Implementation:Zai org CogVideo Export To Video

Overview

Concrete tool for saving generated video frames as an MP4 file provided by the diffusers utility library. This is the final step in the text-to-video inference workflow, converting in-memory PIL Image frames into a playable video file.

Source

inference/cli_demo.py:L195

Signature

export_to_video(
    video_frames: List[PIL.Image.Image],
    output_video_path: str = "./output.mp4",
    fps: int = 16
) -> str

Key Parameters

Parameter	Type	Default	Description
video_frames	List[PIL.Image.Image]	(required)	List of PIL Image frames from pipeline output (`output.frames[0]`)
output_video_path	str	"./output.mp4"	File path for the output MP4 video file
fps	int	16	Frames per second for video playback. Use 16 for Diffusers pipeline, 8 for SAT pipeline.

Inputs

List of PIL Image frames -- The video frames generated by the CogVideoX pipeline, accessed via output.frames[0]. Each frame is a PIL.Image.Image object.

Outputs

.mp4 video file -- An H.264-encoded MP4 video file saved to disk at the specified output path. Returns the output path as a string.

Usage Example

import torch
from diffusers import CogVideoXPipeline, CogVideoXDPMScheduler
from diffusers.utils import export_to_video

# Assume pipeline is loaded, configured, and memory-optimized
pipe = CogVideoXPipeline.from_pretrained(
    "THUDM/CogVideoX1.5-5B",
    torch_dtype=torch.bfloat16
)
pipe.scheduler = CogVideoXDPMScheduler.from_config(
    pipe.scheduler.config,
    timestep_spacing="trailing"
)
pipe.enable_sequential_cpu_offload()
pipe.vae.enable_slicing()
pipe.vae.enable_tiling()

# Generate video
output = pipe(
    prompt="A detailed wooden toy ship gliding over a calm ocean.",
    height=768,
    width=1360,
    num_frames=81,
    num_inference_steps=50,
    guidance_scale=6.0,
    use_dynamic_cfg=True,
    generator=torch.Generator().manual_seed(42),
)

# Export to video file
export_to_video(
    video_frames=output.frames[0],
    output_video_path="./output.mp4",
    fps=16
)

Import

from diffusers.utils import export_to_video

Related Pages

Page Connections

Double-click a node to navigate. Hold to expand connections.

Principle

Implementation

Heuristic

Environment