gary4beatbox app analytics for December 27

gary takes your input audio and runs away with it. this version was designed to continue your beatboxes. record using your mic, with or without noise cancellation, and with a count-in if you need it, and then send it to gary. you can have it continue from where it left off, or you can crop off the end of its output and have it continue from there. if you don't like the extended output, you can retry to have it go from the same place with different settings. some features: use any audiocraft/musicgen model on the huggingface hub by entering it into settings. the 'tame the gary' setting is supposed to focus the output on percussion and drums, and hopefully retain more of your beatbox in its output.  it's experimental and seems to only work well with the model 'thepatch/vanya_ai_dnb_0.1', but who knows... the prompt duration is the number of seconds of audio you want it to use. when you press 'send to gary' that number of seconds is counted from the beginning of the input. when you press 'continue' or 'retry' that number is taken from the end of the audio. share it as a video if you want to share to twitter, or un-toggle that to simply share the audio. because this is free and open source, i'm really not sure how long our backend will stay up.  you can change the backend url if you've built the backend from source: https://github.com/betweentwomidnights/gary-backend-combined. (this uses the same backend as gary4live, the free ableton plugin that does whatever tf it wants)