My ComfyUI workflows for using MiniMax H3

Will the Spectrum Node be included in this workflow in the future? Or this workflow has already included some kind of accelerator node?

How much VRAM is it required to run REF2V with the provided models linked in the workflow ?
I have 16gb, but I get "Cuda error: out of memory" for a single reference image generation, using the default WF settings.

这个工作流和之前的wan2.2一样简单易用。我找到一个能够辅助生成提示词的skill（https://github.com/MiniMax-AI/MiniMax-H3/tree/main/skills/h3-prompt-writing），只要简单描述就能生成一个完整的六段式提示词，试了一些提示，都能补充的非常具体，并且按照提示词生成的视频效果基本符合预期。如果要描述一些nsfw记得先叠甲。

这个工作流可以添加Perfect Loop吗？单靠首位帧提示做循环动画并不连贯。

I am full up to date, but just can not get this workflow to run.

2 Invalid Connections

- Settings's guide input expects MINIMAX_H3_DIRECTOR_GUIDE, but the connected output is IMAGE.

- DaSiWa Advanced LoRA Loader's model input expects MODEL, but the connected output is STRING.

2 Validation failed

The workflow couldn't validate a connected node.

If/Else switch Audio

If/Else Switch Images

Would it be possible to extend the output, by introducing the last 2 seconds of the previous output to gain continuity? Not sure how it would be prompted?

suggestion/recomandation: Add the new Model Preview Override node from KJNodes to get a ppreview of the video while it generates, you can quickly see if the output will match what you wanted and re-prompt if needed. need the taeh3 file from here https://huggingface.co/Kijai/MiniMax-H3-TAE/tree/main/vae_approx that goes into models\vae_approx

Overall a fantastic workflow, but no matter what I try, I can't seem to get SageAttention working right (on Linux). Any tips?

This is going to be a dumb question, but do I simply use "custom" for aspect ratio when I'm using a FL2VA first reference image?

Hi! First of all, thank you very much for all the work you’ve put into the DaSiWa MiniMax H3 workflow and models.

I wanted to kindly ask whether you might consider releasing a GGUF quantized version of the DaSiWa MiniMax H3 model in the future, perhaps aroundQ5 or Q4?

I recently tested a MiniMax H3 Curve Form GGUF version on the same system, and the memory behavior was noticeably better, especially with Q5. That made me wonder whether a similar GGUF release for the DaSiWa MiniMax H3 model might be technically possible.

Of course, I completely understand that creating and maintaining another model format takes additional time and effort, so this is only a suggestion/request if it is something you would be interested in supporting.

Thank you again for your work and for sharing it with the community!

so basically i can't run 30 steps with ref2va with this workflow cause it just doesnt work :/ i2v is fine - no speed loras used - can't still figure out why ref2va with 30 step 20s doesnt work on this workflow. it does work in other workflows though, could you test 30 steps with 15s and see why not working here? Thanks

Hi Dasiwa, any insight on what we can expect from v1.3? And maybe an ETA? I am having a blast with MMH3 and I love your workflow. Can't wait to see where it will go in the future! Cheers! ❤️

... did anyone managed to make ref2va work with this workflow as it should? Really like to see a working example, all i get are results that are nothing like expected.

Hi everyone! 😊✨ I hope someone can help me out—if you’d be so kind. I need to use a fixed seed again, but I don't see the option to do so in the workflow. I'm referring to the MiniMax H3 workflow; it's excellent—kudos to the creator—but I need that feature to compare results. Thanks a lot in advance! 🤗💗

How do I prevent the seed from randomizing when I hit run?

ComfyUI/models/diffusion_models/MiniMaxH3/

why do you make directory like this?  this is fucking bothersome!!!!

First of all, I’m really enjoying using Workflow. Thank you.

I have to manually adjust the aspect ratio every time I use an image—is there a way to do this automatically?

It seems like it would be possible if I could load the image from the Director node into the DaSiWa Resolution Scale Calculator, but it doesn’t appear that I can retrieve the image from the Director node.

is there a limit to how long videos can be before they stop working? not breaking, but like stop falling prompts well or renders get unreasonably long? iirc, wan2.2 you wanted to go no more then 7 seconds before it starts to degrade. i could be wrong on that since it's been long time since i used wan2.2 locally on comfyui.

This is the best workflow for H3 that I have found, especially the way you have set up the ref2Vid, it just works, the only thing missing is some of the newer speedup loras.

I’m not sure if this is feasible, but could a dedicated field for the first frame be added directly to Ref2v? When using prompts, H3 sometimes fails to strictly adhere to the instruction that "<Picture 1> is the opening-frame anchor for [Shot 1]."

This was the best workflow but a recent update made it impossible to just copy prompts from LLMs, now requiring each of the 6 fields to be handled separately. Anyone know how to fix this?

Does anyone have a tip for the director node? It only accepts a prompt. I assume I must be overlooking something simple. It also didn't work with the lora loader provided but I just switched it out. I think I read something about "lanes", but I don't seem to have those, just a prompt

Is there a way to download V1.1? The latest one throws a clip error with the Load Clip node.

Is it possible to loop videos?

a new LORA for face consistency: https://huggingface.co/fal/MiniMax-H3-Realism-People-LoRA

This is a nice setup, works fine straight off the bat. I was surprised to see no preview. Only changes I needed to give it a thorough run-through were adding the preview node and inserting my own sizing templates and output nodes. Nice work on the H3 director. The model is incredibly sensitive to promptage, any automation at all saves a lot of time.

Oh, question. How do you extract input file names from your director node? I can't really use it without that. I use S&R notation %% to call the input filenames, otherwise I can't keep track of anything. My comps are gigantic, many, many clips, and obviously serial FLF needs to line up. Where does this data live in your node?

Are you planning to make your own minimax h3 version like you made with wan and ltx? The base model is great, but it misses the dasiwa touch.^^

Is there a way to change the output format to MP4?

The director update is nice, but when I drag a video into comfyui that had the workflow embedded, my prompt doesn't show up in the director node, it just shows the default text. This essentially means I can no longer use my old prompts.

Additionally, it's quite inconvenient to not be able to set "control after generate" on the noise seed via the main settings panel. Seems like a bit of an oversight no?

When I try to use the last frame extracted from the video as a first frame for the next one in I2VA, it gets progressively darker, making it hard to create a seamless video, Anyone has a tip or fix for that?

Dasiwa Minimax model when ? 🤔

thanks for the constant updates to this workflow, for me everything works expect the preview node it keeps giving me this error [WARNING] Initial noise preview failed (sigmas still sent): mat1 and mat2 shapes cannot be multiplied (1060224x1 and 24x3) in yellow text. any advice?

@Darksidewalker Have you seen people using the ref model as like a qwen edit image editor? I've always wanted to do something this with previous video models and grab the starting image before generating the video.

Generating image to video with this workflow is super slow even with the turbo lora I don't know why.

@Darksidewalker I found out that comfy-kitchen Attention seems to have a better speed then SageAttention. the new comfy attention is already built in comfyui you just dubble click on a new emty spot and you search for ModelAttentionBackend you open the node then you select comfy-kitchen Attention.

SageAttention is slower and has a blurry haze on the videos mostly in the eyes you can see that or hands in lots of movement...
comfy-kitchen Attention is faster and keeps the video sharp

i tried adding it myself but there are so many settings in the backend of the workflow... i would rather wait for the new workflow update where comfy-kitchen Attention is added as a toggle under SageAttention or something

First of all, I really like the workflow you created. However, I've encountered some problems recently, and do you know how to solve them? The error messages are as follows:

4 nodes — 4 errors

2 Invalid connection

Connected nodes are using incompatible input and output types.

Settings - guide

Settings's guide input expects MINIMAX_H3_DIRECTOR_GUIDE, but the connected output is IMAGE.

DaSiWa Advanced LoRA Loader - model

DaSiWa Advanced LoRA Loader's model input expects MODEL, but the connected output is STRING.

1 Validation failed

The workflow could not be validated because a node validation check failed unexpectedly.

DaSiWa Enhanced Video Combine

DaSiWa Enhanced Video Combine failed during validation: tuple index out of range

1 Validation failed

The workflow couldn't validate a connected node.

If/Else Switch - images

If/Else Switch couldn't validate images: tuple index out of range

Minimax H3 Director node now always gives me an error: ## Error Details - Node ID: 1512:2591 - Node Type: SamplerCustomAdvanced - Exception Type: RuntimeError - Exception Message: RuntimeError: shape '[1, 128, 1, 1, 20, 2, 12, 2]' is invalid for input of size 125952. I am up to date, and tried all possibilities with aspect, resolution or scaling - always the same dimension error. i set 768x1280 manually - no change.

Is there a good way to view the prompts for generations made with previous versions of the director node?

When I've tried to reload older generations none of the references or prompt are transferred.

I really like this workflow. Works great and I like the RTX upscaler integration. I'm still on v12 as my comfyui is still on the released version. Would love a way other than copying and pasting the Director node to save various reference setups for characters and scenes. maybe a scene and character nodes that could be used as inputs? maybe this is coming?

Hi! This workflow is really great—thanks a lot for it. I have a question. I'm using reference images, and I also try to use a rough sketch just to show the character's pose and camera angle. In my prompt, I say it's a layout. But Minimax still treats the sketch like it's part of the final video and tries to include all the messy lines. How can I stop that from happening?

Also, is there another way to control character pose and camera position? I need control, but making a first frame for every single shot is just too much work.

Thanks for the workflow! Always love the way you organize it.

I had to replace your custom LoRA loader though because it doesn't seem to support keyboard entry of strength values.

You have once again surpassed every expectation. The moment MMH3 was released, there were a handful of workflows floating around, but nothing worked properly for me. Yours worked right away with no issues. Load the models, and done!~

Your layout is always super easy to follow, and I love that. Version 1.2 was great, but 1.3 never worked for me, so I stayed on 1.2 until 1.4. Now, with the auto aspect ratio and everything else, all my stuff finally works out of the box without me having to crop or resize anything.

One thing I noticed in your "Model Storage Location" guide: on my setup, the fl2va and ref2va models had to go into the unet folder instead of the diffusion_models folder. Not sure if that was a small mistake in the guide or just something specific to my setup, but other than that, I had no issues.

It is also really well optimized. 10 seconds, 2K videos upscaled with RTX Upscaler look so much better, and on my dinky ROG STRIX RTX 3090 24GB, it only takes about 13 minutes, which is pretty crazy.

Keep up the good work! ♥

Dasiwa, you always deliver. Thank you.
I have questions. Will you add Spectrum node in the near future?

Is there any reason you put the Shift video at 10 and shift audio at 5, instead of the official 12/3?

And lastly, I'd love that you make a list of what you changed with each update, because it's not very clear.

Thank you for your work.

the current minimax h3 director and the dasiwa lora loader causing severe lags...im on the newest comfiui build and i did the nightly update with the newest dasiwa nodes(with new aspect settings and so on).

if i pull each node somwhere else on the canvas its running smooth.if im looking on one of the nodes its very laggy...

I really like DSW's 'workflow' node mode and neat layout.

It worked in the previous (2026-08-12) 'MINIMAX-H3 workflow', but after updating the 'Comfyui-Dasiwa-nodes' extension and the new version (08-18) of 'workflow', both the previous and current workflows became unavailable, and 'sage attention' became unavailable. Even if I disabled these options, they were always listed as 'GPU' BOOM', and prompted me to reduce the image resolution. In fact, I always generate with one image specification. I realized that the new 'workflow' also added a large node of 'Model Preview Override', which I knew nothing about, so I did not have the 'tiny_VAE' model and disabled it. The current 'workflow' I cannot run successfully at all. I don't know where the problem lies. Can you explain the updated content and changes in the introduction? I don't know how to fix it now.

Hello sir, could you please tell me where to put the Spectrum apply minimax h3 node in this workflow? I'm quite new in this, so please be patient with me. Thanks.

hello sir as a big fan of your older checkpoints i wonder if this workflow does actually help in doing nsfw stuff - since i guess it wont until we dont get propper loras or checkpoints for MH3 (of which i assume arent any published yet)?

Hi DaSiWa,

I've been comparing MythicAlchemy v15 and v12 side by side (same input, same settings otherwise), and v15's video output looks noticeably more compressed/blocky than v12's.

Digging into the workflow JSON, I found that the quality parameter on the DaSiWa_EnhancedVideoCombine node is set to 28 in v15, versus 20 in v12. Since that's the CRF-style compression setting, an 8-point jump is a pretty visible quality drop on its own.

Was that change to 28 intentional (e.g. to cut file size or export time), or did it slip in unintentionally between versions? Just want to understand the reasoning before I revert it to 20 on my end.

Thanks for the great workflow!

I recently updated your nodes, and it moderately broke the workflow. Are you planning a release to fix that? It was just some nodes not connected properly, but that might cause some people grief. Lol.

Can you please add an autoprompter option?

You can combine 2 turbo LORAs: minimax_h3_fl2v_lightx2v_turbo_4step_v0.1_comfy.safetensors @ 0.75 and minimax_h3_fl2v_turbo_8step_v1.0_comfyui_bf16.safetensors @ 1 -> down to 6 steps now (try ER_SDR, euler or res_multistep for best results).

I just wanted to say that you did a fantastic job. It's incredibly well-made and very user-friendly.

A huge thank you!

your lora node is not working any way to fix it

Latest one is good and easy to work with, thank you. Problem for me is that every time I change a setting in the Director node, the text box resizes. If I swap out a photo, it resizes. If I change resolution, it resizes.

Not sure if this is a node issue or if it's your custom node, so plese let me know.

Thank you for your work. Will you be adding support for Motion Context? Also, I hope you can add an option to lock the seed—that would make testing much easier for me.

It's the new version with kitchen attention better with an RTX 5090 , or should i stick with the sage attention one ?

Dear dasiwa, I just put your native workflow in ComfyUI to the test. I started generating a ten-second video, and the timer showed exactly “Prompt executed in 00:42:06.” Honestly, the output quality is breathtaking, but the downside is that the execution time drags on. For comparison, I ran the same video through another workflow, and there, the time dropped to just “executed in 572.23 seconds.” The result: I’m truly sorry that I have to do without your workflow—which is so well-structured—solely because of this slowness. That said, thank you so much for sharing it and for all the hard work you put in. Hats off to you.

Both videos are here, posted side by side, so everyone can see and judge for themselves.

love ur job ! will u add auto prompt, turbo and sol ?

ComfyUI-H3-SLA-Attention https://github.com/PlagueKind/ComfyUI-PlagueKind-Nodes

i put it after Comfikitchen node(before the guider).it drasticly reduces rendering time at higher resolution 1mp and 15 seconds.mine go down from ~60s/it to ~ 44s/it.at lower resolution there is not much improvements .and pls add fixed seed so i can compare better pls :)

Quick question. when i add a turbo lora, i get garbled audio. Do we know why? Do I need to change workflow settings other than reducing steps?

1.6 is not working for me:

# ComfyUI Error Report ## Error Details - Node ID: 1512:2591 - Node Type: SamplerCustomAdvanced - Exception Type: RuntimeError - Exception Message: RuntimeError: shape mismatch: value tensor of shape [180, 512] cannot be broadcast to indexing result of shape [720, 96]

I deleted all custom nodes reinstalled them, made all the latest updates, but no luck. 1.4 still works.

Thank You DaSiWa. One question - how to set noise to fixed ? There is no option - even inside sub graph its grey out.

i didn't find the keep generate seed button where is it?

There is an aspect ratio option in the configuration. What does “Custom” mean? Does it automatically match the aspect ratio of the input image, or do I need to unlock the width and height settings and enter fixed values manually?

I tried selecting “Custom,” but the generated video still doesn’t seem to match the aspect ratio of the original image.

I don't know what's going on, but this workflow eats up about 2–4 GB of space on my SSD per generation; after just 20 generations, it’s practically filled up 200 GB. Is that normal? Am I going to have to say goodbye to my SSD? Hahaha.

Is it even worth trying with GeForce RTX 3080 12GB or Minimax requires more memory/better videocard?

Is there some trick to preventing smearing/distortion on distant faces and fast detailed motion (e.g. fingers and hair)? I've tweaked a whole bunch of variables (MP, steps, LORA weights, etc) but nothing seems to work.

Hi everyone.

Is there a way to use {choice 1|choice 2} to generate several movie where it pick one of the option ?

Anyone can recommend the best turbo lora for this workflow? One i use gives weird results

Hi Dasiwa,
Great works. .
Using on Comfyui kitchen attention on V 0.32
660 secondes for 14 second R2V clip with 48 fps frame interpolation + upscale *2.
Final size : 1216/2176
This is for your int 8 hybrid H3 checkpoint.
On the other side, i wonder how and so wich of your workflow using to run your W4A8 checkpoint ?
Using the actual Workflow leads to a strange waving clip.
Thanks

Could you add an optional "Unload All Models" step before the H3 sampling stage?

I'm running your workflow on an RTX 4060 Ti 16GB + 32GB RAM, and it seems that the large Qwen encoder remains loaded after text encoding, potentially consuming a lot of RAM during H3 sampling.

Some users report significantly better memory usage after manually unloading Qwen before sampling.

An optional "low RAM" version of the workflow would be amazing for 16GB VRAM / 32GB RAM systems. 🙏

Thanks for the excellent H3 workflow!

大佬，有啥办法可以让抽插的动作加快吗？？2个高评论的lora 都试过了，一个太快都不复合物理学了，另一个虽然快，但人物动作幅度极小。。。

有个问题，我生成的是10秒视频，用的FL2VA首尾帧模式。步数下面那个shift_v 我设置的是10.。现在跑出来的视频最后2秒画面是静止的。。。。是我哪里没设置好么

Hey boss, just wanted to toss this into your community suggestion box:

https://github.com/Carasibana/ComfyUI-H3-FaceRefine

Splendid job as usual , the one issue I'm having is when I select 16:9 aspect , the generation always gets stuck , and I have to force close the comfy ui console to fix it , any idea why it could be happening? with other aspects it works fine.

Has anyone tried motion syncing using REF2VA with this workflow?

When I try to have the character in REFImage follow the motion in REFVideo, it either doesn't follow the movements at all or just lip-syncs to the song included in REFVideo.

Even if I set the prompt to have the character in REFImage follow the movements in REFVideo, it does not use the movements in the video as a reference at all.

Anyone know if there's a youtube video explaining how the heck to use the MinMax H3 Director for Ref2VA? I've uploaded Images and Audio, but have no clue how to fill in all the extra stuff. Hitting Prefill just makes generic "opening frame anchor", which I have no idea what that means. :p

Anyone Know where to Find his FL2VA? Can't seem to find it.

In this workflow, is there a way to use REF2VA but without audio involved? As it create output without any audio, skipping any processing audio related entirely.

Is it possible to change the output file type to like mp4 or something else? Or would I have to route the nodes to a different video combiner?

This Workflow might require some extra steps that if you're not a ComfyUI expert you might need an LLM to walk you through problem solving, but... totally worth it.

Currently testing with the new Eros model of H3 and I'm getting 5sec video on a 4090 with no loras in about 2.5min. I had to modify the subgraph for 0 video and audio shift. but it works great. Full 15sec is about 10min.

Need small advice from you guys. I'm enjoying the workflow (Thanks Dasi) But in terms of video length the workflow hits the wall at 30 sec. I can do 27 just fine but can't go past that. What methods do you guys use to extend the length of the generations inside this workflow?

Im loving this workflow, so clean! Hoping for an extended video update. I've seen a few workflows that use 15 second, "chunks" that keep the voices and scene context the same, but most are full of extra junk that break them unless your a master ComfyUI user.

model link?

Hey have you heard about the Latent Upscale Model? It sounds like people are liking it from this workflow, perhaps you can add it to yours too?

很期待你的长视频工作流

Might be a dumb question, but how to randomize the seed after each gen ? 🤔

This is very very good btw, thanks !

Workflow update AND turbo model!? We are eating good today boys!

Having an issue where even if I change seed, the generation is always the same every gen. Any way around this?

Any one could give an insight on what the latent upscale node settings should be? no matter what I try it either get stuck generating or I get an OOM :( I got a 4070 ti 16vram +32ram

And the bf16 model from here https://huggingface.co/LBH-123-AI/Minimax_h3_latent_Upscaler/tree/main

Using minimax h3 cache give this error:

[INFO] got prompt

[INFO] Found quantization metadata version 1

[INFO] Detected mixed precision quantization

[INFO] Using mixed precision operations

[INFO] Native ops: convrot_w4a4, asym_w4a8_int8, float8_e5m2, float8_e4m3fn, int8_tensorwise , emulated ops: nvfp4, mxfp8

[INFO] model weight dtype torch.bfloat16, manual cast: torch.bfloat16

[INFO] model_type FLOW_AV

[DaSiWa MiniMax H3 Director] mode=I2VA; requested_model=fl2va_model; passed_model=comfy.model_patcher.ModelPatcherDynamic; canvas=768x1152; frames=158; fps=24.0; refs=images:0,videos:0,video_audio:0,audio:0; timeline_items=1

[INFO] VAE load device: cuda:0, offload device: cpu, dtype: torch.float32

[INFO] VAE load device: cuda:0, offload device: cpu, dtype: torch.float16

[INFO] Found quantization metadata version 1

[INFO] Using MixedPrecisionOps for text encoder

[INFO] CLIP/text encoder model load device: cuda:0, offload device: cpu, current: cpu, dtype: torch.float16

[DaSiWa MiniMax H3 Director Guide] mode=FL2VA; upstream=MiniMaxH3ImageToVideo; clip=CLIP; vae=VAE; audio_vae=not-used; frames=158; first_frame=True; last_frame=False

[INFO] Requested to load MiniMaxH3TEModel_

[INFO] Model MiniMaxH3TEModel_ prepared for dynamic VRAM loading. 14257MB Staged. 0 patches attached. Force pre-loaded 310 weights: 1272 KB.

[INFO] Requested to load MiniMaxH3VideoVAE

[INFO] Model MiniMaxH3VideoVAE prepared for dynamic VRAM loading. 4965MB Staged. 0 patches attached. Force pre-loaded 128 weights: 348 KB.

[DaSiWa MiniMax H3 Director Guide] passed forward from MiniMaxH3ImageToVideo: conditioning=list[1]; latent={samples:(1, 24, 47, 72, 48)}

[INFO] Requested to load MiniMaxH3

[INFO] 0 models unloaded.

[INFO] Model MiniMaxH3 prepared for dynamic VRAM loading. 19995MB Staged. 0 patches attached. Force pre-loaded 210 weights: 1175 KB.

[INFO] Patching torch settings: torch.backends.cuda.matmul.allow_fp16_accumulation = True

0%| | 0/8 [00:46<?, ?it/s, Model Initializing ... ]

[INFO] Patching torch settings: torch.backends.cuda.matmul.allow_fp16_accumulation = False

[ERROR] !!! Exception during processing !!! FinalLayer.forward() missing 3 required positional arguments: 'sigma', 'sample_sigmas', and 'shifts'

[ERROR] Traceback (most recent call last):

File "C:\AI\ComfyUI\execution.py", line 545, in execute

output_data, output_ui, has_subgraph, has_pending_tasks = await get_output_data(prompt_id, unique_id, obj, input_data_all, execution_block_cb=execution_block_cb, pre_execute_cb=pre_execute_cb, v3_data=v3_data)

^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^

File "C:\AI\ComfyUI\execution.py", line 344, in get_output_data

return_values = await asyncmap_node_over_list(prompt_id, unique_id, obj, input_data_all, obj.FUNCTION, allow_interrupt=True, execution_block_cb=execution_block_cb, pre_execute_cb=pre_execute_cb, v3_data=v3_data)

^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^

File "C:\AI\ComfyUI\execution.py", line 318, in asyncmap_node_over_list

await process_inputs(input_dict, i)

File "C:\AI\ComfyUI\execution.py", line 306, in process_inputs

result = f(**inputs)

^^^^^^^^^^^

File "C:\AI\ComfyUI\comfy_api\internal\__init__.py", line 149, in wrapped_func

return method(locked_class, **inputs)

^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^

File "C:\AI\ComfyUI\comfy_api\latest\_io.py", line 1990, in EXECUTE_NORMALIZED

to_return = cls.execute(*args, **kwargs)

^^^^^^^^^^^^^^^^^^^^^^^^^^^^

File "C:\AI\ComfyUI\comfy_extras\nodes_custom_sampler.py", line 1055, in execute

samples = guider.sample(noise.generate_noise(latent), latent_image, sampler, sigmas, denoise_mask=noise_mask, callback=callback, disable_pbar=disable_pbar, seed=noise.seed)

^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^

File "C:\AI\ComfyUI\comfy\samplers.py", line 1335, in sample

output = executor.execute(noise, latent_image, sampler, sigmas, denoise_mask, callback, disable_pbar, seed, latent_shapes=latent_shapes)

^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^

File "C:\AI\ComfyUI\comfy\patcher_extension.py", line 114, in execute

return self.wrappers[self.idx](self, args, *kwargs)

^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^

File "C:\AI\ComfyUI\custom_nodes\ComfyUI-DaSiWa-Nodes\nodes\nodes_minimax_h3_cache.py", line 165, in call

return sample_fn(*args, **kwargs)

^^^^^^^^^^^^^^^^^^^^^^^^^^

File "C:\AI\ComfyUI\comfy\patcher_extension.py", line 106, in call

return new_executor.execute(*args, **kwargs)

^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^

File "C:\AI\ComfyUI\comfy\patcher_extension.py", line 114, in execute

return self.wrappers[self.idx](self, args, *kwargs)

^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^

File "C:\AI\ComfyUI\custom_nodes\ComfyUI-KJNodes\nodes\preview_override_node.py", line 685, in call

return executor(noise, latent_image, sampler, sigmas, denoise_mask, new_callback, disable_pbar, seed, latent_shapes=latent_shapes)

^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^

File "C:\AI\ComfyUI\comfy\patcher_extension.py", line 106, in call

return new_executor.execute(*args, **kwargs)

^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^

File "C:\AI\ComfyUI\comfy\patcher_extension.py", line 113, in execute

return self.original(*args, **kwargs)

^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^

File "C:\AI\ComfyUI\comfy\samplers.py", line 1262, in outer_sample

output = self.inner_sample(noise, latent_image, device, sampler, sigmas, denoise_mask, callback, disable_pbar, seed, latent_shapes=latent_shapes)

^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^

File "C:\AI\ComfyUI\comfy\samplers.py", line 1237, in inner_sample

samples = executor.execute(self, sigmas, extra_args, callback, noise, latent_image, denoise_mask, disable_pbar)

^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^

File "C:\AI\ComfyUI\comfy\patcher_extension.py", line 113, in execute

return self.original(*args, **kwargs)

^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^

File "C:\AI\ComfyUI\comfy\samplers.py", line 1005, in sample

samples = self.sampler_function(model_k, noise, sigmas, extra_args=extra_args, callback=k_callback, disable=disable_pbar, **self.extra_options)

^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^

File "C:\AI\ComfyUI\venv\Lib\site-packages\torch\utils\_contextlib.py", line 124, in decorate_context

return func(*args, **kwargs)

^^^^^^^^^^^^^^^^^^^^^

File "C:\AI\ComfyUI\comfy\k_diffusion\sampling.py", line 205, in sample_euler

denoised = model(x, sigma_hat s_in, *extra_args)

^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^

File "C:\AI\ComfyUI\comfy\samplers.py", line 640, in call

out = self.inner_model(x, sigma, model_options=model_options, seed=seed)

^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^

File "C:\AI\ComfyUI\comfy\samplers.py", line 1208, in call

return self.outer_predict_noise(*args, **kwargs)

^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^

File "C:\AI\ComfyUI\comfy\samplers.py", line 1215, in outer_predict_noise

).execute(x, timestep, model_options, seed)

^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^

File "C:\AI\ComfyUI\comfy\patcher_extension.py", line 113, in execute

return self.original(*args, **kwargs)

^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^

File "C:\AI\ComfyUI\comfy\samplers.py", line 1218, in predict_noise

return sampling_function(self.inner_model, x, timestep, self.conds.get("negative", None), self.conds.get("positive", None), self.cfg, model_options=model_options, seed=seed)

^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^

File "C:\AI\ComfyUI\comfy\samplers.py", line 620, in sampling_function

out = calc_cond_batch(model, conds, x, timestep, model_options)

^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^

File "C:\AI\ComfyUI\comfy\samplers.py", line 211, in calc_cond_batch

return calccond_batch_outer(model, conds, x_in, timestep, model_options)

^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^

File "C:\AI\ComfyUI\comfy\samplers.py", line 219, in calccond_batch_outer

return executor.execute(model, conds, x_in, timestep, model_options)

^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^

File "C:\AI\ComfyUI\comfy\patcher_extension.py", line 113, in execute

return self.original(*args, **kwargs)

^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^

File "C:\AI\ComfyUI\comfy\samplers.py", line 335, in calccond_batch

output = model.apply_model(input_x, timestep_, **c).chunk(batch_chunks)

^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^

File "C:\AI\ComfyUI\comfy\model_base.py", line 205, in apply_model

return comfy.patcher_extension.WrapperExecutor.new_class_executor(

^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^

File "C:\AI\ComfyUI\comfy\patcher_extension.py", line 113, in execute

return self.original(*args, **kwargs)

^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^

File "C:\AI\ComfyUI\comfy\model_base.py", line 249, in applymodel

model_output = self.diffusion_model(xc, t, context=context, control=control, transformer_options=transformer_options, **extra_conds)

^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^

File "C:\AI\ComfyUI\venv\Lib\site-packages\torch\nn\modules\module.py", line 1779, in wrappedcall_impl

return self._call_impl(*args, **kwargs)

^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^

File "C:\AI\ComfyUI\venv\Lib\site-packages\torch\nn\modules\module.py", line 1790, in callimpl

return forward_call(*args, **kwargs)

^^^^^^^^^^^^^^^^^^^^^^^^^^^^^

File "C:\AI\ComfyUI\comfy\ldm\minimax\model.py", line 566, in forward

out = comfy.patcher_extension.WrapperExecutor.new_class_executor(

^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^

File "C:\AI\ComfyUI\comfy\patcher_extension.py", line 113, in execute

return self.original(*args, **kwargs)

^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^

File "C:\AI\ComfyUI\custom_nodes\ComfyUI-DaSiWa-Nodes\nodes\nodes_minimax_h3_cache.py", line 277, in patched_forward

video_result, audio_result = self.final_layer(hidden_states, timestep_embedding, video_segment, audio_segment)

^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^

File "C:\AI\ComfyUI\venv\Lib\site-packages\torch\nn\modules\module.py", line 1779, in wrappedcall_impl

return self._call_impl(*args, **kwargs)

^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^

File "C:\AI\ComfyUI\venv\Lib\site-packages\torch\nn\modules\module.py", line 1790, in callimpl

return forward_call(*args, **kwargs)

^^^^^^^^^^^^^^^^^^^^^^^^^^^^^

TypeError: FinalLayer.forward() missing 3 required positional arguments: 'sigma', 'sample_sigmas', and 'shifts'

I got some problem with POV shot, i'm using your 8 steps model with your workflow. It can't produce a POV video as i prompt

When the first video finishes, the rest of the queue instantly produces the same video. So I get the same video multiple times. Is this because of the workflow or am I doing something wrong? The seed is set to random.

Curious, I'm running an RTX 4090 with 24GB VRAM and 64GB system RAM, if I run this workflow with the 3D Latent Upscale on cuda and fp16 at 0.32 MP I get an OOM, if I run on cpu at fp16 it runs but so slow it's not even worth using. Is this normal for a 4090 not to be able to handle a 0.32 MP render with latent upscaling? And my 0.32MP is just what I use to run tests, normally run at 0.65MP then a 2.0 upscale with RTX & Refiner, no issues. Btw, still waiting for this workflow to progress to step 1/8 and it hasn't done it while writing this note, so a no go for me, unless I'm missing something.

dasiwa node pack ver 0.4.27+comfyui version 0.42.3 causes the following error: Advanced LoRA Loader is missing a required input: use_cache

Upscaling Modes - RTX, Model, Simple, H3 Latent

Is chooes only ones or can multip?

and the Model where can search?

I encountered an error when using Latent Upscaler in version v18. I turned it off to generate normally, but the Latent Upscaler feature cannot be used

# ComfyUI Error Report

## Error Details

- Node ID: 1512:2591

- Node Type: SamplerCustomAdvanced

- Exception Type: RuntimeError

- Exception Message: RuntimeError: shape mismatch: value tensor of shape [864, 96] cannot be broadcast to indexing result of shape [3456, 96]

## Stack Trace

```

File "E:\BaiduNetdiskDownload\ComfyUI-aki-v3.2\ComfyUI\execution.py", line 545, in execute

output_data, output_ui, has_subgraph, has_pending_tasks = await get_output_data(prompt_id, unique_id, obj, input_data_all, execution_block_cb=execution_block_cb, pre_execute_cb=pre_execute_cb, v3_data=v3_data)

^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^

File "E:\BaiduNetdiskDownload\ComfyUI-aki-v3.2\ComfyUI\execution.py", line 344, in get_output_data

return_values = await asyncmap_node_over_list(prompt_id, unique_id, obj, input_data_all, obj.FUNCTION, allow_interrupt=True, execution_block_cb=execution_block_cb, pre_execute_cb=pre_execute_cb, v3_data=v3_data)

^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^

File "E:\BaiduNetdiskDownload\ComfyUI-aki-v3.2\ComfyUI\execution.py", line 318, in asyncmap_node_over_list

await process_inputs(input_dict, i)

File "E:\BaiduNetdiskDownload\ComfyUI-aki-v3.2\ComfyUI\execution.py", line 306, in process_inputs

result = f(**inputs)

File "E:\BaiduNetdiskDownload\ComfyUI-aki-v3.2\ComfyUI\comfy_api\internal\__init__.py", line 149, in wrapped_func

return method(locked_class, **inputs)

File "E:\BaiduNetdiskDownload\ComfyUI-aki-v3.2\ComfyUI\comfy_api\latest\_io.py", line 1990, in EXECUTE_NORMALIZED

to_return = cls.execute(*args, **kwargs)

File "E:\BaiduNetdiskDownload\ComfyUI-aki-v3.2\ComfyUI\comfy_extras\nodes_custom_sampler.py", line 1055, in execute

samples = guider.sample(noise.generate_noise(latent), latent_image, sampler, sigmas, denoise_mask=noise_mask, callback=callback, disable_pbar=disable_pbar, seed=noise.seed)

File "E:\BaiduNetdiskDownload\ComfyUI-aki-v3.2\ComfyUI\comfy\samplers.py", line 1335, in sample

output = executor.execute(noise, latent_image, sampler, sigmas, denoise_mask, callback, disable_pbar, seed, latent_shapes=latent_shapes)

File "E:\BaiduNetdiskDownload\ComfyUI-aki-v3.2\ComfyUI\comfy\patcher_extension.py", line 114, in execute

return self.wrappers[self.idx](self, args, *kwargs)

~~~~~~~~~~~~~~~~~~~~~~~^^^^^^^^^^^^^^^^^^^^^^^

File "E:\BaiduNetdiskDownload\ComfyUI-aki-v3.2\ComfyUI\custom_nodes\comfyui-kjnodes\nodes\preview_override_node.py", line 685, in call

return executor(noise, latent_image, sampler, sigmas, denoise_mask, new_callback, disable_pbar, seed, latent_shapes=latent_shapes)

File "E:\BaiduNetdiskDownload\ComfyUI-aki-v3.2\ComfyUI\comfy\patcher_extension.py", line 106, in call

return new_executor.execute(*args, **kwargs)

~~~~~~~~~~~~~~~~~~~~^^^^^^^^^^^^^^^^^

File "E:\BaiduNetdiskDownload\ComfyUI-aki-v3.2\ComfyUI\comfy\patcher_extension.py", line 113, in execute

return self.original(*args, **kwargs)

~~~~~~~~~~~~~^^^^^^^^^^^^^^^^^

File "E:\BaiduNetdiskDownload\ComfyUI-aki-v3.2\ComfyUI\comfy\samplers.py", line 1262, in outer_sample

output = self.inner_sample(noise, latent_image, device, sampler, sigmas, denoise_mask, callback, disable_pbar, seed, latent_shapes=latent_shapes)

File "E:\BaiduNetdiskDownload\ComfyUI-aki-v3.2\ComfyUI\comfy\samplers.py", line 1237, in inner_sample

samples = executor.execute(self, sigmas, extra_args, callback, noise, latent_image, denoise_mask, disable_pbar)

File "E:\BaiduNetdiskDownload\ComfyUI-aki-v3.2\ComfyUI\comfy\patcher_extension.py", line 113, in execute

return self.original(*args, **kwargs)

~~~~~~~~~~~~~^^^^^^^^^^^^^^^^^

File "E:\BaiduNetdiskDownload\ComfyUI-aki-v3.2\ComfyUI\comfy\samplers.py", line 1005, in sample

samples = self.sampler_function(model_k, noise, sigmas, extra_args=extra_args, callback=k_callback, disable=disable_pbar, **self.extra_options)

File "E:\BaiduNetdiskDownload\ComfyUI-aki-v3.2\python\Lib\site-packages\torch\utils\_contextlib.py", line 124, in decorate_context

return func(*args, **kwargs)

File "E:\BaiduNetdiskDownload\ComfyUI-aki-v3.2\ComfyUI\comfy\k_diffusion\sampling.py", line 205, in sample_euler

denoised = model(x, sigma_hat s_in, *extra_args)

File "E:\BaiduNetdiskDownload\ComfyUI-aki-v3.2\ComfyUI\comfy\samplers.py", line 640, in call

out = self.inner_model(x, sigma, model_options=model_options, seed=seed)

File "E:\BaiduNetdiskDownload\ComfyUI-aki-v3.2\ComfyUI\comfy\samplers.py", line 1208, in call

return self.outer_predict_noise(*args, **kwargs)

~~~~~~~~~~~~~~~~~~~~~~~~^^^^^^^^^^^^^^^^^

File "E:\BaiduNetdiskDownload\ComfyUI-aki-v3.2\ComfyUI\comfy\samplers.py", line 1215, in outer_predict_noise

).execute(x, timestep, model_options, seed)

~~~~~~~^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^

File "E:\BaiduNetdiskDownload\ComfyUI-aki-v3.2\ComfyUI\comfy\patcher_extension.py", line 113, in execute

return self.original(*args, **kwargs)

~~~~~~~~~~~~~^^^^^^^^^^^^^^^^^

File "E:\BaiduNetdiskDownload\ComfyUI-aki-v3.2\ComfyUI\comfy\samplers.py", line 1218, in predict_noise

return sampling_function(self.inner_model, x, timestep, self.conds.get("negative", None), self.conds.get("positive", None), self.cfg, model_options=model_options, seed=seed)

File "E:\BaiduNetdiskDownload\ComfyUI-aki-v3.2\ComfyUI\comfy\samplers.py", line 620, in sampling_function

out = calc_cond_batch(model, conds, x, timestep, model_options)

File "E:\BaiduNetdiskDownload\ComfyUI-aki-v3.2\ComfyUI\comfy\samplers.py", line 211, in calc_cond_batch

return calccond_batch_outer(model, conds, x_in, timestep, model_options)

File "E:\BaiduNetdiskDownload\ComfyUI-aki-v3.2\ComfyUI\comfy\samplers.py", line 219, in calccond_batch_outer

return executor.execute(model, conds, x_in, timestep, model_options)

~~~~~~~~~~~~~~~~^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^

File "E:\BaiduNetdiskDownload\ComfyUI-aki-v3.2\ComfyUI\comfy\patcher_extension.py", line 113, in execute

return self.original(*args, **kwargs)

~~~~~~~~~~~~~^^^^^^^^^^^^^^^^^

File "E:\BaiduNetdiskDownload\ComfyUI-aki-v3.2\ComfyUI\comfy\samplers.py", line 335, in calccond_batch

output = model.apply_model(input_x, timestep_, **c).chunk(batch_chunks)

~~~~~~~~~~~~~~~~~^^^^^^^^^^^^^^^^^^^^^^^^^

File "E:\BaiduNetdiskDownload\ComfyUI-aki-v3.2\ComfyUI\comfy\model_base.py", line 205, in apply_model

return comfy.patcher_extension.WrapperExecutor.new_class_executor(

~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~

...<2 lines>...

comfy.patcher_extension.get_all_wrappers(comfy.patcher_extension.WrappersMP.APPLY_MODEL, transformer_options)

~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~

).execute(x, t, c_concat, c_crossattn, control, transformer_options, **kwargs)

~~~~~~~~~^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^

File "E:\BaiduNetdiskDownload\ComfyUI-aki-v3.2\ComfyUI\comfy\patcher_extension.py", line 113, in execute

return self.original(*args, **kwargs)

~~~~~~~~~~~~~^^^^^^^^^^^^^^^^^

File "E:\BaiduNetdiskDownload\ComfyUI-aki-v3.2\ComfyUI\comfy\model_base.py", line 249, in applymodel

model_output = self.diffusion_model(xc, t, context=context, control=control, transformer_options=transformer_options, **extra_conds)

File "E:\BaiduNetdiskDownload\ComfyUI-aki-v3.2\python\Lib\site-packages\torch\nn\modules\module.py", line 1778, in wrappedcall_impl

return self._call_impl(*args, **kwargs)

~~~~~~~~~~~~~~~^^^^^^^^^^^^^^^^^

File "E:\BaiduNetdiskDownload\ComfyUI-aki-v3.2\python\Lib\site-packages\torch\nn\modules\module.py", line 1789, in callimpl

return forward_call(*args, **kwargs)

File "E:\BaiduNetdiskDownload\ComfyUI-aki-v3.2\ComfyUI\comfy\ldm\minimax\model.py", line 540, in forward

out = comfy.patcher_extension.WrapperExecutor.new_class_executor(

...<2 lines>...

comfy.patcher_extension.get_all_wrappers(comfy.patcher_extension.WrappersMP.DIFFUSION_MODEL, transformer_options)

).execute(x, timestep, context, transformer_options, minimax_payload=minimax_payload,

denoise_mask=denoise_mask, audio_denoise_mask=audio_denoise_mask, **kwargs)

File "E:\BaiduNetdiskDownload\ComfyUI-aki-v3.2\ComfyUI\comfy\patcher_extension.py", line 113, in execute

return self.original(*args, **kwargs)

~~~~~~~~~~~~~^^^^^^^^^^^^^^^^^

File "E:\BaiduNetdiskDownload\ComfyUI-aki-v3.2\ComfyUI\comfy\ldm\minimax\model.py", line 654, in _forward

all_video_rows[~img_update] = cond_video_rows

~~~~~~~~~~~~~~^^^^^^^^^^^^^

```

## System Information

- ComfyUI Version: 0.34.2

- Arguments: E:\BaiduNetdiskDownload\ComfyUI-aki-v3.2\ComfyUI\main.py --windows-standalone-build --fast --listen 0.0.0.0 --enable-cors-header * --use-sage-attention --auto-launch

- OS: win32

- Python Version: 3.13.11 (tags/v3.13.11:6278944, Dec 5 2025, 16:26:58) [MSC v.1944 64 bit (AMD64)]

- Embedded Python: false

- PyTorch Version: 2.13.0+cu130

## Devices

- Name: cuda:0 NVIDIA GeForce RTX 5080 : native

- Type: cuda

- VRAM Total: 17066033152

- VRAM Free: 13333088768

- Torch VRAM Total: 58720256

- Torch VRAM Free: 15124992

## Logs

```

2026-08-30T00:28:26.562072 - 2026-08-30T00:28:26.688759 - 2026-08-30T00:28:26.688802 - [2026-08-30T00:28:26.688810 - 2KFetch error: 2026-08-30T00:28:26.688814 -

2026-08-30T00:28:26.688820 - E:\BaiduNetdiskDownload\ComfyUI-aki-v3.2026-08-30T00:28:26.688830 - 22026-08-30T00:28:26.688836 - \ComfyUI\custom_nodes\ComfyUI_Custom_No2026-08-30T00:28:26.688840 -

2026-08-30T00:28:26.688843 - des_AlekPet2026-08-30T00:28:26.688847 -

2026-08-30T00:28:26.689619 -

2026-08-30T00:28:26.689644 - Cmd2026-08-30T00:28:26.689653 - (2026-08-30T00:28:26.689662 - 'git'2026-08-30T00:28:26.689670 - )2026-08-30T00:28:26.689676 - failed due to: exit 2026-08-30T00:28:26.689683 - code2026-08-30T00:28:26.689690 - (2026-08-30T00:28:26.689697 - 1282026-08-30T00:28:26.689704 - )2026-08-30T00:28:26.689709 -

2026-08-30T00:28:26.689713 - cmdline: git fetch -v -- origin2026-08-30T00:28:26.689716 -

2026-08-30T00:28:26.689722 - stderr: 2026-08-30T00:28:26.689727 - 'fatal: unable to access 2026-08-30T00:28:26.689733 -

2026-08-30T00:28:26.689746 - '2026-08-30T00:28:26.689754 - https://github.com/AlekPet/ComfyUI_Custom_Nodes_AlekPet/2026-08-30T00:28:26.689761 - ': Failed to connect 2026-08-30T00:28:26.689766 -

2026-08-30T00:28:26.689772 - to github.com:443 after 21071 ms: Could not connect to server'2026-08-30T00:28:26.689777 -

2026-08-30T00:28:26.689781 - CUSTOM NODE CHECK: Error2026-08-30T00:28:26.689784 -

2026-08-30T00:28:26.689788 -

2026-08-30T00:28:26.689941 - 

Fetching: E:\BaiduNetdiskDownload\ComfyUI-aki-v3.2\ComfyUI\custom_nodes\ComfyUI-VideoHelperSuite2026-08-30T00:28:26.689946 - 2026-08-30T00:28:26.695232 - 2026-08-30T00:28:26.695279 - [2026-08-30T00:28:26.695295 - 2KFetch error: 2026-08-30T00:28:26.695304 -

2026-08-30T00:28:26.695312 - E:\BaiduNetdiskDownload\ComfyUI-aki-v3.2026-08-30T00:28:26.695329 - 22026-08-30T00:28:26.695338 - \ComfyUI\custom_nodes\ComfyUI_UltimateS2026-08-30T00:28:26.695344 -

2026-08-30T00:28:26.695351 - DUpscale2026-08-30T00:28:26.695356 -

2026-08-30T00:28:26.696294 -

2026-08-30T00:28:26.696337 - Cmd2026-08-30T00:28:26.696351 - (2026-08-30T00:28:26.696363 - 'git'2026-08-30T00:28:26.696373 - )2026-08-30T00:28:26.696383 - failed due to: exit 2026-08-30T00:28:26.696392 - code2026-08-30T00:28:26.696402 - (2026-08-30T00:28:26.696412 - 1282026-08-30T00:28:26.696421 - )2026-08-30T00:28:26.696429 -

2026-08-30T00:28:26.696434 - cmdline: git fetch -v -- origin2026-08-30T00:28:26.696441 -

2026-08-30T00:28:26.696447 - stderr: 2026-08-30T00:28:26.696454 - 'fatal: unable to access 2026-08-30T00:28:26.696463 -

2026-08-30T00:28:26.696472 - '2026-08-30T00:28:26.696482 - https://github.com/ssitu/ComfyUI_UltimateSDUpscale/2026-08-30T00:28:26.696492 - ': Failed to connect to 2026-08-30T00:28:26.696499 -

2026-08-30T00:28:26.696507 - github.com:443 after 21093 ms: Could not connect to server'2026-08-30T00:28:26.696514 -

2026-08-30T00:28:26.696519 - CUSTOM NODE CHECK: Error2026-08-30T00:28:26.696523 -

2026-08-30T00:28:26.696528 -

2026-08-30T00:28:26.696735 - 

Fetching: E:\BaiduNetdiskDownload\ComfyUI-aki-v3.2\ComfyUI\custom_nodes\ComfyUI-WanVideoWrapper2026-08-30T00:28:26.696741 - 2026-08-30T00:28:48.005479 - 2026-08-30T00:28:48.005514 - [2026-08-30T00:28:48.005522 - 2KFetch error: 2026-08-30T00:28:48.005527 -

2026-08-30T00:28:48.005532 - E:\BaiduNetdiskDownload\ComfyUI-aki-v3.2026-08-30T00:28:48.005543 - 22026-08-30T00:28:48.005549 - \ComfyUI\custom_nodes\ComfyUI_IPAdapter2026-08-30T00:28:48.005552 -

2026-08-30T00:28:48.005556 - _plus2026-08-30T00:28:48.005560 -

2026-08-30T00:28:48.005984 -

2026-08-30T00:28:48.006002 - Cmd2026-08-30T00:28:48.006011 - (2026-08-30T00:28:48.006020 - 'git'2026-08-30T00:28:48.006028 - )2026-08-30T00:28:48.006034 - failed due to: exit 2026-08-30T00:28:48.006042 - code2026-08-30T00:28:48.006049 - (2026-08-30T00:28:48.006057 - 1282026-08-30T00:28:48.006064 - )2026-08-30T00:28:48.006069 -

2026-08-30T00:28:48.006074 - cmdline: git fetch -v -- origin2026-08-30T00:28:48.006079 -

2026-08-30T00:28:48.006083 - stderr: 2026-08-30T00:28:48.006088 - 'fatal: unable to access 2026-08-30T00:28:48.006095 -

2026-08-30T00:28:48.006100 - '2026-08-30T00:28:48.006109 - https://github.com/cubiq/ComfyUI_IPAdapter_plus/2026-08-30T00:28:48.006117 - ': Failed to connect to 2026-08-30T00:28:48.006122 -

2026-08-30T00:28:48.006129 - github.com:443 after 21079 ms: Could not connect to server'2026-08-30T00:28:48.006134 -

2026-08-30T00:28:48.006139 - CUSTOM NODE CHECK: Error2026-08-30T00:28:48.006142 -

2026-08-30T00:28:48.006145 -

2026-08-30T00:28:48.006280 - 

Fetching: E:\BaiduNetdiskDownload\ComfyUI-aki-v3.2\ComfyUI\custom_nodes\ComfyUI-RMBG2026-08-30T00:28:48.006284 - 2026-08-30T00:28:48.154364 - 2026-08-30T00:28:48.154399 - [2026-08-30T00:28:48.154407 - 2KFetch error: 2026-08-30T00:28:48.154412 -

2026-08-30T00:28:48.154417 - E:\BaiduNetdiskDownload\ComfyUI-aki-v3.2026-08-30T00:28:48.154428 - 22026-08-30T00:28:48.154433 - \ComfyUI\custom_nodes\ComfyUI-WanVideoW2026-08-30T00:28:48.154437 -

2026-08-30T00:28:48.154441 - rapper2026-08-30T00:28:48.154444 -

2026-08-30T00:28:48.154867 -

2026-08-30T00:28:48.154888 - Cmd2026-08-30T00:28:48.154921 - (2026-08-30T00:28:48.154953 - 'git'2026-08-30T00:28:48.154964 - )2026-08-30T00:28:48.154970 - failed due to: exit 2026-08-30T00:28:48.154977 - code2026-08-30T00:28:48.154984 - (2026-08-30T00:28:48.154992 - 1282026-08-30T00:28:48.154998 - )2026-08-30T00:28:48.155005 -

2026-08-30T00:28:48.155009 - cmdline: git fetch -v -- origin2026-08-30T00:28:48.155013 -

2026-08-30T00:28:48.155016 - stderr: 2026-08-30T00:28:48.155021 - 'fatal: unable to access 2026-08-30T00:28:48.155026 -

2026-08-30T00:28:48.155034 - '2026-08-30T00:28:48.155041 - https://github.com/kijai/ComfyUI-WanVideoWrapper/2026-08-30T00:28:48.155048 - ': Failed to connect to 2026-08-30T00:28:48.155053 -

2026-08-30T00:28:48.155061 - github.com:443 after 21097 ms: Could not connect to server'2026-08-30T00:28:48.155065 -

2026-08-30T00:28:48.155069 - CUSTOM NODE CHECK: Error2026-08-30T00:28:48.155072 -

2026-08-30T00:28:48.155076 -

2026-08-30T00:28:48.155210 - 

Fetching: E:\BaiduNetdiskDownload\ComfyUI-aki-v3.2\ComfyUI\custom_nodes\ComfyUI-LTXVideo2026-08-30T00:28:48.155214 - 2026-08-30T00:28:48.157382 - 2026-08-30T00:28:48.157410 - [2026-08-30T00:28:48.157418 - 2KFetch error: 2026-08-30T00:28:48.157422 -

2026-08-30T00:28:48.157427 - E:\BaiduNetdiskDownload\ComfyUI-aki-v3.2026-08-30T00:28:48.157437 - 22026-08-30T00:28:48.157442 - \ComfyUI\custom_nodes\ComfyUI-VideoHelp2026-08-30T00:28:48.157446 -

2026-08-30T00:28:48.157449 - erSuite2026-08-30T00:28:48.157453 -

2026-08-30T00:28:48.158806 -

2026-08-30T00:28:48.158824 - Cmd2026-08-30T00:28:48.158833 - (2026-08-30T00:28:48.158842 - 'git'2026-08-30T00:28:48.158849 - )2026-08-30T00:28:48.158854 - failed due to: exit 2026-08-30T00:28:48.158861 - code2026-08-30T00:28:48.158867 - (2026-08-30T00:28:48.158874 - 1282026-08-30T00:28:48.158880 - )2026-08-30T00:28:48.158885 -

2026-08-30T00:28:48.158889 - cmdline: git fetch -v -- origin2026-08-30T00:28:48.158893 -

2026-08-30T00:28:48.158896 - stderr: 2026-08-30T00:28:48.158901 - 'fatal: unable to access 2026-08-30T00:28:48.158906 -

2026-08-30T00:28:48.158912 - '2026-08-30T00:28:48.158919 - https://github.com/Kosinkadink/ComfyUI-VideoHelperSuite/2026-08-30T00:28:48.158927 - ': Failed to connect 2026-08-30T00:28:48.158932 -

2026-08-30T00:28:48.158937 - to github.com:443 after 21104 ms: Could not connect to server'2026-08-30T00:28:48.158942 -

2026-08-30T00:28:48.158945 - CUSTOM NODE CHECK: Error2026-08-30T00:28:48.158948 -

2026-08-30T00:28:48.158951 -

2026-08-30T00:29:09.508274 - 2026-08-30T00:29:09.508308 - [2026-08-30T00:29:09.508316 - 2KFetch error: 2026-08-30T00:29:09.508320 -

2026-08-30T00:29:09.508325 - E:\BaiduNetdiskDownload\ComfyUI-aki-v3.2026-08-30T00:29:09.508335 - 22026-08-30T00:29:09.508340 - \ComfyUI\custom_nodes\ComfyUI-RMBG2026-08-30T00:29:09.508344 -

2026-08-30T00:29:09.508765 -

2026-08-30T00:29:09.508784 - Cmd2026-08-30T00:29:09.508793 - (2026-08-30T00:29:09.508801 - 'git'2026-08-30T00:29:09.508810 - )2026-08-30T00:29:09.508815 - failed due to: exit 2026-08-30T00:29:09.508821 - code2026-08-30T00:29:09.508828 - (2026-08-30T00:29:09.508835 - 1282026-08-30T00:29:09.508841 - )2026-08-30T00:29:09.508846 -

2026-08-30T00:29:09.508850 - cmdline: git fetch -v -- origin2026-08-30T00:29:09.508853 -

2026-08-30T00:29:09.508857 - stderr: 2026-08-30T00:29:09.508862 - 'fatal: unable to access '2026-08-30T00:29:09.508869 - https://github.com/1038lab/ComfyUI-RMBG/2026-08-30T00:29:09.508876 - ': 2026-08-30T00:29:09.508881 -

2026-08-30T00:29:09.508886 - Failed to connect to github.com:443 after 21143 ms: Could not connect to 2026-08-30T00:29:09.508891 -

2026-08-30T00:29:09.508897 - server'2026-08-30T00:29:09.508906 -

2026-08-30T00:29:09.508909 - CUSTOM NODE CHECK: Error2026-08-30T00:29:09.508913 -

2026-08-30T00:29:09.508916 -

2026-08-30T00:29:09.624784 - 2026-08-30T00:29:09.624810 - [2026-08-30T00:29:09.624816 - 2KFetch error: 2026-08-30T00:29:09.624819 -

2026-08-30T00:29:09.624823 - E:\BaiduNetdiskDownload\ComfyUI-aki-v3.2026-08-30T00:29:09.624830 - 22026-08-30T00:29:09.624834 - \ComfyUI\custom_nodes\ComfyUI-LTXVideo2026-08-30T00:29:09.624837 -

2026-08-30T00:29:09.625166 -

2026-08-30T00:29:09.625177 - Cmd2026-08-30T00:29:09.625184 - (2026-08-30T00:29:09.625190 - 'git'2026-08-30T00:29:09.625195 - )2026-08-30T00:29:09.625198 - failed due to: exit 2026-08-30T00:29:09.625203 - code2026-08-30T00:29:09.625207 - (2026-08-30T00:29:09.625212 - 1282026-08-30T00:29:09.625217 - )2026-08-30T00:29:09.625220 -

2026-08-30T00:29:09.625223 - cmdline: git fetch -v -- origin2026-08-30T00:29:09.625225 -

2026-08-30T00:29:09.625229 - stderr: 2026-08-30T00:29:09.625233 - 'fatal: unable to access 2026-08-30T00:29:09.625236 -

2026-08-30T00:29:09.625240 - '2026-08-30T00:29:09.625246 - https://github.com/Lightricks/ComfyUI-LTXVideo/2026-08-30T00:29:09.625250 - ': Failed to connect to 2026-08-30T00:29:09.625254 -

2026-08-30T00:29:09.625257 - github.com:443 after 21114 ms: Could not connect to server'2026-08-30T00:29:09.625261 -

2026-08-30T00:29:09.625263 - CUSTOM NODE CHECK: Error2026-08-30T00:29:09.625265 -

2026-08-30T00:29:09.625267 -

2026-08-30T00:30:23.060587 - 2026-08-30T00:30:23.060631 - [2026-08-30T00:30:23.060643 - 2KFetch error: 2026-08-30T00:30:23.060650 -

2026-08-30T00:30:23.060658 - E:\BaiduNetdiskDownload\ComfyUI-aki-v3.2026-08-30T00:30:23.060672 - 22026-08-30T00:30:23.060680 - \ComfyUI\custom_nodes\ComfyUI-Manager2026-08-30T00:30:23.060686 -

2026-08-30T00:30:23.061824 -

2026-08-30T00:30:23.061849 - Cmd2026-08-30T00:30:23.061864 - (2026-08-30T00:30:23.061873 - 'git'2026-08-30T00:30:23.061880 - )2026-08-30T00:30:23.061887 - failed due to: exit 2026-08-30T00:30:23.061893 - code2026-08-30T00:30:23.061900 - (2026-08-30T00:30:23.061907 - 1282026-08-30T00:30:23.061914 - )2026-08-30T00:30:23.061920 -

2026-08-30T00:30:23.061924 - cmdline: git fetch -v -- origin2026-08-30T00:30:23.061928 -

2026-08-30T00:30:23.061932 - stderr: 2026-08-30T00:30:23.061937 - 'fatal: unable to access 2026-08-30T00:30:23.061942 -

2026-08-30T00:30:23.061947 - '2026-08-30T00:30:23.061954 - https://github.com/Comfy-Org/ComfyUI-Manager/2026-08-30T00:30:23.061962 - ': Recv failure: Connection was 2026-08-30T00:30:23.061967 -

2026-08-30T00:30:23.061973 - reset'2026-08-30T00:30:23.061978 -

2026-08-30T00:30:23.061981 - CUSTOM NODE CHECK: Error2026-08-30T00:30:23.061985 -

2026-08-30T00:30:23.061988 -

2026-08-30T00:30:23.062817 - 

Fetching done.2026-08-30T00:30:23.062829 -

2026-08-30T00:30:23.269453 - FETCH DATA from: https://raw.githubusercontent.com/ltdrdata/ComfyUI-Manager/main/extension-node-map.json2026-08-30T00:30:23.269464 - 2026-08-30T00:30:24.743315 - [DONE]2026-08-30T00:30:24.743336 -

2026-08-30T00:31:38.862633 - [INFO] got prompt

2026-08-30T00:31:39.386850 - [INFO] 0 models unloaded.

2026-08-30T00:31:39.439162 - [INFO] Model MiniMaxH3 prepared for dynamic VRAM loading. 19995MB Staged. 208 patches attached. Force pre-loaded 210 weights: 1175 KB.

2026-08-30T00:31:39.452899 - [INFO] Patching torch settings: torch.backends.cuda.matmul.allow_fp16_accumulation = True

2026-08-30T00:31:39.546503 -

0%| | 0/4 [00:00<?, ?it/s, Model Initializing ... ]2026-08-30T00:31:39.550874 -

0%| | 0/4 [00:00<?, ?it/s, Model Initializing ... ]2026-08-30T00:31:39.550892 -

2026-08-30T00:31:39.550986 - [INFO] Patching torch settings: torch.backends.cuda.matmul.allow_fp16_accumulation = False

2026-08-30T00:31:39.643806 - [ERROR] !!! Exception during processing !!! shape mismatch: value tensor of shape [864, 96] cannot be broadcast to indexing result of shape [3456, 96]

2026-08-30T00:31:39.646897 - [ERROR] Traceback (most recent call last):

File "E:\BaiduNetdiskDownload\ComfyUI-aki-v3.2\ComfyUI\execution.py", line 545, in execute

output_data, output_ui, has_subgraph, has_pending_tasks = await get_output_data(prompt_id, unique_id, obj, input_data_all, execution_block_cb=execution_block_cb, pre_execute_cb=pre_execute_cb, v3_data=v3_data)

^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^

File "E:\BaiduNetdiskDownload\ComfyUI-aki-v3.2\ComfyUI\execution.py", line 344, in get_output_data

return_values = await asyncmap_node_over_list(prompt_id, unique_id, obj, input_data_all, obj.FUNCTION, allow_interrupt=True, execution_block_cb=execution_block_cb, pre_execute_cb=pre_execute_cb, v3_data=v3_data)

^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^

File "E:\BaiduNetdiskDownload\ComfyUI-aki-v3.2\ComfyUI\execution.py", line 318, in asyncmap_node_over_list

await process_inputs(input_dict, i)

File "E:\BaiduNetdiskDownload\ComfyUI-aki-v3.2\ComfyUI\execution.py", line 306, in process_inputs

result = f(**inputs)

File "E:\BaiduNetdiskDownload\ComfyUI-aki-v3.2\ComfyUI\comfy_api\internal\__init__.py", line 149, in wrapped_func

return method(locked_class, **inputs)

File "E:\BaiduNetdiskDownload\ComfyUI-aki-v3.2\ComfyUI\comfy_api\latest\_io.py", line 1990, in EXECUTE_NORMALIZED

to_return = cls.execute(*args, **kwargs)

File "E:\BaiduNetdiskDownload\ComfyUI-aki-v3.2\ComfyUI\comfy_extras\nodes_custom_sampler.py", line 1055, in execute

samples = guider.sample(noise.generate_noise(latent), latent_image, sampler, sigmas, denoise_mask=noise_mask, callback=callback, disable_pbar=disable_pbar, seed=noise.seed)

File "E:\BaiduNetdiskDownload\ComfyUI-aki-v3.2\ComfyUI\comfy\samplers.py", line 1335, in sample

output = executor.execute(noise, latent_image, sampler, sigmas, denoise_mask, callback, disable_pbar, seed, latent_shapes=latent_shapes)

File "E:\BaiduNetdiskDownload\ComfyUI-aki-v3.2\ComfyUI\comfy\patcher_extension.py", line 114, in execute

return self.wrappers[self.idx](self, args, *kwargs)

~~~~~~~~~~~~~~~~~~~~~~~^^^^^^^^^^^^^^^^^^^^^^^

File "E:\BaiduNetdiskDownload\ComfyUI-aki-v3.2\ComfyUI\custom_nodes\comfyui-kjnodes\nodes\preview_override_node.py", line 685, in call

return executor(noise, latent_image, sampler, sigmas, denoise_mask, new_callback, disable_pbar, seed, latent_shapes=latent_shapes)

File "E:\BaiduNetdiskDownload\ComfyUI-aki-v3.2\ComfyUI\comfy\patcher_extension.py", line 106, in call

return new_executor.execute(*args, **kwargs)

~~~~~~~~~~~~~~~~~~~~^^^^^^^^^^^^^^^^^

File "E:\BaiduNetdiskDownload\ComfyUI-aki-v3.2\ComfyUI\comfy\patcher_extension.py", line 113, in execute

return self.original(*args, **kwargs)

~~~~~~~~~~~~~^^^^^^^^^^^^^^^^^

File "E:\BaiduNetdiskDownload\ComfyUI-aki-v3.2\ComfyUI\comfy\samplers.py", line 1262, in outer_sample

output = self.inner_sample(noise, latent_image, device, sampler, sigmas, denoise_mask, callback, disable_pbar, seed, latent_shapes=latent_shapes)

File "E:\BaiduNetdiskDownload\ComfyUI-aki-v3.2\ComfyUI\comfy\samplers.py", line 1237, in inner_sample

samples = executor.execute(self, sigmas, extra_args, callback, noise, latent_image, denoise_mask, disable_pbar)

File "E:\BaiduNetdiskDownload\ComfyUI-aki-v3.2\ComfyUI\comfy\patcher_extension.py", line 113, in execute

return self.original(*args, **kwargs)

~~~~~~~~~~~~~^^^^^^^^^^^^^^^^^

File "E:\BaiduNetdiskDownload\ComfyUI-aki-v3.2\ComfyUI\comfy\samplers.py", line 1005, in sample

samples = self.sampler_function(model_k, noise, sigmas, extra_args=extra_args, callback=k_callback, disable=disable_pbar, **self.extra_options)

File "E:\BaiduNetdiskDownload\ComfyUI-aki-v3.2\python\Lib\site-packages\torch\utils\_contextlib.py", line 124, in decorate_context

return func(*args, **kwargs)

File "E:\BaiduNetdiskDownload\ComfyUI-aki-v3.2\ComfyUI\comfy\k_diffusion\sampling.py", line 205, in sample_euler

denoised = model(x, sigma_hat s_in, *extra_args)

File "E:\BaiduNetdiskDownload\ComfyUI-aki-v3.2\ComfyUI\comfy\samplers.py", line 640, in call

out = self.inner_model(x, sigma, model_options=model_options, seed=seed)

File "E:\BaiduNetdiskDownload\ComfyUI-aki-v3.2\ComfyUI\comfy\samplers.py", line 1208, in call

return self.outer_predict_noise(*args, **kwargs)

~~~~~~~~~~~~~~~~~~~~~~~~^^^^^^^^^^^^^^^^^

File "E:\BaiduNetdiskDownload\ComfyUI-aki-v3.2\ComfyUI\comfy\samplers.py", line 1215, in outer_predict_noise

).execute(x, timestep, model_options, seed)

~~~~~~~^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^

File "E:\BaiduNetdiskDownload\ComfyUI-aki-v3.2\ComfyUI\comfy\patcher_extension.py", line 113, in execute

return self.original(*args, **kwargs)

~~~~~~~~~~~~~^^^^^^^^^^^^^^^^^

File "E:\BaiduNetdiskDownload\ComfyUI-aki-v3.2\ComfyUI\comfy\samplers.py", line 1218, in predict_noise

return sampling_function(self.inner_model, x, timestep, self.conds.get("negative", None), self.conds.get("positive", None), self.cfg, model_options=model_options, seed=seed)

File "E:\BaiduNetdiskDownload\ComfyUI-aki-v3.2\ComfyUI\comfy\samplers.py", line 620, in sampling_function

out = calc_cond_batch(model, conds, x, timestep, model_options)

File "E:\BaiduNetdiskDownload\ComfyUI-aki-v3.2\ComfyUI\comfy\samplers.py", line 211, in calc_cond_batch

return calccond_batch_outer(model, conds, x_in, timestep, model_options)

File "E:\BaiduNetdiskDownload\ComfyUI-aki-v3.2\ComfyUI\comfy\samplers.py", line 219, in calccond_batch_outer

return executor.execute(model, conds, x_in, timestep, model_options)

~~~~~~~~~~~~~~~~^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^

File "E:\BaiduNetdiskDownload\ComfyUI-aki-v3.2\ComfyUI\comfy\patcher_extension.py", line 113, in execute

return self.original(*args, **kwargs)

~~~~~~~~~~~~~^^^^^^^^^^^^^^^^^

File "E:\BaiduNetdiskDownload\ComfyUI-aki-v3.2\ComfyUI\comfy\samplers.py", line 335, in calccond_batch

output = model.apply_model(input_x, timestep_, **c).chunk(batch_chunks)

~~~~~~~~~~~~~~~~~^^^^^^^^^^^^^^^^^^^^^^^^^

File "E:\BaiduNetdiskDownload\ComfyUI-aki-v3.2\ComfyUI\comfy\model_base.py", line 205, in apply_model

return comfy.patcher_extension.WrapperExecutor.new_class_executor(

~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~

...<2 lines>...

comfy.patcher_extension.get_all_wrappers(comfy.patcher_extension.WrappersMP.APPLY_MODEL, transformer_options)

~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~

).execute(x, t, c_concat, c_crossattn, control, transformer_options, **kwargs)

~~~~~~~~~^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^

File "E:\BaiduNetdiskDownload\ComfyUI-aki-v3.2\ComfyUI\comfy\patcher_extension.py", line 113, in execute

return self.original(*args, **kwargs)

~~~~~~~~~~~~~^^^^^^^^^^^^^^^^^

File "E:\BaiduNetdiskDownload\ComfyUI-aki-v3.2\ComfyUI\comfy\model_base.py", line 249, in applymodel

model_output = self.diffusion_model(xc, t, context=context, control=control, transformer_options=transformer_options, **extra_conds)

File "E:\BaiduNetdiskDownload\ComfyUI-aki-v3.2\python\Lib\site-packages\torch\nn\modules\module.py", line 1778, in wrappedcall_impl

return self._call_impl(*args, **kwargs)

~~~~~~~~~~~~~~~^^^^^^^^^^^^^^^^^

File "E:\BaiduNetdiskDownload\ComfyUI-aki-v3.2\python\Lib\site-packages\torch\nn\modules\module.py", line 1789, in callimpl

return forward_call(*args, **kwargs)

File "E:\BaiduNetdiskDownload\ComfyUI-aki-v3.2\ComfyUI\comfy\ldm\minimax\model.py", line 540, in forward

out = comfy.patcher_extension.WrapperExecutor.new_class_executor(

...<2 lines>...

comfy.patcher_extension.get_all_wrappers(comfy.patcher_extension.WrappersMP.DIFFUSION_MODEL, transformer_options)

).execute(x, timestep, context, transformer_options, minimax_payload=minimax_payload,

denoise_mask=denoise_mask, audio_denoise_mask=audio_denoise_mask, **kwargs)

File "E:\BaiduNetdiskDownload\ComfyUI-aki-v3.2\ComfyUI\comfy\patcher_extension.py", line 113, in execute

return self.original(*args, **kwargs)

~~~~~~~~~~~~~^^^^^^^^^^^^^^^^^

File "E:\BaiduNetdiskDownload\ComfyUI-aki-v3.2\ComfyUI\comfy\ldm\minimax\model.py", line 654, in _forward

all_video_rows[~img_update] = cond_video_rows

~~~~~~~~~~~~~~^^^^^^^^^^^^^

RuntimeError: shape mismatch: value tensor of shape [864, 96] cannot be broadcast to indexing result of shape [3456, 96]

2026-08-30T00:31:39.650932 - [INFO] Prompt executed in 0.76 seconds

```

## Attached Workflow

Please make sure that workflow does not contain any sensitive information such as API keys or passwords.

```

Workflow too large. Please manually upload the workflow from local file system.

```

## Additional Context

(Please add any additional context or steps to reproduce the error here)

Hi Darkside/ or anyone having this- im having a serious issue with the workflow here.. whenever i try to change the video duration located inside the director node, it gives me "Node widget error" inside of the node, and comfyui says : "can't access property "options", this is undefined"

it's making the whole workflow pretty much useless cause im just stuck with 5 seconds by default... i downloaded all the missing nodes from Comfy manager. but still no luck... help please??

Hi darkside is there a workflow instead create 30 sec long vid but also in 1 run make 3 video but seamless in 1 prompt?it Will help with low vram People like me

I would like to suggest a feature for REF2VA. I want to be able to hook up the image to ollama node so that it can generate a prompt based on the image used, and then be able to feed that prompt into the detailed_description and maybe even the other fields. The goal is to be able to generate it's own story based on the image.

That means the node would need to be able to have an output for image and it would need an input for prompt.

Just a suggestion, I know your busy and might want to keep the flow simplistic. Keep up the great work.

this seems like it would be a good feature for your workflow, been having good success with it myself. on github hradec/ComfyUI-HR-Endless-Sampler