[For Patreon Supporters] [Workflow Update] Wan_Vace-Native-V2V-CN LongVideo (Unlimited Length) 20250813 (Patreon)

Related Posts:

(Update) benji_Wan_Vace-Native-V2V-CN Long Video (Unlimited Length) Ver 20250710 https://www.patreon.com/posts/update-benji-wan-133758242

(For Patreon Supporters) Video Editing Inpaint & Outpaint Combo 2 Workflows! For Long Video Length https://www.patreon.com/posts/for-patreon-2-134342230

Native Wan VACE Outpaint - Long Video Gen - Version 20250719 https://www.patreon.com/posts/update-native-134436685

[Workflow Update ]Wan Vace 2.1 - Image2Video LongLengthGen - Ver.20250807 https://www.patreon.com/posts/for-patreon-wan-136018367Wan 2.2 Fun Control : https://www.patreon.com/posts/136389936

------------------------------------------------------------------------------------------------------------

Updates:

Looping logic bug fixed, it can loop through all frames up to as many as your computer can handle.

Cut off some unnecessary custom nodes to make it lighter-weight

Improve the Control Net motion guidance by using more than one Vace control_video input

Adding the frame interpolations, a little tricks for double up the frame rate

Add Ollama Vision for ImageToText

------------------------------------------------------------------------------------------------------------

How to use it?

Pretty easy! First , config the 3 green zones in the workflow.

Basic Video Setting, width , height, mostly you are going to change, other value in this setting group, suggest don't touch it for beginners.

Load Video Reference, you can now using Load by File Path or By Upload to Comfy.

By just Bypass either one Load Video node to-do this. (No need to touch any connections in here,)

Load Image Reference, you can now using Load by File Path or By Upload to Comfy.

By just Bypass either one Load Image node to-do this. (No need to touch any connections in here,)

Text Prompt , very important part for VACE , or any in-context AI.

You need as detail as you can to describe your reference image to get the most Accurate character consistent style. Like other Wan VACE Long Length Gen workflow, such as Outpaint-Inpaint , Mask Editing, Image2Video. , I added Ollama Vision , to have an effective prompt for Wan VACE. You can disconnect the "img_caption" GET node , and ByPass Auto Prompt Group if you don't need this.

Model Loader Setting.

I am using Wan VACE FP16 as the base model. You don't have to follow my setting if your computer have limited VRam to handle this, FP16 SFT, FP8 SFT, or GGUF. Again , you should measure how much your computer can handle accordingly.

Sage Att, use it if you have. Torch Compile, not really a need for me so by default I turn it off.

LoRA loader, basically, I am using FusionX Lora , DetailEnhancer for most of the time. You can try out other T2V LoRA for other effect as well.

And wait for the looping generation, you will get a smooth animation like this.

Please don't scream, when you are successfully generate this XD

------------------------------------------------------------------------------------------------------------

AI Models, LoRA, Files

there are too much AI Models and Lora Models, as we are keep trying new one.

But most essential AI model file for this workflow listed on here : https://www.patreon.com/posts/ai-video-upgrade-133652570

Feel free to ask questions about this workflow in the commend section below, or in our Patreon Discord.

Have Fun ya'll!!!

Attached the workflow below.