Week #5 Update:
Got the model trained, an overfitted one, due to lack of more data for the model (Can replicate and produce the training data but struggles with new generations due to silences in data which were not trimmed and potentially due to some instrumentals clips)
Well seems like we are making progress on the model with speech next thing left to do is start shifting it towards simple music and then singing vocals.
Taking the idea to my toy version of the project. Excited to finish up on the dataset along with my co-founder we did some unconditional generation for the toy version as a test run its all noisy but we plan to improve it next week
@_buildspace@_nightsweekends
@cdt5058 @_buildspace@_nightsweekends@FarzaTV will it be multilingual or just english mode only, would be also nice to have like a short summary of the whole podcast along with it
@dhasandev@FarzaTV@_buildspace@_nightsweekends would be cool if it can be automated and linked to your email so it auto analyzes and updates you from time to time on the docs and legal stuff or like send a summary to your mail
@_buildspace@_nightsweekends
Idea is to let everyone create music whether you are a professional or not, have the base & arch brainstormed for it, let's see where we end up in 6 weeks