JustAnotherCoder

I’ve finished PodAngelEX’s GUI! Now the user can search for any podcast in Apple Podcast’s library, using the Itunes Search API!

Furthermore, users can now ‘subscribe’ to albums, adding them to their library tab for quick access.

You can view every episode of any podcast. To reduce impact, I implemented lazy loading with a ‘load more episodes’ button at the bottom of the list.

You can check off as many episodes as you’d like to transcribe, hit the transcribe button, and watch it go! One thing is that transcription will not work if you close the application, so please just minimize it if you prefer background running. All episodes go to an Output folder, in the same location as the app and its dependencies, though I may update it so that cleaned episodes display in their own little area.

All of the settings you could change in PodAngel’s CLI can be edited in the Settings tab, to adjust PodAngelEX to exactly how you like it!

If you find any bugs with the GUI, shoot me a DM over Slack, @JustAnotherCoder, and I’ll get to it as soon as possible.

Please note, as much as I’d love to use GPU acceleration, I’m on Windows with an AMD card, which is a notoriously bad pairing for GPU acceleration. Once Ubuntu’s latest LTS releases, I’ll be able to enable GPU acceleration, but until then, it’s going to have to be CPU…sorry. If you have an Apple though, it won’t be too bad, with unified memory.

One final thing. I really tried my hardest to keep all my work original, but I did use more AI in the making of this GUI update than I did with the CLI, due to just how complex it got. Thanks for (hopefully) understanding, and enjoy PodAngel 2.0!

Album view completed, though a bit buggy. Album view shows title of podcast, artist, description, episodes, their publish date, and their description. Lazy loading integrated. Next step is to add checkboxes for each episode to mark them for cleaning, as well as enabling the user to add podcasts to their library and edit settings in the Settings tab. Image quality for album view is a little low res, due to rescaling.

What a grindfest. I’m working on updating PodAngel with a GUI, so a user can search for podcasts, subscribe to them, download and transcribe episodes all from the application. Using ITunes Search API, PodAngel would have access to basically every podcast located on Apple’s Podcast app. Currently I have a very basic GUI with tab functionality, a search bar that returns the searched podcasts, and like, a quit button. Waiting for an Ubuntu release to continue working on StreamAngel, as Linux has way better libraries for AMD GPUs. Stupid Windows. Oh, I almost forgot, all of the GUI is made with Custom Tkinter, which is a wrapper for Tkinter

Worked on some QOL and packaged for PYPI. The package won’t show up when searched though, which is strange, although direct link worked fine when I gave the code to friends to test.

Alright, I’ve finally completed PodAngelEX! All of the functionality is…well, functional, along with a few changes.

For example, I’ve made it cut audio rather than mute it, to prevent long stretches of silent audio.

Enjoy PodAngel!

I’ve finally figured out how to not only make whisper_timestamped work, but I’ve added the option to run multiple models at the same time, meaning if you have enough VRAM, you can process at 2x, 3x, 4x speed. However, 4x is not the limit, as the way I’ve done it allows for infinite workers to be run concurrently. However, and I cannot stress this enough, if you don’t have the VRAM to run 50 10B param models, DONT.

Alright, I got all the stuff working. Well, not all the stuff. I added a whole ton of functions to have a config menu appear when the application is run for the first time. The last step is to add the actually meat and potatoes of the program, which is the whisper transcription and Detoxify run-throughs.

Decided to take a new direction, as OmegaMap hit a very tall, wide, and all together annoying wall. Instead, I will be making a new version of an old passion project, to be used to listen to podcasts and such without (as much*) swears

I can’t promise complete muting, as that isn’t quite possible in the technological scope of things, but I can claim, from personal use, that it lowers the swear amount by somewhere between 80-90%

Added limited map functionality. Now, the user can input one of the game names in “currated_map_urls.txt” (I know it’s misspelled), and the program will return the URL for the corresponding map, and then that URL is loaded on the website. Currently, .strip and .lower are not working as I intended them to, so the input is incredibly specific, and to cleanly generate another map, one must refresh the page. Even with all that in mind though, my program now has one of its most basic functions up and running!

After much deliberation, decided to use NiceGUI as the GUI for this project, over Tkinter. Maps display properly. Figuring out how to make coordinates work, to find collectables.

I’m working on my first project! This is so exciting. I can’t wait to share more updates as I build.