restart ComfyUI.
2. If Nodes Stay Red After Installation (Version Mismatch)
Sometimes, even after installation, nodes stay red because the developer updated their internal structure.
Delete the red nodes from the workspace.
Right-click on empty space -> Add Node -> Search for Qwen3-VL Model Loader and Qwen3-VL Run (or whichever ones were red) and add them manually from the menu.
Reconnect: Simply plug the qwen_instruction output from my Cinematic Node into the prompt or text input of your newly added Qwen nodes.
Ā
Description
FAQ
Comments (13)
Qwen3_VL_Run,DownloadAndLoadQwen3_VLModel I tried all the ways, updated, rebooted (they're burning red and I don't know what the reason is, they were installed through the manager, but they're still red, I honestly couldn't find a replacement for them
Same here :-(
"Iām sorry youāre experiencing issues! If the nodes are red, it means there is an import error or missing dependencies. Update Everything: Make sure your ComfyUI and ComfyUI-Manager are fully updated.
Check the Console: Look at the black terminal window (console) that opens with ComfyUI. Look for any text or lines saying [IMPORT ERROR] or Failed to load.
Copy and paste this error message here, maybe I can help you.
my Cinematic Node is universal! It generates a standard text instruction that can work with ANY Qwen-VL node set. If those specific nodes are 'red' for you, you can easily replace them with a different implementation. For example https://github.com/1038lab/ComfyUI-QwenVL
To swap them: Install the alternative Qwen nodes.
Connect the qwen_instruction output from my node to the prompt or text input of the new Qwen node.
Connect your image to the new node as well, and it will work perfectly!"
@Krug_LovĀ Hmm, very strange: if I delete the red nodes from your workflow and load them from the menu they work (the only difference is that "Qwen3-VL Model Loader" is identical, while "Qwen3-VL Run" has many more options).
@faustoserone393Ā Great catch! This confirms it was a version mismatch.
The developer of the Qwen nodes likely updated them and added new features/inputs after I saved the workflow. When ComfyUI sees an 'old' node structure that doesn't match your 'new' installed version, it turns red.
For anyone else having this issue: > 1. Delete the red nodes.
2. Right-click -> Add Node -> and find the Qwen3-VL Model Loader and Qwen3-VL Run manually.
3. Simply reconnect the qwen_instruction output from my Cinematic Node to the prompt input of your new Qwen Run node.
Thank you for sharing this solution, it will help many others!"
@Krug_LovĀ Thank you for the workflow :-)
@faustoserone393Ā "I'm glad you liked it! Please feel free to share your cinematic results in the comments/showcase ā Iād love to see what you create with it!"
@Krug_LovĀ I find it more comfortable to use Qwen VL (advanced). Anyway, I posted a couple of images. :-)
Everything has earned a mega cool thing, a big bow to the Author!!!
Wow, thank you! š
Really appreciate the kind words ā glad youāre enjoying it.
Iāll definitely keep pushing this further!