CLI tool for converting articles to podcasts using AI Text-to-Speech APIs. I have added ElevenLabs basic functionanlity, but it's very simple, and I still use OpenAI more for it's cheapness.
You need to have ffmpeg installed before running this CLI tool.
brew install ffmpegSince JS based articles can't be rendered with requests we're using playwright and chromium web driver to tackle that:
pip install playwright
playwright install chromiumInstall audio-reads with:
pipx install audio-readsaudio-reads --help
Usage: python -m audio_reads [OPTIONS]
Options:
--url TEXT URL of the article to be fetched.
--file-url-list FILE Path to a file with URLs placed on every new
line.
--directory DIRECTORY Directory where the output audio file will
be saved. The filename will be derived from
the article title.
--audio-format [mp3|opus|aac|flac|pcm]
The audio format for the output file.
Default is mp3.
--model [tts-1|tts-1-hd] The model to be used for text-to-speech
conversion.
--voice [alloy|echo|fable|onyx|nova|shimmer]
The voice to be used for the text-to-speech
conversion. Voice options: alloy: A
balanced and neutral voice. echo: A more
dynamic and engaging voice. fable: A
narrative and storytelling voice. onyx: A
deep and resonant voice. nova: A bright
and energetic voice. shimmer: A soft and
soothing voice. Experiment with different
voices to find one that matches your desired
tone and audience. The current voices are
optimized for English.
--strip INTEGER RANGE By what number of chars to strip the text to
send to OpenAI. [5<=x<=2000]
--help Show this message and exit.export OPENAI_API_KEY="your-api-key"
audio-reads \
--url 'https://blog.kubetools.io/kopylot-an-ai-powered-kubernetes-assistant-for-devops-developers'ElevenLabs:
export ELEVEN_API_KEY="your-api-key"
audio-reads \
--url 'https://incident.io/blog/psychological-safety-in-incident-management' \
--vendor elevenlabs \
--directory ~/Downloads/PodcastsIf you're using Nix you can start running the tool by entering:
nix developexport OPENAI_API_KEY="your-api-key"
python \
-m audio_reads \
--model tts-1-hd \
--voice nova \
--directory . \
--url 'https://blog.kubetools.io/kopylot-an-ai-powered-kubernetes-assistant-for-devops-developers'If you used nix develop all necessary dependencies should have already
been installed, so you can just run:
pytest- Bypass Cloudflare block
- Minimize costs on tests
- Add ability to render images to text and send over to text to speech as well
- Shorten filename created
- Shorten article title printed to console
- Send to device right away
- Replace print with logger
- Remove redundant warnings in pytest
- Make sure pytest shows quota errors
- OPENAI_API_KEY secret was added to repository secrets
- PYPI_TOKEN was added to release environment secrets
- Elevenlabs test do not require api key for small size requests
- Long frustration of unread articles
- https://github.com/simonw/ospeak
