Wan Vace(Native) V2V With Extend Long Video Workflow - Version 20250616
Added 2025-06-15 21:11:55 +0000 UTCFinally!
from Vace Start End Frames, got inspired with this.
Like some recent Vace workflow, it's native. but we need the WanVideoWrapper node pack, because of the WanVideo VACE Start To End Frame
This node is awesome! Or should say Wan VACE architecture is awesome.
From now we can stitching long video in ONE generation process.
The workflow applied 4x129 frames but you can apply more if you want to.
Each sampling pass few frames for next batch of frames to start on another next 129 frames.
I apply this Last Overlap frames concept , similar like Skyreel V2 DF. So, the next batch of video are going to look smooth after stitching in the final step.
I try few models, Self-Forcing (which will be coming up in another Yotube video) , FusionX merged , and normal Wan Vace + CausVid.
Using FusionX merged , it improve the speed and video quality. So it would be the current best option to go for.
I am not going to post this workflow on Youtube , because it will be too many explain for in between for each math and logic. Too many mediocre out there will be confuse if do so.
So , I will just post on here, and if you do not understand how to operate from the setting.
Feel free to leave a comment here or ask in the Patreon Discord.
Workflow attached below.
p.s: depends on the reference video, if face are far shot, you will need the Face Refiner workflow to process it after this long video.
Thanks.
Comments
ah didn't notice you already did a lora training guide ! thx !
Melting Province
2025-06-21 22:33:18 +0000 UTChi, thx for the wf, great job ! Do you plan making a quickstart guide on creating wan loras ?
Melting Province
2025-06-21 22:19:40 +0000 UTCStart index is out of range.
Benjamin Law
2025-06-16 13:53:38 +0000 UTCHey Benji, love the concept, I tried running it and got this error: Patching comfy attention to use sageattn 100%|███████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████| 10/10 [05:28<00:00, 32.87s/it] Restoring initial comfy attention loaded completely 0.0 242.02829551696777 True !!! Exception during processing !!! Start index is out of range Traceback (most recent call last): File "/workspace/ComfyUI/execution.py", line 349, in execute output_data, output_ui, has_subgraph = get_output_data(obj, input_data_all, execution_block_cb=execution_block_cb, pre_execute_cb=pre_execute_cb) ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ File "/workspace/ComfyUI/execution.py", line 224, in get_output_data return_values = _map_node_over_list(obj, input_data_all, obj.FUNCTION, allow_interrupt=True, execution_block_cb=execution_block_cb, pre_execute_cb=pre_execute_cb) ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ File "/workspace/ComfyUI/execution.py", line 196, in _map_node_over_list process_inputs(input_dict, i) File "/workspace/ComfyUI/execution.py", line 185, in process_inputs results.append(getattr(obj, func)(**inputs)) ^^^^^^^^^^^^^^^^^^^^^^^^^^^^ File "/workspace/ComfyUI/custom_nodes/ComfyUI-KJNodes/nodes/image_nodes.py", line 1695, in imagesfrombatch raise ValueError("Start index is out of range") ValueError: Start index is out of range Prompt executed in 1818.94 seconds did you run into it?
yoni cohen
2025-06-16 13:05:45 +0000 UTCThat's absolutely great ! Next step is ti find way how to use free init latent save node. To save last latent and use it in encode decode like first latnet. init latent block you know )
Vladimir C
2025-06-16 11:58:09 +0000 UTC
