subtitler

Listens to your computer's microphone, transcribes it to text, translates it from English to Portuguese, by default, and displays subtitles as a screen overlay - live.

Feels like magic... And fun, and silly! Maybe useful, you never know.

Requirements

(on a works-for-me after fiddling-with-it-for-a-couple-of-hours basis)

• A computer running macOS (the UI overlay needs tweaks for other platforms).
• Internet connectivity.
• Python 3.14 (should work with earlier versions, I suppose).
• An API key from Deepgram for speech to text. [see note 1]
• An API key from DeepL for translation. [see note 2]

Notes, as of this writing:

1. There's a free USD 200.00 credit in a pay as you go model; enough for 400+ hours of speech to text with the most expensive model.
2. The free version allows for 50k characters/month.

Limitations

• The subtitles are displayed in a rectangular transparent window sitting on top of all other windows, towards the bottom of the screen - click through does not work.

• Behaviour with multiple displays is unpredictable.

• macOS fullscreen applications won't be overlayed, they take a "virtual display" of their own!

Running

• Set the DEEPGRAM_API_KEY environment variable.
• Set the DEEPL_API_KEY environment variable.
• uv run subtitler

Tweaking

• Play with the src/subtitler/config.py module.
• Things you can easily change:
  • Spoken language.
  • Subtitle language.
  • Font family and size.
  • Subtitle timing.

The background story

On April 1st, 2026, a person, isi, joined the Python Lisbon Meetup Discord channel and asked:

Hello, it’s a pleasure to meet you all. I’ve just started learning Python and noticed that you organise meet-ups on this programming language. Do you think I’ll be able to keep up, even though I’m a beginner, or are the topics too advanced? (I only speak Spanish or Portuguese )

As an organizer, I promptly responded:

If you’re interested in Python you will be welcome… Join us tomorrow and come say hi. As for languages, we speak all of them… But mostly Python! 😊

A few minutes later, Rodrigo, a co-organizer, noted, in private, that I didn't mention that talks tend to be in English.

And then, in a friendly tone, he also responded to isi, in Portuguese:

Todos são bem-vindos, independentemente do nível de experiência . 😄 As palestras costumam ser dadas em inglês por causa do número de estrangeiros que frequentam os nossos encontros. Como disseste que só falas português ou espanhol, mas escreveste a mensagem em inglês, fiquei na dúvida se isto podia ser um problema ou não :/

To which, I replied, sharing a thought:

(pensamento: conseguiremos ter palestras em inglês, por exemplo, legendadas em português?!… graçola ou pensamento sério?) 🤔

Then Marco provocatively (in a funny way) replied:

É só arranjar um estenografo

My last message, to Marco, was:

…ah, “só” isso. Resolvido, então! 🤷🏻‍♂️

Development Notes (for those concerned about AI)

• I chatted a bit with Google Gemini in exploring ideas and identifying speech to text and translation services I could easily use from Python.

• I chatted a bit with Claude via llm, a CLI tool I use all the time, to get my AsyncIO skills up to speed (hadn't touched async code in a while).

• Having prior experience with sounddevice and with tkinter as a tool to create transparent text overlays, the remaining code was mostly adapted from Deepgram/DeepL examples and docs.

• All of the code was written by myself, mostly as an experiment, and of course as a fun thing to bring to our monthly meetup, tomorrow, April 2nd, 2026.

License

Copyright © 2026 Tiago Montes, MIT License - see the LICENSE file.