Skip to main content
Open in Kaggle  Open in Colab  Download Notebook
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.
Automatically create preview thumbnails from video files at specific timestamps or intervals.

​
Problem

You have video files that need preview thumbnails for galleries, search results, or video players. Manually extracting frames doesn’t scale.

​
Solution

What’s in this recipe:
  • Extract thumbnail at a specific timestamp
  • Generate multiple thumbnails per video
  • Resize thumbnails to standard dimensions
You add computed columns that extract frames from videos. Thumbnails are generated automatically when you insert new videos.

​
Setup

%pip install -qU pixeltable

import pixeltable as pxt
from pixeltable.functions import video as video_fn

​
Load videos

# Create a fresh directory
pxt.drop_dir('thumbnail_demo', force=True)
pxt.create_dir('thumbnail_demo')

Connected to Pixeltable database at: postgresql+psycopg://postgres:@/pixeltable?host=/Users/pjlb/.pixeltable/pgdata
Created directory ‘thumbnail_demo’.
<pixeltable.catalog.dir.Dir at 0x14d3a1990>
# Create table for videos
videos = pxt.create_table('thumbnail_demo.videos', {'video': pxt.Video})

Created table ‘videos’.
# Insert sample videos from public S3 bucket
s3_prefix = 's3://multimedia-commons/'
video_paths = [
    'data/videos/mp4/ffe/ffb/ffeffbef41bbc269810b2a1a888de.mp4',
    'data/videos/mp4/ffe/feb/ffefebb41485539f964760e6115fbc44.mp4',
]

videos.insert([{'video': s3_prefix + path} for path in video_paths])

Inserting rows into `videos`: 2 rows [00:00, 382.20 rows/s]
Inserted 2 rows with 0 errors.
2 rows inserted, 4 values computed.
# View videos
videos.collect()

​
Extract thumbnail at timestamp

Extract a single frame at a specific time (e.g., 1 second into the video): 
# Extract frame at 1 second as thumbnail
videos.add_computed_column(
    thumbnail=video_fn.extract_frame(videos.video, timestamp=1.0)
)

Added 2 column values with 0 errors.
2 rows updated, 2 values computed.
# View thumbnails
videos.select(videos.video, videos.thumbnail).collect()

​
Resize thumbnails

Create standard-sized thumbnails for consistent display: 
# Resize thumbnail to 320x180 (16:9 aspect ratio)
videos.add_computed_column(
    thumbnail_small=videos.thumbnail.resize((320, 180))
)

Added 2 column values with 0 errors.
2 rows updated, 2 values computed.
# View resized thumbnails with dimensions
videos.select(
    videos.thumbnail_small,
    videos.thumbnail_small.width,
    videos.thumbnail_small.height
).collect()

​
Multiple thumbnails with FrameIterator

For preview strips or timeline thumbnails, use FrameIterator to extract multiple frames: 
from pixeltable.iterators import FrameIterator

# Create a view with frames extracted at 0.5 fps (one frame every 2 seconds)
frames = pxt.create_view(
    'thumbnail_demo.frames',
    videos,
    iterator=FrameIterator.create(video=videos.video, fps=0.5)
)

Inserting rows into `frames`: 17 rows [00:00, 9736.88 rows/s]
# View extracted frames (multiple per video)
frames.select(frames.frame, frames.pos).head(10)

​
Explanation

Thumbnail extraction methods:
Common thumbnail sizes:

​
See also