Download gary4beatbox App V. 2.4
Apple AppStore - Google Play
gary4beatbox App V. 2.4
by Kevin Griffing
gary takes your input audio and runs away with it. the first version was designed to continue your beatboxes. this new version has the stable-audio-open-small model that you can use to generate drum - More Info...

Available Since: 2024-10-06
Developer: Kevin Griffing
Seller: Kevin Griffing
Price: 0 USD
Current version: 2.4
Last Updated: 2025-06-30

Ratings: 5 out of 5 stars.
Based on 3 reviews. (Rate/Review)
Genre: Music
Filesize: 5.21MB
IOS Bundle ID: com.thecollabagepatch.gary-for-beatbox
Age Rating: 4+
Languages: EN

Videos: gary4beatbox App Videos

gary takes your input audio and runs away with it.

the first version was designed to continue your beatboxes. this new version has the stable-audio-open-small model that you can use to generate drums first, and then have gary continue it, or terry transform it.

if you do beatbox, you can still 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.

now in the app, we have 'terry'. it's the new model by meta, melodyflow. It can transform either your recorded audio or the audio you got back from gary.

undo what terry just did to try again. make sure to set the 'flowstep' which is kinda like the 'strength' of the transformation. it would seem higher is almost always better if you want it to truly retain the shape of what you gave to terry.

use presets or enter your own custom prompt to have terry transform the audio. this is very experimental. no other application in the world has this model in it.

learn more about terry here https://huggingface.co/spaces/facebook/Melodyflow

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)

Want BIG updates? Enter your email address:

Delivered Once Every Tuesday.