Consistent Multishots Made Easy: Qwen Image Edit Multi Angles LoRA (ComfyUI Workflow)
Added 2026-01-12 14:28:20 +0000 UTCLink video: https://youtu.be/4pm1Fp8lt6I
Want to turn a single image into cinematic multi-shots while keeping your character perfectly consistent? In this video, I’ll show you how to use Qwen Image Edit combined with the new Multiple Angles LoRA to achieve 360° character control for FREE in ComfyUI.
This open-source solution allows you to generate 96 different camera angles—from side profiles to back views—without losing character details. It’s a game-changer for AI filmmakers and storyboard artists!
In this tutorial, SOTAI covers:
✅ How to set up Qwen Image Edit & Multiple Angles LoRA.
✅ The secret prompting formula: Azimuth, Elevation & Distance.
✅ Building an optimized ComfyUI workflow with a custom "drag-and-drop" node.
✅ Real-world stress tests: Wide shots, High angles, and the difficult Back view. Stop struggling with inconsistent AI characters. Watch now to master this powerful workflow!
⚙️ Build Workflow
Step 1 - Install the custom node
Install QwenMultiAngle node: https://github.com/jtydhr88/ComfyUI-qwenmultiangle
Install KJnode: https://github.com/kijai/ComfyUI-KJNodes
Search in ComfyUI Manager or instal via Git clone
Step 2 - Download the model and place it in the correct folder
Place the file in: "models/unet"
Place the file in: "models/loras"
Place the file in: "models/loras"
Place the file in:
"models/vae"
Place the file in:
"models/text_encoders"
Full workflow in the attachments.
Comments
For my first test it worked reasonably well, then when I wanted to do other tests with other images the angle no longer moves. If you want I can send you screenshots.
Gattat Nidhal
2026-02-18 11:09:08 +0000 UTCSorry for the oversight in my instructions. Please install the following custom node to get “SetNode” and “GetNode”: https://github.com/kijai/ComfyUI-KJNodes
SOTAI
2026-02-06 13:20:37 +0000 UTCHi. I get an error when running your flow: "Node 'Set_model' has no class_type." I can see in your video that your "Set_xxxx" nodes are of type "Constant", but mine are of type "Unknown". I can't fathom how to change that. Any ideas?
Joe Black
2026-02-04 16:48:43 +0000 UTCHello, may I ask a question? When I click “Run,” ComfyUI suddenly disconnects. What could be causing this issue?
kaohakaoha
2026-01-23 07:25:00 +0000 UTCHiya, this might be a noob question but the text edit wont let me put my own prompt in as well. Do I have to do the changes I want first to produce the edited picture first then send that picture through your workflow. Your workflow works great and thank you for the work your doing. Thank you
Littlejim Wolage
2026-01-17 16:10:12 +0000 UTCHey, is it possible to upload multiple images of the same scene as input so the output has a more accurate background?
raspberry pi
2026-01-14 03:20:47 +0000 UTCHello, is that speed something you’re experiencing with my workflow or with another one? Also, the generation speed gradually slowing down and then stabilizing is normal behavior in ComfyUI, so there’s no need to worry too much about it.
SOTAI
2026-01-14 02:30:55 +0000 UTCHello, Sorry if this question is unrelated, when i tried to use Qwen Image edit, During Generation my Comfy UI Become extremely slluggish, I'm using 5080 GPU with 64GB of rams, i don't know if this model is supposed to be slow or not but i took me 10 minute for first generation which is understandable, 5 mins for second image and 1 minute for the third and then went back to around 4-5min ??. it's quite weird to me. is it due to my comfy ui setting making it laggy sometimes?
Ludy Winterhowl
2026-01-13 11:29:29 +0000 UTCYes, it’s definitely possible. However, it will only fit the specific scene it was trained on and won’t generalize to other environments.
SOTAI
2026-01-13 10:47:42 +0000 UTCDo you think this LoRA could be finetuned with a specific location? The goal would be that it would know what should be in the background if it orbits 90 degrees, for example, rather than it inferring what could potentially be there again each time and coming up with wildly different scenery?
Mike K
2026-01-13 06:20:42 +0000 UTCYes, it was really silly of me to miss the main LoRA file. I’ve updated the post now. Thank you for pointing it out.
SOTAI
2026-01-13 03:32:53 +0000 UTCHello, could you explain it in more detail?
SOTAI
2026-01-13 03:29:32 +0000 UTChello i have a problem the node loraloadermodelonly doesn t work for me
Hib00 Gaming
2026-01-12 22:02:58 +0000 UTCAm I just missing the link to the LoRA itself, I grabe the other models, but I don't see the MA Lora. Asked, too soon, it is halfway down the node page. https://huggingface.co/fal/Qwen-Image-Edit-2511-Multiple-Angles-LoRA/tree/main
TheD1C
2026-01-12 21:33:41 +0000 UTC
