mli/autocut
Autocut
Autocut is a text-based video editor and automatic speech recognition tool. It allows users to cut and merge video clips by modifying a text transcript instead of using a traditional timeline.
The system operates as an FFmpeg video processor and subtitle manipulation utility. It converts spoken audio into text and compacts subtitle files into simplified formats, enabling the removal of unwanted video segments by deleting corresponding sentences from a transcription file.
The project covers automated video transcription, non-linear video cutting, and subtitle file management. It supports hardware acceleration to increase the processing speed of transcription and video manipulation tasks.
Features
- Transcript-Based Editing - Enables non-linear video cutting by mapping text timestamps to video frames within a transcription file.
- Automatic Speech Recognition - Functions as an automatic speech recognition system that converts video audio into editable text files.
- Speech to Text Transcription - Converts spoken audio from video files into text formats using configurable speech-to-text models.
- Automated Video Transcribers - Automates the conversion of spoken audio from video files into time-synced text transcripts.
- Transcript-to-Timestamp Mapping - Generates precise video edit points by mapping text indices from a transcript to specific timecodes.
- FFmpeg Wrappers - Provides a processing engine that leverages FFmpeg for the precise cutting and merging of video segments.
- Automated Processors - Uses an automated FFmpeg-based pipeline to execute precise cuts and concatenations based on timestamped text.
- Transcript-Driven Editors - Provides a text-based video editor that enables cutting and merging by editing a transcript instead of a timeline.
- Non-Linear Video Editing - Allows removing unwanted video segments through the modification of a transcription file instead of manual scrubbing.
- Text-Based Cutting - Implements a system for cutting and merging video clips by deleting or rearranging sentences in a text transcript.
- Text-Based Trimming - Allows the removal of unwanted video segments by deleting corresponding sentences from a transcription or subtitle file.
- GPU Acceleration - Implements hardware acceleration via GPU to speed up the deep learning models used for audio transcription.
- Lossless Concatenations - Ships the ability to merge pre-cut video portions losslessly without re-encoding the entire file.
- Subtitle Processing - Provides utilities for cleaning and compacting subtitle files to make manual text editing more efficient.
- Subtitle Format Converters - Provides a utility for converting standard subtitle files into compact versions for easier manual cleaning.
- Video Assembly - Assembles final video clips by combining segments based on a modified text transcript.
- Video Editing - Combines multiple edited video segments into a single output file based on a selection list.
