LTX-2.3 THE BEST VIDEO AI KING WITH VIDEO EXTEND & MULTI-FRAME!
Added 2026-03-21 17:54:24 +0000 UTC
Comments
finally got it working by just installing it from scratch on it's own portable. Word of advice, don't even bother trying to install on an existing comfy install. not worth the headache.
brian
2026-03-23 01:33:26 +0000 UTCI've been getting the following error when I run it: Node 'ID #1537:1530' has no class_type. The workflow may be corrupted or a custom node is missing.
Instruments of Destruction
2026-03-22 23:54:50 +0000 UTCAmazing videos, but question. The more I use this the slower it gets. Now it's stuck using image to video after like 20 videos. Is there a cache to clear or something?
Pizza
2026-03-22 23:26:24 +0000 UTCThe error is coming from VAELoaderKJ on the audio VAE. Try updating KJNodes or delete the folder from the ComfyUI_windows_portable\ComfyUI\custom_nodes folder, then use the LTX-2-3-MODELS-NODES_INSTALL.bat put that installer into the ComfyUI_windows_portable\ComfyUI folder and run it, it will reinstall all the missing node
Aitrepreneur
2026-03-22 21:36:14 +0000 UTCyes, go to civitai and filter for lora for ltx 2.0 and 2.3, 2.0 loras work for 2.3 as well
Aitrepreneur
2026-03-22 21:34:52 +0000 UTCyou can dm it to me
Aitrepreneur
2026-03-22 21:33:58 +0000 UTCNo need for that, I've changed the way the previous workflow works, if you want better quality, you can just change the sampler from euler to something like euler_ancestral_cfg_pp but it will take longer to generate
Aitrepreneur
2026-03-22 21:21:22 +0000 UTCYeah it's fine, as I said it's just a bit of banter, just a random example I found of a low quality video with bad lighting and a character that ltx doesn't know so that the voice cloning would work fine
Aitrepreneur
2026-03-22 21:19:27 +0000 UTCYou can actually, there is a special node to do that, I wanted to implement it but after multiple tests, there are too many flickers and bad transitions between frames, 3 frames is often the limit for correct frames transition, at least for now
Aitrepreneur
2026-03-22 21:18:03 +0000 UTCcouple issues with this. 1st the json file uploaded if tried to use straight out of the download doesn't like the 2.3 VAE video and audio files provided. it give me this error: "Error(s) in loading state_dict for VideoVAE: size mismatch for decoder.conv_in.conv.weight: copying a param with shape torch.Size([1024, 128, 3, 3, 3]) from checkpoint, the shape in current model is torch.Size([256, 128, 3, 3, 3]). size mismatch for decoder.conv_in.conv.bias: copying a param with shape torch.Size([1024]) from checkpoint, the shape in current model is torch.Size([256])." the error goes away after i switch it to the "LTX2_video_vae_bf16" and "LTX2_audio_vae_bf16" but then it gives the error: 'Linear' object has no attribute 'weight' this is on the comfy installer he made for LTX2 3 months ago. EDIT: updated comfyui and the VAE are passing at least initially now but the Checkpoint "ltx-2.3-22b-dev-Q8_0.gguf" doesn't work and gives me RuntimeError: Error(s) in loading state_dict for LTXAVModel: size mismatch for audio_embeddings_connector.learnable_registers: copying a param with shape torch.Size([128, 2048]) from checkpoint, the shape in current model is torch.Size([128, 3840]). size mismatch for audio_embeddings_connector.transformer_1d_blocks.0.attn1.q_norm.weight: copying a param with shape torch.Size([2048]) from checkpoint, the shape in current model is torch.Size([3840]). size mismatch for audio_embeddings_connector.transformer_1d_blocks.0.attn1.k_norm.weight: copying a param with shape torch.Size([2048]) from checkpoint, the shape in current model is torch.Size([3840]).
brian
2026-03-22 20:30:50 +0000 UTCRunpod constantly times out and is always trying to reconnect, until it just dies. Anyone else have that issue?
Josh Atencio
2026-03-22 19:55:49 +0000 UTCthnx for this. Is it NSFW? I tried oe prompts with no success. thnx in advance
Akind
2026-03-22 19:23:11 +0000 UTCRuns fine starts generating then I am getting the following error "ValueError: not enough values to unpack (expected 4, got 3)" I have the log if anyone wants to see it.
BadgerWi Johnson
2026-03-22 16:13:38 +0000 UTCYop, how to add more steps for the complex scene ?
Benjamin GULMEZ
2026-03-22 15:24:05 +0000 UTCHey ! Thank your for the workflow and the video! The example with Asmongold is .... waou... but be careful, his community could shit storm you for this :s I'm a bit worried....
RavenWT
2026-03-22 14:03:29 +0000 UTCThat would be nice to have multi frame generation feature, like eg. 10 frames and can generate >1 minute long video.
PJ
2026-03-22 12:06:28 +0000 UTCMake sure to enable "No Custom Audio IMG2VID" if you don't have an audio file uploaded into the Load Audio node. That fixed it for me.
Skippy
2026-03-22 02:08:29 +0000 UTCIs there any plans to add sage attention to these?
Manic
2026-03-22 00:51:28 +0000 UTCI get errors like: ...\comfy_extras\nodes_lt.py", line 671, in execute latents = av_latent["samples"].unbind() ~~~~~~~~~^^^^^^^^^^^ TypeError: 'NoneType' object is not subscriptable Prompt executed in 1.07 seconds I did "update all"/ the .bat update. Looking at the comfyui nodes in the workflow, quite a few have conflicts.
cool1
2026-03-21 23:12:50 +0000 UTC
