twitch_auto_clipper

Automatically create Twitch clips when chat activity spikes.

Twitch Auto Clipper monitors Twitch chat and detects moments where a large number of viewers send the same emote/message within a short period of time. When activity exceeds a configurable threshold, it automatically creates a Twitch clip.

https://github.com/user-attachments/assets/de26cc92-f84c-4ed1-9942-328684f99298

Installation

pip install twitch_auto_clipper

Requirements

Minimum Python version supported by twitch_auto_clipper is 3.10.

Quick start

from twitch_auto_clipper.TwitchAutoClipper import TwitchAutoClipper

clipper = TwitchAutoClipper(
    ["Marlon", "Lacy"],
    "YOUR_CLIENT_ID",
    "YOUR_CLIENT_SECRET",
    on_clip=lambda clip: print(clip),
    common_value=1,
    emote_value=2,
    clipable_message_ratio=0.5,
)

clipper.start()
clipper.join()

Features

• Automatically handles Twitch OAuth token refresh
• Automatically monitors multiple Twitch streams
• Supports custom emotes/messages as clipping triggers
• Optional 7TV emote support
• Runs each streamer in separate thread

How it works

Twitch chat
     │
     ▼
Collect messages
     │
     ▼
Count messages / emotes
     │
     ▼
Detect activity spike
     │
     ▼
Threshold exceeded?
     │
    YES
     ▼
Create Twitch clip
     │
     ▼
Call on_clip callback

Usage

There are two classes that you will typically need:

• TwitchAutoClipper — monitors Twitch chat and automatically creates clips when a message reaches the configured threshold.
• Clip — contains information about a clip that was created.

TwitchAutoClipper

TwitchAutoClipper(
    streamers_names,
    client_id,
    client_secret,
    on_clip,
    counter_interval_seconds=...,
    clipable_message_ratio=...,
    clipable_wait=...,
    common_value=...,
    emote_value=...,
    excluded_words=...,
    logging_handlers=...,
    logging_level=...
)

Clip

A Clip instance contains information about the clip that was created.

Motivation

There are a lot of AI clippers that are mostly paid services, largely as a result of their high inference costs. But why use AI when, on popular streams, you constantly have thousands of people evaluating the current state of the stream through chat?

By using the audience's reactions, you can extract clips from Twitch with almost no processing power.

Roadmap

• Analyze Twitch chat through IRC
• Twitch OAuth authentication
• Automatically create clips
• Monitor multiple live streamers
• Save clips to a log
• Add automated tests
• Pip package
• Web page for constants editing