Implementation:Datajuicer Data juicer VideoSplitByDurationMapper

Knowledge Sources	Datajuicer_Data_juicer
Domains	Data_Processing, Mapping
Last Updated	2026-02-14 16:00 GMT

Overview

Concrete tool for splitting videos into fixed-duration segments provided by Data-Juicer.

Description

VideoSplitByDurationMapper is a mapper operator that splits videos into fixed-duration segments, creating multiple shorter clips from each input video to produce uniform-length training samples. It divides each video into segments of a configurable duration (default: 10 seconds), discards the last segment if shorter than a minimum threshold, saves split clips using either FFmpeg or PyAV backends, and updates the sample's video references and text placeholders to reflect the new segments, with optional original sample preservation.

Usage

Use when you need to create uniform-length video clips required by many video generation and understanding models that expect fixed-duration inputs.

Code Reference

Source Location

• Repository: Datajuicer_Data_juicer
• File: data_juicer/ops/mapper/video_split_by_duration_mapper.py

Signature

@OPERATORS.register_module("video_split_by_duration_mapper")
class VideoSplitByDurationMapper(Mapper):
    def __init__(self, split_duration: float = 10, min_last_split_duration: float = 0, keep_original_sample: bool = True, save_dir: str = None, video_backend: str = "ffmpeg", ffmpeg_extra_args: str = "", *args, **kwargs):

Import

from data_juicer.ops.mapper.video_split_by_duration_mapper import VideoSplitByDurationMapper

I/O Contract

Inputs

Name	Type	Required	Description
split_duration	float	No	Duration of each video split in seconds (default: 10)
min_last_split_duration	float	No	Minimum allowable duration for the last split; shorter splits are discarded (default: 0)
keep_original_sample	bool	No	Whether to keep the original sample (default: True)
save_dir	str	No	Directory for generated video files; if not specified, saves alongside input files
video_backend	str	No	Video backend: "ffmpeg" or "av" (default: "ffmpeg")
ffmpeg_extra_args	str	No	Extra FFmpeg args for splitting video (default: "")

Outputs

Name	Type	Description
samples	Dict	Transformed samples with split video segment file paths

Usage Examples

process:
  - video_split_by_duration_mapper:
      split_duration: 10
      min_last_split_duration: 3
      keep_original_sample: false

Related Pages

• Environment:Datajuicer_Data_juicer_Python_Runtime_Environment