4 ms·
Really impressive. If we are able to generate such detailed images, is there anything similar for text to music? I would I though that it would be simpler to ac
by y04nn 4y ago
Really impressive. If we are able to generate such detailed images, is there anything similar for text to music? I would I though that it would be simpler to achieve than text to image.
- nomel 4y agoCompare the size of a raw image file to a raw music file, to get an idea of the complexity difference.
- redox99 4y agoOur language is much more effective at describing images than music.
- tomatowurst 4y agowhy stop at audio? the pinnacle of this would be text-to-videos, equally indistinguishable from real thing.
- burlesona 4y agoThe way things look when still is much easier to fake than the way things move. I would expect AI development to follow a similar path to digital media generally, as its following the increasing difficulty and space requirements of digitally representing said media: text < basic sounds < images < advanced audio < video. What’s more impressive to me is how far ahead text-to-speech is, but I think the explanation is straightforward (the accessibility value has motivated us to work on that for a lot longer).
- touringa 4y agoSymphonyNet: https://youtu.be/m4tT5fx_ih8 https://youtu.be/m4tT5fx_ih8