DrDictaphone

Dictation app for the terminal and Neovim, using Whisper for transcription and ChatGPT for post-processing.

Installation

You can use the installation script:

curl https://raw.githubusercontent.com/olekli/DrDictaphone/main/script/install.sh | sh

Or create a virtual environment and do:

pip install drdictaphone
python -m drdictaphone.cli install ~/DrDictaphone

Place OpenAI API key in ~/DrDictaphone/config/openai_api_key.

Running

To start the standalone app, do ./drdictaphone.

To start only the server, do ./drdictaphone server.

Shutdown a running server by doing ./drdictaphone shutdown.

Neovim Plugin

If you are not already using Python plugins in Neovim, you need to create a virtual environment for Neovim to use. Tell Neovim about it by adding to your init.vim:

let g:python3_host_prog = '~/.neovim-venv/bin/python'

(Or wherever your venv is located.)

Inside this virtual environment, install the Neovim plugin:

pip install drdictaphone-neovim-plugin

Now you need to add the plugin to your Neovim config directory:

ln -s ~/.neovim-venv/lib/python3.11/site-packages/drdictaphone_neovim/DrDictaphone.py ~/.config/nvim/rplugin/python3/.

(Your paths may vary.)

Then start the server. Do :UpdateRemotePlugins once in Neovim, restart. Use DrDictaphoneSetProfile, DrDictaphoneToggle vim commands.

Controlling Standalone App:

• s: select profile
• p: start / stop and transcribe recording
• d: stop and discard recording
• q: exit

Profiles

Profiles consist of:

• topic for transcribing and post-processing, a list of strings
• language to use for the transcriber, a string
• output directory, a string
• output_command to pipe output to
• enable_vad whether or not to enable VAD, a bool, defaults to false

Output will be written to a timestamped file in the output directory.

VAD will filter recordings for parts with voice before processing them.

Post-Processor

The Post-Processor specs consist of:

• instructions for the post-processor, either a filename to load from or a list of strings
• gpt_model to use for post-processing, either a filename to load from or an object
• options to use for post-processing, either a filename to load from or an object
• tools to use for the post-processor, either a filename to load from or an object

The context for the post-processor is built from the profile and the post-processor specs. Settings in the profile take precedence over settings in the specs.