LTX Audio VAE is not found.

[3dcinetv]

hi. I'm using the Comfy UI template to load the image to video workflow, I downloaded all the required nodes, but when I press Run, I get this error:
LTXVAudioVAELoader
Metadata is required for audio VAE

Please help. Where do I get (connect/download ) this from?

[Sean-Bradley]

ComfyUI release 0.8.0 is out now. Upgrade to that since these new LTX 2 nodes use code made available in release 0.8.0

[3dcinetv]

Ok, updated. Thanks!
Also: I tried to downsample a 360x240 , 15 fps, video with 60 frames (4 seconds), and it takes more than 1 hour in my RTX 3060 with 12 GB Vram. I don't undestand what else do I need to downsample to make it faster. The LTXVae tile was reduced from 4096 to 512 and other parameters cut in half. But still the GPU takes forever to create a 6 second talking video in SD resolution.
What other things do I need to check?
I'm using ltx-2-19b-dev-fp4.safetensors

[michaellightricks]

try to use distilled model and one of the fp8 comfy flags

[michaellightricks]

make sure you use flow suitable for distilled model

[pgiblock]

Different user, same error here.

I am using the template exactly as it comes shipped - and downloaded the models it reports missing via the links presented in the missing models dialog. It not longer complains about missing models. All the nodes in the main and subgraph appear to be satisfied. But I get the error "Metadata is required for audio VAE" upon run.

• ComfyUI Version: 0.8.0
• Arguments: /home/comfyui/ComfyUI/main.py --enable-manager --disable-api-nodes --cuda-device 1 --port 8189
• OS: linux
• Python Version: 3.12.11 (main, Sep 12 2025, 21:12:46) [GCC 12.2.0]
• Embedded Python: false
• PyTorch Version: 2.8.0+cu129