Tharsis Souza

Results 21 issues of Tharsis Souza

best config 1. OpenAI o1 for transcript generation 2. Google TTS Multispeaker for audio generation Upload to README. This is going to be fun!

HIGH Priority

mypy podcastfy/*.py ``` podcastfy/content_generator.py:56: error: Unexpected keyword argument "max_output_tokens" for "ChatGoogleGenerativeAI" [call-arg] podcastfy/content_generator.py:56: error: Incompatible types in assignment (expression has type "ChatGoogleGenerativeAI", variable has type "Llamafile") [assignment] podcastfy/content_generator.py:62: error: Unexpected...

"evals are surprisingly often all you need" But here we are evaluating a pretty novel dimension: How can we systematically quantify generated text/audio is engaging, follows a target configuration while...

enhancement

Integrate with docling instead of building our own content parsers from scratch: https://github.com/DS4SD/docling Improved support, maintainability as well as coverage of courses since docling supports (PDF, DOCX, PPTX, Images, HTML,...

good first issue
Fund

https://github.com/soimort/you-get Enable download of vídeos, images etc from input website

Currently, the tool supports URLs and PDF files as input. Users can also provide a file containing a set of links for processing. We aim to add an option allowing...

good first issue
HIGH Priority

Add support for additional tags dynamically given TTS model. Right now it's the intersection of OpenAI/MS Edgeand ElevenLabs supported tags fixed values. Modify text_to_speech.py::clean_tss_markup

enhancement

Add option that allow users to process provided website url + its subpages.

enhancement

String should have at most 4096 characters before passing to TSS. We should split the transcript into 4096 chunks -> send to TSS then stitch them together to generate the...

bug

Add a way for users to know how much a podcast generation will cost before running it

enhancement