Skip to content
This repository has been archived by the owner on Nov 1, 2024. It is now read-only.
/ squawk Public archive

Automatic subtitles for DaVinci Resolve with OpenAI Whisper

License

Notifications You must be signed in to change notification settings

in03/squawk

Repository files navigation

squawk

Squawk logo

GitHub Code style: black GitHub branch checks state pre-commit.ci status

GitHub last commit GitHub Repo stars


Important

This repository is no longer maintained

Davinci Resolve 18.5 supports automatic subtitles natively. It is recommended to upgrade. If you must use an older version of Resolve and still want to use Squawk, it will still function, but there are no plans for updates.


Automatic subtitles for DaVinci Resolve with OpenAI Whisper ✨

Save time writing subtitles yourself and don't pay an online service to do it. Squawk generates subtitles for your videos in DaVinci Resolve! It renders audio of your timeline, transcribes it (speech-to-text) with Whisper, and automatically imports the subtitles into Resolve. Whisper can even translate other languages into English before transcribing.

Note

Currently only Resolve 18 is supported. Resolve 17 and older require Python 3.6, which is now EOL. Some dependencies have started dropping support for it. A Resolve 17 branch doesn't currently exist, and will not unless someone asks nicely for it or forks the project.

Installation

Install with pipx

pipx install git+https://github.com/in03/squawk

Usage

Simply run:

squawk timeline

to render, transcribe and import the subtitles of the active timeline.

You can pass "Timeline Name" as an optional argument to select a specific timeline:

squawk timeline "My Timeline"

If you already have the file you're transcribing rendered, pass the file:

squawk file "My File.mov"

Transcription Accuracy

Depending on the quality of your audio, you can try different language models. The larger the model, the higher chance of an accurate transcription, but the slower the analysis. You can choose the model used in the user configuration file.

If your audio is clear and high-quality like a professional VO, you might get away just fine with the tiny model. If your audio contains dialogue with lots of sound effects, music and people talking over each other, you might want to start with medium. Try large as a last resort if the transcription contains a lot of incorrect words.

You can see a comparison of Whisper's language models here.

Commands

 _____ _____ _____ _____ _ _ _ _____ 
|   __|     |  |  |  _  | | | |  |  |
|__   |  |  |  |  |     | | | |    -|
|_____|__  _|_____|__|__|_____|__|__|
         |__|


Git build (cloned 🛠 ) '0141446'
Run squawk --help for a list of commands
editor@EMA:~$ squawk --help

✅  Checking settings...

 Usage: squawk [OPTIONS] COMMAND [ARGS]...

╭─ Options ──────────────────────────────────────────────────────────────────────────────────────────────────╮
│ --install-completion   Install completion for the current shell.                                           │
│ --show-completion      Show completion for the current shell, to copy it or customize the installation.    │
│ --help                 Show this message and exit.                                                         | 
╰────────────────────────────────────────────────────────────────────────────────────────────────────────────╯

╭─ Commands ─────────────────────────────────────────────────────────────────────────────────────────────────╮
│ config     Open user settings configuration file for editing                                               │
│ file       Transcribe a media file and import the transcription into Resolve.                              │
│ timeline   Transcribe a Resolve timeline and import the transcription into Resolve.                        | 
╰────────────────────────────────────────────────────────────────────────────────────────────────────────────╯