Split data into multiple rows with iterators

This documentation page is also available as an interactive notebook. You can launch the notebook in Kaggle or Colab, or download it for use with an IDE or local Jupyter installation, by clicking one of the above links.

Transform a single document, video, image, or audio file into multiple rows for granular processing. What’s in this recipe:

Split documents into text chunks for RAG
Extract frames or segments from videos
Tile images for high-resolution analysis
Chunk audio files for transcription

Problem

You have documents, videos, or text that you need to break into smaller pieces for processing. A PDF needs to be split into chunks for retrieval-augmented generation. A video needs individual frames for analysis. Text needs to be divided into sentences or sliding windows. You need a way to transform one source row into multiple output rows automatically.

Solution

You create views with iterator functions that split source data into multiple rows. Pixeltable provides built-in iterators for documents, videos, images, audio, and strings.

Setup

%pip install -qU pixeltable spacy tiktoken click
!python -m spacy download en_core_web_sm -q

Split documents into chunks

Use document_splitter to break documents (PDF, HTML, Markdown, TXT) into text chunks.

import pixeltable as pxt
from pixeltable.functions.document import document_splitter

pxt.drop_dir('split_demo', force=True)
pxt.create_dir('split_demo')

docs = pxt.create_table('split_demo/docs', {'doc': pxt.Document})
docs.insert(
    [
        {
            'doc': 'https://raw.githubusercontent.com/pixeltable/pixeltable/main/docs/resources/rag-demo/Jefferson-Amazon.pdf'
        }
    ]
)

Inserted 1 row with 0 errors in 0.13 s (7.68 rows/s)
1 row inserted.

chunks = pxt.create_view(
    'split_demo/doc_chunks',
    docs,
    iterator=document_splitter(
        docs.doc, separators='sentence,token_limit', limit=300
    ),
)
chunks.select(chunks.text).limit(3).collect()

Available separators:

heading — Split on HTML/Markdown headings
sentence — Split on sentence boundaries (requires spacy)
token_limit — Split by token count (requires tiktoken)
char_limit — Split by character count
page — Split by page (PDF only)

SDK Reference: document_splitter

Extract frames from videos

Use frame_iterator to extract frames at specified intervals.

from pixeltable.functions.video import frame_iterator

videos = pxt.create_table('split_demo/videos', {'video': pxt.Video})
videos.insert(
    [
        {
            'video': 'https://github.com/pixeltable/pixeltable/raw/main/docs/resources/bangkok.mp4'
        }
    ]
)

Inserted 1 row with 0 errors in 1.28 s (0.78 rows/s)
1 row inserted.

frames = pxt.create_view(
    'split_demo/frames',
    videos,
    iterator=frame_iterator(videos.video, fps=1.0),
)
frames.select(frames.frame, frames.frame_attrs).limit(3).collect()

frame_iterator options:

fps — Frames per second to extract
num_frames — Extract exact number of frames (evenly spaced)
keyframes_only — Extract only keyframes

SDK Reference: frame_iterator

Split videos into segments

Use video_splitter to divide videos into smaller clips.

from pixeltable.functions.video import video_splitter

segments = pxt.create_view(
    'split_demo/segments',
    videos,
    iterator=video_splitter(
        videos.video, duration=5.0, min_segment_duration=1.0
    ),
)
segments.select(
    segments.segment_start, segments.segment_end, segments.video_segment
).limit(3).collect()

video_splitter options:

duration — Duration of each segment in seconds
overlap — Overlap between segments in seconds
min_segment_duration — Drop last segment if shorter than this

SDK Reference: video_splitter

Split strings into sentences

Use string_splitter to divide text into sentences.

from pixeltable.functions.string import string_splitter

texts = pxt.create_table('split_demo/texts', {'content': pxt.String})
texts.insert(
    [
        {
            'content': 'AI data infrastructure simplifies ML workflows. Declarative pipelines update incrementally. This makes development faster and more maintainable.'
        }
    ]
)

Inserted 1 row with 0 errors in 0.03 s (38.38 rows/s)
1 row inserted.

sentences = pxt.create_view(
    'split_demo/sentences',
    texts,
    iterator=string_splitter(texts.content, separators='sentence'),
)
sentences.select(sentences.text).collect()

SDK Reference: string_splitter

Tile images for analysis

Use tile_iterator to divide large images into a grid of smaller tiles. This is useful for processing high-resolution images that are too large to analyze at once, or for running object detection on different regions.

from pixeltable.functions.image import tile_iterator

images = pxt.create_table('split_demo/images', {'image': pxt.Image})
images.insert(
    [
        {
            'image': 'https://raw.githubusercontent.com/pixeltable/pixeltable/main/docs/resources/pixeltable-logo-large.png'
        }
    ]
)

Inserted 1 row with 0 errors in 0.09 s (11.69 rows/s)
1 row inserted.

tiles = pxt.create_view(
    'split_demo/tiles',
    images,
    iterator=tile_iterator(images.image, tile_size=(100, 100)),
)

tile_iterator options:

tile_size — Size of each tile as (width, height)
overlap — Overlap between adjacent tiles as (width, height)

SDK Reference: tile_iterator

tiles.select(tiles.tile_coord, tiles.tile).sample(n=4).collect()

Split audio into chunks

Use audio_splitter to divide audio files into time-based segments for transcription or analysis.

from pixeltable.functions.audio import audio_splitter

audio = pxt.create_table('split_demo/audio', {'audio': pxt.Audio})
audio.insert(
    [
        {
            'audio': 'https://raw.githubusercontent.com/pixeltable/pixeltable/main/docs/resources/10-minute%20tour%20of%20Pixeltable.mp3'
        }
    ]
)

Inserted 1 row with 0 errors in 0.67 s (1.50 rows/s)
1 row inserted.

audio_segments = pxt.create_view(
    'split_demo/audio_chunks',
    audio,
    iterator=audio_splitter(audio.audio, duration=30.0, overlap=2.0),
)
audio_segments.select(
    audio_segments.segment_start, audio_segments.segment_end
).limit(5).collect()

audio_splitter options:

duration — Duration of each chunk in seconds
overlap — Overlap between chunks in seconds
min_segment_duration — Drop last chunk if shorter than this

SDK Reference: audio_splitter

Welcome to Pixeltable

Tools

Core Concepts

How-To

Split data into multiple rows with iterators

Problem

Solution

Setup

Split documents into chunks

Extract frames from videos

Split videos into segments

Split strings into sentences

Tile images for analysis

Split audio into chunks

See also

​Problem

​Solution

​Setup

​Split documents into chunks

​Extract frames from videos

​Split videos into segments

​Split strings into sentences

​Tile images for analysis

​Split audio into chunks

​See also

Problem

Solution

Setup

Split documents into chunks

Extract frames from videos

Split videos into segments

Split strings into sentences

Tile images for analysis

Split audio into chunks

See also