Uploaded February 2026 | Updated November 2026, 2 days ago
ComfyUI Workflow: Stop the flicker. Seriously. If you’ve tried video-to-video style transfer, you know the pain of "zombie faces"—those glitches where the AI just loses the plot.
I built the TeleStyle node to fix it. It runs Wan 2.1 on just 6GB VRAM, locks the structure, and cuts image render times by 40%.
DFans – AI Influencer Platform:
Join here: dfans.co/?ref=LA63HR
Invitation code: LA63HR
The "Frame Zero" Fix (Video) Standard video-to-video creates a mess because the AI forgets what the subject looked like three frames ago. I found a workaround. It’s called the "Frame Zero" secret. You grab the very first frame. Style that single image. Then, you force the TeleStyle node to use it.
It makes the AI "use" that one style across the whole clip. No morphing. No zombies.
The Speed Hack (Images) The official image node is heavy and it tries to load 60GB of files and takes 4 minutes per render. I hated that. So I added a Turbo Mode. By enable on enable_tf32 and dropping min_edge to 512, we get the generation down to seconds. It saves your GPU from crashing and keeps the quality tight.
Download The Workflow: aistudynow.com/how-to-fix-slow-style-transfer-in-comfyui-run-telestyle-on-6gb-vram
Github: github.com/aistudynow/Comfyui-tetestyle-image-video
Safety Check: I scanned these files myself. Clean. No malware.
1. TeleStyle Custom Node: github.com/aistudynow/Comfyui-tetestyle-image-video
2. TeleStyle LoRA Hack: huggingface.co/Tele-AI/TeleStyle/blob/main/weights/diffsynth_Qwen-Image-Edit-2509-telestyle.safetensors
ComfyUI Workflow: Stop the flicker. Seriously. If you’ve tried video-to-video style transfer, you know the pain of "zombie faces"—those glitches where the AI just loses the plot.
I built the TeleStyle node to fix it. It runs Wan 2.1 on just 6GB VRAM, locks the structure, and cuts image render times by 40%.
DFans – AI Influencer Platform:
Join here: dfans.co/?ref=LA63HR
Invitation code: LA63HR
The "Frame Zero" Fix (Video) Standard video-to-video creates a mess because the AI forgets what the subject looked like three frames ago. I found a workaround. It’s called the "Frame Zero" secret. You grab the very first frame. Style that single image. Then, you force the TeleStyle node to use it.
It makes the AI "use" that one style across the whole clip. No morphing. No zombies.
The Speed Hack (Images) The official image node is heavy and it tries to load 60GB of files and takes 4 minutes per render. I hated that. So I added a Turbo Mode. By enable on enable_tf32 and dropping min_edge to 512, we get the generation down to seconds. It saves your GPU from crashing and keeps the quality tight.
Download The Workflow: aistudynow.com/how-to-fix-slow-style-transfer-in-comfyui-run-telestyle-on-6gb-vram
Github: github.com/aistudynow/Comfyui-tetestyle-image-video
Safety Check: I scanned these files myself. Clean. No malware.
1. TeleStyle Custom Node: github.com/aistudynow/Comfyui-tetestyle-image-video
2. TeleStyle LoRA Hack: huggingface.co/Tele-AI/TeleStyle/blob/main/weights/diffsynth_Qwen-Image-Edit-2509-telestyle.safetensors





