[model] LTX Video 0.9.8

[a-r-r-o-w]

code

import torch
from diffusers import LTXConditionPipeline, LTXLatentUpsamplePipeline
from diffusers.pipelines.ltx.pipeline_ltx_condition import LTXVideoCondition
from diffusers.pipelines.ltx.modeling_latent_upsampler import LTXLatentUpsamplerModel
from diffusers.utils import export_to_video, load_video

pipeline = LTXConditionPipeline.from_pretrained("Lightricks/LTX-Video-0.9.8-13B-distilled", torch_dtype=torch.bfloat16)
upsampler = LTXLatentUpsamplerModel.from_pretrained("a-r-r-o-w/LTX-0.9.8-Latent-Upsampler", torch_dtype=torch.bfloat16)
pipe_upsample = LTXLatentUpsamplePipeline(vae=pipeline.vae, latent_upsampler=upsampler).to(torch.bfloat16)
pipeline.to("cuda")
pipe_upsample.to("cuda")
pipeline.vae.enable_tiling()

def round_to_nearest_resolution_acceptable_by_vae(height, width):
    height = height - (height % pipeline.vae_spatial_compression_ratio)
    width = width - (width % pipeline.vae_spatial_compression_ratio)
    return height, width

prompt = """The camera pans over a snow-covered mountain range, revealing a vast expanse of snow-capped peaks and valleys.The mountains are covered in a thick layer of snow, with some areas appearing almost white while others have a slightly darker, almost grayish hue. The peaks are jagged and irregular, with some rising sharply into the sky while others are more rounded. The valleys are deep and narrow, with steep slopes that are also covered in snow. The trees in the foreground are mostly bare, with only a few leaves remaining on their branches. The sky is overcast, with thick clouds obscuring the sun. The overall impression is one of peace and tranquility, with the snow-covered mountains standing as a testament to the power and beauty of nature."""
# prompt = """A woman walks away from a white Jeep parked on a city street at night, then ascends a staircase and knocks on a door. The woman, wearing a dark jacket and jeans, walks away from the Jeep parked on the left side of the street, her back to the camera; she walks at a steady pace, her arms swinging slightly by her sides; the street is dimly lit, with streetlights casting pools of light on the wet pavement; a man in a dark jacket and jeans walks past the Jeep in the opposite direction; the camera follows the woman from behind as she walks up a set of stairs towards a building with a green door; she reaches the top of the stairs and turns left, continuing to walk towards the building; she reaches the door and knocks on it with her right hand; the camera remains stationary, focused on the doorway; the scene is captured in real-life footage."""
negative_prompt = "bright colors, symbols, graffiti, watermarks, worst quality, inconsistent motion, blurry, jittery, distorted"
expected_height, expected_width = 480, 832
downscale_factor = 2 / 3
# num_frames = 161
num_frames = 361

# 1. Generate video at smaller resolution
downscaled_height, downscaled_width = int(expected_height * downscale_factor), int(expected_width * downscale_factor)
downscaled_height, downscaled_width = round_to_nearest_resolution_acceptable_by_vae(downscaled_height, downscaled_width)
latents = pipeline(
    prompt=prompt,
    negative_prompt=negative_prompt,
    width=downscaled_width,
    height=downscaled_height,
    num_frames=num_frames,
    timesteps=[1000, 993, 987, 981, 975, 909, 725, 0.03],
    decode_timestep=0.05,
    decode_noise_scale=0.025,
    image_cond_noise_scale=0.0,
    guidance_scale=1.0,
    guidance_rescale=0.7,
    generator=torch.Generator().manual_seed(0),
    output_type="latent",
).frames

# 2. Upscale generated video using latent upsampler with fewer inference steps
# The available latent upsampler upscales the height/width by 2x
upscaled_height, upscaled_width = downscaled_height * 2, downscaled_width * 2
upscaled_latents = pipe_upsample(
    latents=latents,
    adain_factor=1.0,
    tone_map_compression_ratio=0.6,
    output_type="latent"
).frames

# 3. Denoise the upscaled video with few steps to improve texture (optional, but recommended)
video = pipeline(
    prompt=prompt,
    negative_prompt=negative_prompt,
    width=upscaled_width,
    height=upscaled_height,
    num_frames=num_frames,
    denoise_strength=0.999,  # Effectively, 4 inference steps out of 5
    timesteps=[1000, 909, 725, 421, 0],
    latents=upscaled_latents,
    decode_timestep=0.05,
    decode_noise_scale=0.025,
    image_cond_noise_scale=0.0,
    guidance_scale=1.0,
    guidance_rescale=0.7,
    generator=torch.Generator().manual_seed(0),
    output_type="pil",
).frames[0]

# 4. Downscale the video to the expected resolution
video = [frame.resize((expected_width, expected_height)) for frame in video]

export_to_video(video, "output.mp4", fps=24)

moutain.mp4

woman-door.mp4

[HuggingFaceDocBuilderDev]

The docs for this PR live here. All of your documentation changes will be reflected on that endpoint. The docs are available until 30 days after the last update.

[yiyixuxu]

thanks!

[a-r-r-o-w]

@yiyixuxu I forgot to clarify here. This PR is slightly incomplete -- we still need the autoregressive sampling I believe. I wrote more about it here: https://huggingface.slack.com/archives/C065E480NN9/p1754568827299629?thread_ts=1752682810.233009&cid=C065E480NN9

Let's not merge for now :) I was debugging a different PR of mine so still need to look for a bit at their comfy codebase

[vladmandic]

just checking on what's the status of this pr?

[a-r-r-o-w]

The infinite generation does not work yet. I couldn't get the outputs to match with the ComfyUI code after a few days of debugging, but did make significant progress. However, I no longer work at HF and it's hard for me to find the time to contribute this atm.

Anyone interested can feel free to take over! It mostly works but there's temporal jitter at every context boundary - which is the only thing that remains to be debugged I think

[vladmandic]

imo, the value of ltxvideo-0.9.8 exists even without infinite generation.
infinite generation is a net-new feature, but regardless of it ltxvideo-0.9.8 is higher than previous versions, so i'd suggest to include basic support and then follow-up on jitter that you've mentioned separately.

[a-r-r-o-w]

Since I don't have access to edit this branch, @yiyixuxu if you revert to this commit, LTX 098 should be good to merge without the infinite generation capability: 47d93ce

[yiyixuxu]

thanks @a-r-r-o-w!
I merged it here #12493

will close this PR for now :)