Implementation:NVIDIA NeMo Curator ClipWriterStage

Knowledge Sources	NeMo Curator
Domains	Data_Curation, Video_Processing, Data_Engineering
Last Updated	2026-02-14 17:00 GMT

Overview

Concrete tool for writing video clips, metadata, embeddings, and previews to structured storage provided by NeMo Curator.

Description

The ClipWriterStage writes all artifacts from a fully-processed VideoTask to organized storage directories. It supports parallel writing via thread pool, configurable output selection (clips, embeddings, previews, captions), and deterministic naming via content hashing.

Usage

Import this stage as the final step in a video curation pipeline.

Code Reference

Source Location

Repository: NeMo Curator
File: nemo_curator/stages/video/io/clip_writer.py
Lines: L34-429

Signature

@dataclass
class ClipWriterStage(ProcessingStage[VideoTask, VideoTask]):
    output_path: str
    input_path: str
    upload_clips: bool
    dry_run: bool
    generate_embeddings: bool
    generate_previews: bool
    generate_captions: bool
    embedding_algorithm: str = "cosmos-embed1"
    caption_models: list[str] | None = None
    enhanced_caption_models: list[str] | None = None
    verbose: bool = False
    max_workers: int = 6
    log_stats: bool = False
    name: str = "clip_writer"

Import

from nemo_curator.stages.video.io.clip_writer import ClipWriterStage

I/O Contract

Inputs

Name	Type	Required	Description
task	VideoTask	Yes	Fully processed video with clips, embeddings, captions

Outputs

Name	Type	Description
task	VideoTask	Same task after writing artifacts to storage

Usage Examples

from nemo_curator.stages.video.io.clip_writer import ClipWriterStage

writer = ClipWriterStage(
    output_path="./output/curated_videos",
    input_path="./data/raw_videos",
    upload_clips=True,
    dry_run=False,
    generate_embeddings=True,
    generate_previews=True,
    generate_captions=True,
    max_workers=6,
)

Related Pages

Implements Principle

Page Connections

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

Principle

Implementation

Heuristic

Environment