LTX-Video

LTX-Video é um diffusion transformer projetado para a geração rápida e em tempo real de vídeos em alta resolução a partir de texto e imagens. O principal recurso do LTX-Video é o Video-VAE. O Video-VAE possui uma taxa de compressão de pixel para latente mais alta (1:192), o que permite um processamento de dados de vídeo mais eficiente e uma velocidade de geração mais rápida. Para suportar e evitar que detalhes mais finos sejam perdidos durante a geração, o decodificador do Video-VAE executa a conversão de latente para pixel e a última etapa de desruido.

Você pode encontrar todos os checkpoints originais do LTX-Video sob a organização Lightricks.

Clique nos modelos LTX-Video na barra lateral à direita para ver mais exemplos de outras tarefas de geração de vídeo.

O exemplo abaixo demonstra como gerar um vídeo otimizado para memória ou velocidade de inferência.

memória

velocidade de inferência

Observações

Consulte as configurações recomendadas a seguir para geração a partir do repositório LTX-Video.
- O dtype recomendado para o transformer, o VAE e o text encoder é torch.bfloat16. O VAE e o text encoder também podem ser torch.float32 ou torch.float16.
- Para as variantes destiladas por orientação do LTX-Video, defina guidance_scale como 1.0. O guidance_scale para qualquer outro modelo deve ser definido com um valor maior, como 5.0, para uma boa qualidade de geração.
- Para as variantes de VAE com consciência de timestep (LTX-Video 0.9.1 e superiores), defina decode_timestep como 0.05 e image_cond_noise_scale como 0.025.
- Para as variantes que suportam interpolação entre múltiplas imagens e vídeos de condicionamento (LTX-Video 0.9.5 e superiores), utilize imagens e vídeos semelhantes para obter os melhores resultados. A divergência em relação às entradas de condicionamento pode levar a transições abruptas no vídeo gerado.

O LTX-Video 0.9.7 inclui um upscaler espacial de latentes e um transformer com 13B de parâmetros. Durante a inferência, um vídeo de baixa resolução é gerado rapidamente primeiro e, em seguida, é ampliado e refinado.

Mostrar código de exemplo

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

pipeline = LTXConditionPipeline.from_pretrained("Lightricks/LTX-Video-0.9.7-dev", dtype=torch.bfloat16)
pipeline_upsample = LTXLatentUpsamplePipeline.from_pretrained("Lightricks/ltxv-spatial-upscaler-0.9.7", vae=pipeline.vae, dtype=torch.bfloat16)
pipeline.to("cuda")  # or "mps", "xpu", "cpu"
pipe_upsample.to("cuda")
pipeline.vae.enable_tiling()

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

video = load_video(
    "https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/diffusers/cosmos/cosmos-video2world-input-vid.mp4"
)[:21]  # only use the first 21 frames as conditioning
condition1 = LTXVideoCondition(video=video, frame_index=0)

prompt = """
The video depicts a winding mountain road covered in snow, with a single vehicle 
traveling along it. The road is flanked by steep, rocky cliffs and sparse vegetation. 
The landscape is characterized by rugged terrain and a river visible in the distance. 
The scene captures the solitude and beauty of a winter drive through a mountainous region.
"""
negative_prompt = "worst quality, inconsistent motion, blurry, jittery, distorted"
expected_height, expected_width = 768, 1152
downscale_factor = 2 / 3
num_frames = 161

# 1. Generate video at smaller resolution
# Text-only conditioning is also supported without the need to pass `conditions`
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(
    conditions=[condition1],
    prompt=prompt,
    negative_prompt=negative_prompt,
    width=downscaled_width,
    height=downscaled_height,
    num_frames=num_frames,
    num_inference_steps=30,
    decode_timestep=0.05,
    decode_noise_scale=0.025,
    image_cond_noise_scale=0.0,
    guidance_scale=5.0,
    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,
    output_type="latent"
).frames

# 3. Denoise the upscaled video with few steps to improve texture (optional, but recommended)
video = pipeline(
    conditions=[condition1],
    prompt=prompt,
    negative_prompt=negative_prompt,
    width=upscaled_width,
    height=upscaled_height,
    num_frames=num_frames,
    denoise_strength=0.4,  # Effectively, 4 inference steps out of 10
    num_inference_steps=10,
    latents=upscaled_latents,
    decode_timestep=0.05,
    decode_noise_scale=0.025,
    image_cond_noise_scale=0.0,
    guidance_scale=5.0,
    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)

O modelo destilado LTX-Video 0.9.7 é destilado em relação ao guidance e aos timesteps para acelerar a geração. É necessário definir guidance_scale como 1.0 e num_inference_steps deve estar entre 4 e 10 para uma boa qualidade de geração. Você também deve usar os seguintes timesteps personalizados para obter os melhores resultados.

Inferência do modelo base para preparar o upscaling: [1000, 993, 987, 981, 975, 909, 725, 0.03].
Upscaling: [1000, 909, 725, 421, 0].

Mostrar código de exemplo

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

pipeline = LTXConditionPipeline.from_pretrained("Lightricks/LTX-Video-0.9.7-distilled", dtype=torch.bfloat16)
pipe_upsample = LTXLatentUpsamplePipeline.from_pretrained("Lightricks/ltxv-spatial-upscaler-0.9.7", vae=pipeline.vae, dtype=torch.bfloat16)
pipeline.to("cuda")  # or "mps", "xpu", "cpu"
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 = """
artistic anatomical 3d render, utlra quality, human half full male body with transparent 
skin revealing structure instead of organs, muscular, intricate creative patterns, 
monochromatic with backlighting, lightning mesh, scientific concept art, blending biology 
with botany, surreal and ethereal quality, unreal engine 5, ray tracing, ultra realistic, 
16K UHD, rich details. camera zooms out in a rotating fashion
"""
negative_prompt = "worst quality, inconsistent motion, blurry, jittery, distorted"
expected_height, expected_width = 768, 1152
downscale_factor = 2 / 3
num_frames = 161

# 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,
    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)

O modelo destilado LTX-Video 0.9.8 é semelhante à variante 0.9.7. Ele é destilado em relação ao guidance e ao timestep, e um código de inferência semelhante ao mostrado acima pode ser utilizado. Uma melhoria desta versão é o suporte à geração de vídeos muito longos. Além disso, ele suporta o uso de tone mapping para melhorar a qualidade do vídeo gerado, por meio do parâmetro tone_map_compression_ratio. O valor padrão de 0.6 é recomendado.

Mostrar código de exemplo

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

pipeline = LTXConditionPipeline.from_pretrained("Lightricks/LTX-Video-0.9.8-13B-distilled", dtype=torch.bfloat16)
# TODO: Update the checkpoint here once updated in LTX org
upsampler = LTXLatentUpsamplerModel.from_pretrained("a-r-r-o-w/LTX-0.9.8-Latent-Upsampler", dtype=torch.bfloat16)
pipe_upsample = LTXLatentUpsamplePipeline(vae=pipeline.vae, latent_upsampler=upsampler).to(torch.bfloat16)
pipeline.to("cuda")  # or "mps", "xpu", "cpu"
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)

O LTX-Video suporta LoRAs com load_lora_weights().

Mostrar código de exemplo

import torch
from diffusers import LTXConditionPipeline
from diffusers.utils import export_to_video, load_image

pipeline = LTXConditionPipeline.from_pretrained(
    "Lightricks/LTX-Video-0.9.5", dtype=torch.bfloat16
)

pipeline.load_lora_weights("Lightricks/LTX-Video-Cakeify-LoRA", adapter_name="cakeify")
pipeline.set_adapters("cakeify")

# use "CAKEIFY" to trigger the LoRA
prompt = "CAKEIFY a person using a knife to cut a cake shaped like a Pikachu plushie"
image = load_image("https://huggingface.co/Lightricks/LTX-Video-Cakeify-LoRA/resolve/main/assets/images/pikachu.png")

video = pipeline(
    prompt=prompt,
    image=image,
    width=576,
    height=576,
    num_frames=161,
    decode_timestep=0.03,
    decode_noise_scale=0.025,
    num_inference_steps=50,
).frames[0]
export_to_video(video, "output.mp4", fps=26)

O LTX-Video suporta o carregamento a partir de arquivos únicos, como checkpoints GGUF, com loaders.FromOriginalModelMixin.from_single_file() ou loaders.FromSingleFileMixin.from_single_file().

Exibir código de exemplo

import torch
from diffusers.utils import export_to_video
from diffusers import LTXPipeline, AutoModel, GGUFQuantizationConfig

transformer = AutoModel.from_single_file(
    "https://huggingface.co/city96/LTX-Video-gguf/blob/main/ltx-video-2b-v0.9-Q3_K_S.gguf"
    quantization_config=GGUFQuantizationConfig(compute_dtype=torch.bfloat16),
    dtype=torch.bfloat16
)
pipeline = LTXPipeline.from_pretrained(
    "Lightricks/LTX-Video",
    transformer=transformer,
    dtype=torch.bfloat16
)

Diffusers

LTX-Video

Observações

LTXI2VLongMultiPromptPipeline

classe diffusers.LTXI2VLongMultiPromptPipeline

__call__

encode_prompt

prepare_latents

vae_decode_tiled

LTXPipeline

classe diffusers.LTXPipeline

__call__

encode_prompt

LTXImageToVideoPipeline

classe diffusers.LTXImageToVideoPipeline

__call__

encode_prompt

LTXConditionPipeline

classe diffusers.LTXConditionPipeline

__call__

add_noise_to_image_conditioning_latents

encode_prompt

trim_conditioning_sequence

LTXLatentUpsamplePipeline

classe diffusers.LTXLatentUpsamplePipeline

__call__

adain_filter_latent

tone_map_latents

LTXPipelineOutput

classe diffusers.pipelines.ltx.pipeline_output.LTXPipelineOutput

call

call

call

call

call