Weights are byte-identical to the upstream release; only the layout has been rearranged so LTX2LatentUpsamplerModel.from_pretrained and LTX2LatentUpsamplePipeline can consume it directly.

Usage

python

import torch
from diffusers.pipelines.ltx2 import LTX2LatentUpsamplerModel
from diffusers.pipelines.ltx2.pipeline_ltx2_latent_upsample import LTX2LatentUpsamplePipeline

latent_upsampler = LTX2LatentUpsamplerModel.from_pretrained(
    '<your-username>/<your-repo>',
    subfolder='latent_upsampler',
    torch_dtype=torch.bfloat16,
)
upsample_pipe = LTX2LatentUpsamplePipeline(vae=your_pipe.vae, latent_upsampler=latent_upsampler)

License

Apache 2.0, inherited from the upstream Lightricks/LTX-2.3 release.

CalamitousFelicitousness/LTX-2.3-Spatial-Upsampler-x2-1.0-Diffusers

LTX-2.3 Spatial Upscaler x2 1.0 (Diffusers)

Usage

License

CalamitousFelicitousness/
LTX-2.3-Spatial-Upsampler-x2-1.0-Diffusers