1- A lot of options to influence/edit the generated output. What AUTOMATIC1111 is building for Stable Diffusion is the right direction (open extensions and manoeuvrability). I hope harmonai will get the same treatment.
2- Smaller minimum training requirements and simple retraining workflows. AudioLM is outstanding in this regard (but fails the first point)
3- Prod-level quality for end-user tools : Style/tone transfer and cloning plugins like DDSP-VST, mawf and yours (RAVE) sounds at best like DALLE-1 level quality. do you think we could make a DALLE-2 kind of jump soon?
And we do indeed lack public gathering places for this scene ! (or do we ?)
1- A lot of options to influence/edit the generated output. What AUTOMATIC1111 is building for Stable Diffusion is the right direction (open extensions and manoeuvrability). I hope harmonai will get the same treatment.
2- Smaller minimum training requirements and simple retraining workflows. AudioLM is outstanding in this regard (but fails the first point)
3- Prod-level quality for end-user tools : Style/tone transfer and cloning plugins like DDSP-VST, mawf and yours (RAVE) sounds at best like DALLE-1 level quality. do you think we could make a DALLE-2 kind of jump soon?
And we do indeed lack public gathering places for this scene ! (or do we ?)
Here is a personal list I made of some AI Music creation tools: https://rentry.co/Music-Creation-AI-Tools
Merci IRCAM !