[1/6] 🔊 We release CLAP2 and LAION-Audio-XL: 365M audio clips (1.39M hours) with 707M captions, used to train five CLAP2 models from scratch.
🏆CLAP2-L reaches 53.6% on MAEB-30, the best CLAP model on the leaderboard.
https://t.co/nxvgVuwo9k
Christoph (@ChristophSchuh6) made a niffty audio segmentation demo using the CLAP2 models we released last week that he shared on his YouTube. He doesn't post much here these days so check it out: https://t.co/4Glb6L1rY3
@foerster_ai I played around with the new models (medium and large)
They are a HUGE step forward from previous encoders, embeddings are much more informative, great work!
As a side note, a caption generator, be it OpenVision2 style, CoCa, whatever, can be really interesting for audio 🤔
[1/6] 🔊 We release CLAP2 and LAION-Audio-XL: 365M audio clips (1.39M hours) with 707M captions, used to train five CLAP2 models from scratch.
🏆CLAP2-L reaches 53.6% on MAEB-30, the best CLAP model on the leaderboard.
https://t.co/nxvgVuwo9k
@InOneProjects@rms80 We just released CLAP2 in collaboration with LAION. It was trained an a lot more data including video game sounds and general sound effects e.g. League of Legends gameplay. https://t.co/QY98ry7fmE
(disclaimer: I'm one of the authors)