Update 2026-05-13: LTX 2.3 All-in-One v3.0 workflow published
VideoFlow LTX 2.3 distilled 1.1 All-in-One v3.0
New Features:
Text-to-Video Support with a pre-configured set of LoRAs for creating photorealistic videos.
Image-to-Video Support for both first and last frame.
Optional Audio Integration (audio-to-video) with the ability to extract voice from recordings (file or recorded clip) to remove background noise.
Consistent Character Voice through voice cloning with just a 5-second reference audio (file or recorded clip).
Video Filters for adjusting brightness, contrast, saturation, sharpness, blur, and enhancing edges and details.
Film grain for a cinematic or analog effect.
50fps Support via frame interpolation.
Improvements:
Now using LTX 2.3 distilled 1.1 resulting in better emotions, movements and audio.
Faster, less memory-intensive color correction.
More explanations and guidance integrated.
Fixes:
Audio and video are now always perfectly in sync.
Resolution input (video dimensions) for image-to-video generation now works properly.
Update 2026-04-14: LTX 2.3 I2V workflow updated
VideoFlow LTX 2.3 distilled I2V v2.0
VideoFlow 2.0 is here, bringing major performance upgrades, better quality, and more flexibility to your workflow.
Key Improvements:
Much Faster Generation: Thanks to improved samplers and schedulers, videos generate approx. twice as fast compared to Version 1.0.
Higher Quality Output: Despite the speed boost, image quality, audio quality, and prompt adherence have all been significantly improved.
Flexible Model Support: You can now freely choose between multiple model types:
Checkpoint
GGUF UNet
Diffusion model
Optimized for Low VRAM Systems: With GGUF support, VideoFlow now runs much more efficiently on systems with limited GPU memory.
Optional Sampler Preview: Disable the sampler preview to further reduce generation time.
Improved Usability: Additional guidance and hint texts help you get the most out of the workflow.
Update 2026-03-15: LTX 2.3 I2V workflow added
VideoFlow LTX 2.3 distilled I2V v1.0
This workflow provides an easy-to-use image-to-video solution for LTX 2.3, designed to work seamlessly with the distilled LoRA model. It focuses on high-quality, realistic output, with the first-stage scheduler's sigma values finely optimized for best performance.
Subgraphs are used to keep the main workflow streamlined and easy to navigate. A live preview is displayed during generation, allowing you to monitor progress and stop the process early if desired. Additionally, the first-stage video can be decoded for quick previewing. This feature lets you watch a lower-resolution version of the final video and cancel immediately if the result doesn’t meet expectations.
As the distilled LoRA already delivers impressive quality in the first stage, you can skip the second stage entirely if your hardware has limited performance. An optional color-correction node is included to compensate for LTX’s tendency to introduce subtle color and lighting shifts, ensuring consistent visual quality.
Update 2025-08-24: Wan 2.2 I2V workflow added
VideoFlow Wan 2.2 I2V v1.0
VideoFlow is now fully optimized for Wan 2.2. It supports resolutions from 480p up to 720p, with the option to upscale smoothly to 1440p at 32fps. The process is accelerated by integrating Lightning LoRA during the final two-thirds of the generation steps, ensuring faster results without compromising quality. Importantly, Lightning LoRA does not influence the initial generation steps, preserving natural and fluid movements throughout the video. SageAttention with Triton is supported but not required. Instructions on how to set up and use the workflow are included within the workflow itself.
VideoFlow Wan 2.1 I2V v1.0
This image-to-video workflow is designed to generate smooth, realistic videos at 32 fps with a strong emphasis on fast, high-resolution output. At least 16 GB of VRAM is recommended for optimal performance. For additional speed improvements, you may also install SageAttention and Triton, though these are optional.
It's fast 🚀!
Sample videos were rendered at 768 × 1152 resolution and 16 fps, consisting of 81 frames, each video taking about 6 minutes to generate. The upscaling and frame interpolation to 1536 × 2304 resolution and 32 fps took approximately another 6 minutes on an RTX 4080 with 16 GB VRAM. Lower resolutions render even faster.
Key configuration for the sample videos:
Video model: Wan2.1 SkyReels V2 I2V 14B 720P
LoRA: Lightx2v
Steps: 4
Sampler: dpmpp_sde_gpu
Scheduler: beta
💡Comprehensive usage details and instructions are provided within the workflow itself.
Sample images for input were created with my PhotoFlow workflow.
The download of the workflow contains all sample videos, including the input image with its own workflow, the initial generated video and its upscaled counterpart, allowing for convenient side-by-side comparison.
Leave a 👍 if you like the workflow 🙂.
Description
New Features:
Text-to-Video Support with a pre-configured set of LoRAs for creating photorealistic videos.
Image-to-Video Support for both first and last frame.
Optional Audio Integration (audio-to-video) with the ability to extract voice from recordings (file or recorded clip) to remove background noise.
Consistent Character Voice through voice cloning with just a 5-second reference audio (file or recorded clip).
Video Filters for adjusting brightness, contrast, saturation, sharpness, blur, and enhancing edges and details.
Film grain for a cinematic or analog effect.
50fps Support via frame interpolation.
Improvements:
Now using LTX 2.3 distilled 1.1 resulting in better emotions, movements and audio.
Faster, less memory-intensive color correction.
More explanations and guidance integrated.
Fixes:
Audio and video are now always perfectly in sync.
Resolution input (video dimensions) for image-to-video generation now works properly.
FAQ
Comments (82)
First of all: this is the best workflow I've ever seen; everything is mapped out so beautifully that I could solve almost all issues on my own.
However I have two issues that keeps popping up randomly when I try to generate a video with audio and both are connected to the LTXV Audio VAE Encode -node. They are:
IndexError: tuple index out of range
And
RuntimeError: Input type (float) and bias type (struct c10::BFloat16) should be the same
The former one I have no clue what the issue is but the 2nd one appears to hint that there is a BF32 model in the workflow somewhere that doesn't work with the BF16 VAE. Any idea how to get around this? :)
And thank you again for the amazing workflow. <3
Thank you for your kind words! Maybe you selected a wrong file in the Load Models node? Maybe a wrong VAE? Are ComfyUI and all installed nodes up-to-date?
Hey again! You were right! I didn't consider that node since I hadn't made any changes but it turns out that the little bastard was running on bf16 instead of bf32! Had to open the advanced input tab to find it but when I changed it back it worked right away!
Now I just need to figure out the blurring and stuff but this is already looking good. Will post videos and reviews soon, thanks for the help! <3
@Vexy_Kink So are you now running the VAE with fp32? The default is a bf16 file with the bf16 selected, so no changed required in the default. A common mistake is to use a LTX 2 VAE instead the LTX 2.3 one.
@ai839 Nah I'm using only the VAE's, models and other dependencies you outlined in the guide. It was set to BF16 before but when I run it with bf16 the runtime error occurs. I take it that I'm not doing it right based on your reaction. :)
Sorry for the confusion. I am running it with fp32 atm. That's the only way I could get past the error message. Running a RTC 4090 and using checkpoints if that helps.
This wf can do NSFW? Or there is a similar WF meant to NSFW? Thanks for your work
Yes, you can do whatever you want with this workflow. For NSFW you probably will need other LoRAs and/or model like Sulphur or 10eros.
OMG I open your WF and have 26 ERRORS + several conflicts. The manager warns that some nodes cannot be installed because they haven't registered in the Confy UI system. I have the latest version of Confy. What should I do?
You are the first with this problem. I cannot reproduce it. Do you use the ComfyUI Manager or the built in manager? As mentioned in the installation guide I recommend the ComfyUI Manager: https://github.com/Comfy-Org/ComfyUI-Manager?tab=readme-ov-file#comfyui-manager
btw: The ComfyUI Manager also lists "ComfyUI-GGUF-FantasyTalking" as missing node but it is not part of the workflow and can be skipped. If I try to install it nevertheless, it fails for this node.
I use Pixaroma compilation.
I also had a ton of errors. Updating ComfyUI and installing everything mentioned in the workflow fixed everything except for the Film Grain node which is still missing and cannot be found in the build-in ComfyUI manager, leading to the WF not running. I tried installing the one linked here but I have not been successful to make it work, it doesn't show up in my ComfyUI but this is an issue mentioned on the github page.
@makeb38136834 I installed a fresh ComfyUI portable today and the Film Grain node was installed without any problems. It might fail if you have installed other nodes with conflicting dependencies. You are right that there is a discussion about this issue in a pull request: https://github.com/vrgamegirl19/comfyui-vrgamedevgirl/pull/82
@Kagi13 I don't know Pixorama and cannot tell if this can cause problems during the installation of nodes. The installation of the ComfyUI Manager should be possible.
@makeb38136834 I faced the same did quite a number of things:
1. Cloned manually the node, as instructed to custom nodes
https://github.com/vrgamegirl19/comfyui-vrgamedevgirl/tree/main
2. Ran the Python installer of requirements, from there started facing some errors with missing wheels, missing C++ dependencies (google or GPT to troubleshoot), in my case had to install vs_BuildTools to continue resolving the compilation issues.
3. Eventually everything "worked" but the node was still missing in Comfy. Troubleshooting with GPT, helped me discover a dependency "librosa" was not running (despite of it "being installed"). So to install it for Comfy use
"C:\ComfyUI\.venv\Scripts\python.exe -m pip install librosa"
Problem was it installed in the local installation of Python which is different than the version being used by ComfyUI.
This workflow keeps getting better. It's the only one I've used that once I download the used nodes, it works on the first try. The voice cloning works so much better than I anticipated.
The only change I make is swapping the lora loader for the ND-Super-Nodes one, which is way better for filtering and categorizing.
The lora loader sounds interesting. I don't know it yet but will check it.
Couldn't get 3.0 to work. Any ideas? Your ksampler supgraphs are throwing a large host of errors. Updated ComfyUI to latest version 0.21.1. The workflow loads fine without errors, but when actually running the workflow;
\edit\* Nevermind, had to update kjnodes.
I was trying all day to run the wf but comfyui stills saying I´m loosing this node, does anyone know where can I find it or how Can I solve this trouble?
comfyui-aspect-ratio-crop-node (1)
Click the "Update All" button in the ComfyUI Manager. I guess one of the installed nodes is not up-to-date. Ensure ComfyUI is up-to-date, too (at least v0.20.1).
Worked for me doing this:
1. Cloned manually the node, as instructed to custom nodes
https://github.com/vrgamegirl19/comfyui-vrgamedevgirl/tree/main
2. Ran the Python installer of requirements, from there started facing some errors with missing wheels, missing C++ dependencies (google or GPT to troubleshoot), in my case had to install vs_BuildTools to continue resolving the compilation issues.
3. Eventually everything "worked" but the node was still missing in Comfy. Troubleshooting with GPT, helped me discover a dependency "librosa" was not running (despite of it "being installed"). So to install it for Comfy use
"C:\ComfyUI\.venv\Scripts\python.exe -m pip install librosa"
Problem was it installed in the local installation of Python which is different than the version being used by ComfyUI.
@MrTonyB thx this worked for me too
First off, wow. Fast initial video generation and the quality is amazing. However, when I try to add custom 5 sec audio to the generated video for lipsyn, all i get is the orginial created video and the custom audio in the background. What am I missing? any help is appreciated.
In the Troubleshooting Guide on the top right I wrote down some lines regarding missing lip-sync. The prompt or, if it is image-to-video, the first frame (your input image) can be the problem, if the audio does not fit good enough to your scene. So my first question is: Is it text- or image-to-video?
btw: Maybe it is already clear for you: 5 sec of audio is fine, but your are not restricted to 5 seconds. Only for voice cloning 5 seconds are a fixed length for the reference audio (video can still have a different length). Ensure the video and audio length are the same, if you use audio-to-video.
The prompt should mention that the person is saying something otherwise chances are high the input audio will just be background audio.
@ai839 Thank you for your rapid response. It's image to video. I did leave a simple prompt, sublte natural movements then speaks , but it just plays the included audio in the background while my model looks in to the camera.
@franklynsotelo72838 Try another prompt and seed. You could ask an AI like ChatGPT or Perplexity to generate a prompt for you. It can also be helpful to enter the words from the audio clip word for word (as they are spoken, not as they are spelled) into the prompt.
@ai839 great idea. I'll give it a shot!
@franklynsotelo72838 Hi! I had the same issue as you but that was solved when I realized I had missed one setting. Go to the very top node in the section Voice cloning/Audio to video -section and note the description: False = Clone voice, True = Audio to Video. In my case it was set to true by default. But once I set it to false it worked flawlessly. Try that. :D
Oh and always use quotation marks for dialogue like this: "Great workflow, man!" because LTX is always looking for them and will prioritize anything between "'s. If you get weird random dialogue it might be due to an extra " hiding somewhere in the prompt, trust me I've seen that. :P
I'm having an issue where Stage 1 generates successfully, but Stage 2 (video generation) never starts. Instead, Stage 1 keeps restarting and looping forever.
Best wf i use so far. Thanks a lot, you're a hero. I have only one question, if my final video is not quite sharp, then the instructions "raise the sigmas_start value" refer to stage 2, or should I raise the first one too?
Create the video of Stage 1, too. If it's sharp don't make changes there. 0.80 as sigmas_start should be high enough. Increase the steps of Stage 2 to 8 and sigmas_terminal to 0.00 or 0.10. If this does not help, try sigmas_start of 0.40 and sigmas_terminal of 0.20. Is it a text- or image-to-video generation? What is your configured final video size (dimensions)?
Excited to test this workflow but i'm getting an error related to the LTXVLatentUpsampler
AttributeError: 'NoneType' object has no attribute 'per_channel_statistics'
Any help would be great. Thanks!
Possible reasons: You selected a wrong or non existing file or ComfyUI is not up-to-date. Maybe your upscaler file is corrupt, download it again, if you are unsure.
thx!
I'm unable to find the "ComfyUI Aspect Ratio Crop" node in my setup.
Could you please let me know which custom node pack I need to install to get this node?
This was discussed in another comment: https://civitai.red/models/1815300?dialog=commentThread&commentId=1193905
Why not fix the workflow instead of having everyone deal with that issue...
@nCurious The problem isn't with the workflow. Otherwise, more than four out of over 2,300 people would have commented on it. It's end edge case in the local ComfyUI setup.
@ai839 that doesnt make sense - fix the workflow. everybody who will use this workflow in the future will run into this issue...
can't run 2nd stage.
the first stage of workflow completes without an error within 5 minutes with all lora include, and include voice tub enabled. but 2nd stage stays at 0% till more than 6 hours and shows OOM error. i'm having rtx 4070 super 12gb gpu, 32 gb ram.
without voice dub it can generate the video upto 15 sec with some kind of gibirish language, finished 2 stage around 16 minutes.
Ensure no other software blocks VRAM. Close everything you don't need. You can save a little bit of VRAM by deactivating hardware accelleration in the browser. Maybe this is already enough to solve the problem. Otherwise lower the resolution a little bit. Adding a "Clean VRAM Used" node directly before Stage 2 might also help.
having issues with Stage 1 Sampler
AttributeError: 'AudioVAE' object has no attribute 'per_channel_statistics'
unbypassed Load First frame
any suggestions?
figured out the issue and everything is working at the moment. will let you know if i run into other issues!
needs update for 10S nodes, its fixes colors on 2 stage and have nodes for face :3
Anyone found a sollution for "Node 'ID #668:875' has no class_type. The workflow may be corrupted or a custom node is missing." ?
How do I delete my own comment for "Answered already, see the effing guide." reasons?
@ihojyjosdowbacjffb112 how did you solve, I am getting a node error too
@Sherlock669 Honest to god, I have no idea anymore (scatterbrained af). Maybe I will see it today at some point. But looking at my own reaction to my own question... I think there is something in the readme.md or requirements.txt or the description of this project OR (probably the banger) inside the workflow file (the text notes that show up after importing the workflow, look around, there are text notes somewhere inside the workflow).
Hello. Just want to thank you, and say that "LTX 2.3 All-in-One 3.0" workflow is, by my opinion, the best LTX workflow I've seen, and I've seen them all, probably, diving deep into LTX for the past couple of months after Grok died. It simply produces better quality videos than even RuneXX workflows that are gold standard today, and it is MUCH more stable and mature than DaSiwa workflows (second best in video quality). Thank you :) I've just finished my first serious project in LTX, complex music video, and it looks fantastic thanks to this workflow.
Thank you very much for your kind words ♥️.
Getting this error:
TypeError: AudioVAE.__init__() takes 2 positional arguments but 3 were given
I suppose it'd be nice to fix it, but I also would love to just disable all of the audio stuff entirely if possible!
Nevermind - had to delete then reinstall KJNodes via git rather than the ComfyUI Manager.
Hi! Thanks for the amazing workflow.
I have a question regarding the MXFP8 model and LoRAs. I'm running this on an RTX 4090. When I try to use the recommended ltx-2.3-22b-distilled-1.1_transformer_only_mxfp8_block32.safetensors model together with the optional style LoRAs enabled, the generation fails and I get this exact error:
BufferError: float8 types are not supported by dlpack
If I bypass/disable all the LoRAs, the MXFP8 model generates the video perfectly and really fast. Do you know why this float8/dlpack conflict happens when LoRAs are introduced, and is there any known fix or workaround to use LoRAs with this specific FP8 model on an RTX 40xx?
Thanks in advance for your help!
mxfp8 only works with blackwell GPUs (RTX 50XX) and not with ada GPUs (RTX 40XX). My info in the workflow is wrong. I don't know why it works without the LoRAs. My recommendation is to use a fp8 model.
When I run I2V, background music and subtitles keep being generated no matter what. Even after adding strong negative instructions to the positive prompt like: "No subtitles, no captions, no text, no letters, no words, no logos, no watermark, no UI, no music, no singing, no narration, no voice, silent scene" they still keep appearing.
I think I need to connect the LTX2.3 NAG node, but I have no idea how to wire it properly. Can anyone show me the correct way to connect it or tell me what else I should do to completely remove music and subtitles?
Any help or workflow example would be really appreciated. Thanks!
is "last frame" saving supported?
Not by default. Feel free to add a save node for that.
comfyui-aspect-ratio-crop-node
does not exist.
For everyone wondering:
Double Click the "Stage 2 Decoding" Node to open it.
In there you will find the "Film Grain" Node that does not exist anymore.
Simply delete the node, add simple "Image Grain Add" Node, set weight to 0 and attach input and output. Then the Workflow works again.
@chr_isso is it worth trying this one?
I just tried it and first 2 tries goes with comfyui crash. and third try my win 10 pc blue screen of death it said memory management system. be careful guys. it seems like it gives me that error from pytorch when i activate 2. segment of the video
The same thing happened to me when I started Stage 2.
A workflow that runs smoothly without any headaches caused by missing or broken nods. Thank you.
The was-ns nodes are missing. I have the nove manager installed but they are nust not showing up in the list. Can you share a link to where I can manually downlaod if you have one?
"WAS Node Suite (Revised)" is the name. Ensure your ComfyUI Manager registry is up-to-date. The repo url is: https://github.com/ltdrdata/was-node-suite-comfyui
@ai839 Thank you.
i am getting error with defealt settings that it came configured with i just activated sage attention patch on auto thats it
[INFO] got prompt
[WARNING] WARNING: PlaySound.IS_CHANGED() missing 1 required positional argument: 'self'
[WARNING] WARNING: PlaySound.IS_CHANGED() missing 1 required positional argument: 'self'
[INFO] VAE load device: cuda:0, offload device: cpu, dtype: torch.float32
[INFO] VAE load device: cuda:0, offload device: cpu, dtype: torch.bfloat16
[INFO] VAE load device: cuda:0, offload device: cpu, dtype: torch.bfloat16
[INFO] Found quantization metadata version 1
[INFO] Using MixedPrecisionOps for text encoder
[INFO] Interrupting prompt 19eb192a-44bd-44b8-b83b-30a521fbaf8f
[INFO] Requested to load LTXAVTEModel_
[INFO] loaded completely; 11201.91 MB loaded, full load: True
[INFO] CLIP/text encoder model load device: cuda:0, offload device: cpu, current: cuda:0, dtype: torch.float16
[INFO] Processing interrupted
[INFO] Prompt executed in 27.29 seconds
[INFO] got prompt
[WARNING] WARNING: PlaySound.IS_CHANGED() missing 1 required positional argument: 'self'
[WARNING] WARNING: PlaySound.IS_CHANGED() missing 1 required positional argument: 'self'
[INFO] Found quantization metadata version 1
[INFO] Detected mixed precision quantization
[INFO] Using mixed precision operations
[INFO] Native ops: , emulated ops: float8_e5m2, mxfp8, float8_e4m3fn, nvfp4
[INFO] model weight dtype torch.bfloat16, manual cast: torch.bfloat16
[INFO] model_type FLUX
[rgthree-comfy][Power Lora Loader] Lora "AmateurHour_01_rank16.safetensors" not found, skipping.
[rgthree-comfy][Power Lora Loader] Lora "LTX2.3_Soft_Enhance.safetensors" not found, skipping.
[rgthree-comfy][Power Lora Loader] Lora "LTX23-GalaxyAce.safetensors" not found, skipping.
[rgthree-comfy][Power Lora Loader] Lora "LTX2.3_Crisp_Enhance.safetensors" not found, skipping.
[rgthree-comfy][Power Lora Loader] Lora "Luxe_Sensual.safetensors" not found, skipping.
[INFO] Using sage attention mode: auto
[INFO] Interrupting prompt d362dcb6-0d76-445f-a594-69d147d9dabe
[INFO] Processing interrupted
[INFO] Prompt executed in 6.75 seconds
[INFO] got prompt
[WARNING] WARNING: PlaySound.IS_CHANGED() missing 1 required positional argument: 'self'
[WARNING] WARNING: PlaySound.IS_CHANGED() missing 1 required positional argument: 'self'
[INFO] Using sage attention mode: auto
[INFO] Requested to load VideoVAE
[INFO] loaded completely; 10357.54 MB usable, 1384.94 MB loaded, full load: True
[INFO] Requested to load LTXAV
[INFO] loaded partially; 16681.53 MB usable, 16598.22 MB loaded, 6319.52 MB offloaded, 112.54 MB buffer reserved, lowvram patches: 0
100%|████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████| 4/4 [07:59<00:00, 119.93s/it]
[INFO] Requested to load AudioVAE
[INFO] loaded completely; 693.46 MB loaded, full load: True
[INFO] Requested to load VideoVAE
[INFO] 0 models unloaded.
[INFO] loaded partially; 0.00 MB usable, 0.00 MB loaded, 1384.94 MB offloaded, 378.02 MB buffer reserved, lowvram patches: 0
[INFO] loaded completely; 20330.85 MB usable, 1384.94 MB loaded, full load: True
[INFO] VAE load device: cuda:0, offload device: cpu, dtype: torch.bfloat16
[INFO] VAE load device: cuda:0, offload device: cpu, dtype: torch.bfloat16
[INFO] Found quantization metadata version 1
[INFO] Using MixedPrecisionOps for text encoder
[INFO] Requested to load LTXAVTEModel_
[INFO] loaded completely; 11201.91 MB loaded, full load: True
[INFO] CLIP/text encoder model load device: cuda:0, offload device: cpu, current: cuda:0, dtype: torch.float16
[ERROR] !!! Exception during processing !!! The paging file is too small for this operation to complete. (os error 1455)
[ERROR] Traceback (most recent call last):
File "C:\Comfy\ComfyUI-Easy-Install\ComfyUI\execution.py", line 542, 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:\Comfy\ComfyUI-Easy-Install\ComfyUI\execution.py", line 341, 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:\Comfy\ComfyUI-Easy-Install\ComfyUI\custom_nodes\comfyui-lora-manager\py\metadata_collector\metadata_hook.py", line 171, in async_map_node_over_list_with_metadata
results = await original_map_node_over_list(
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "C:\Comfy\ComfyUI-Easy-Install\ComfyUI\execution.py", line 315, in asyncmap_node_over_list
await process_inputs(input_dict, i)
File "C:\Comfy\ComfyUI-Easy-Install\ComfyUI\execution.py", line 303, in process_inputs
result = f(**inputs)
^^^^^^^^^^^
File "C:\Comfy\ComfyUI-Easy-Install\ComfyUI\nodes.py", line 965, in load_unet
model = comfy.sd.load_diffusion_model(unet_path, model_options=model_options)
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "C:\Comfy\ComfyUI-Easy-Install\ComfyUI\comfy\sd.py", line 2054, in load_diffusion_model
sd, metadata = comfy.utils.load_torch_file(unet_path, return_metadata=True)
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "C:\Comfy\ComfyUI-Easy-Install\ComfyUI\comfy\utils.py", line 149, in load_torch_file
raise e
File "C:\Comfy\ComfyUI-Easy-Install\ComfyUI\comfy\utils.py", line 133, in load_torch_file
with safetensors.safe_open(ckpt, framework="pt", device=device.type) as f:
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
OSError: The paging file is too small for this operation to complete. (os error 1455)
The most likely cause is insufficient Windows virtual memory: the page file is too small or already full when ComfyUI loads the model. This usually happens when RAM is exhausted and Windows cannot allocate more swap space.
Possible fixes:
- Increase the Windows page file size.
- Set virtual memory to automatic management.
- Free disk space on the system drive.
- Close other memory-heavy programs.
- Use a smaller model.
- Add more physical RAM if this happens often.
Workflow runs pretty good. Thanks for the work. The one thing I am trying to figure out is why my 720x1280 portrait image keeps zooming into a close up of the face. any idea?
Be more specific with your prompt. Describe the last frame.
Can you know how to remove this zoom-in effect that keeps happening in an image-to-video? The music is awful sometimes. Can it not be added or turned on and off with a switch?
For enable frist frame, when the seconds are like over 30, like for talking videos from frist frame then its very bad. if i do for less than 20 seconds its great, but when put 40 - 45 seconds. it turns out very bad quality always.
can you guide what can be the issue or what setting i might have to do ?
These are the limitations of LTX 2.3 ot at least of my workflow but maybe there are other models, worklows or something else to achive this with a single generation. I don't have experience with that so I cannot help, sorry.
@ai839 how long of a single video generation you have tried?
@abhituls55982 over 20 secs causes problems
Just downloaded the workflow and the models, but i'm getting an error:
module 'ast' has no attribute 'Num'
Stage 1 Video Width
i have no idea what it is
Do you have version with 10SNodes integrated?