OpenSora Releases its first trained checkpoints (2-5 SEC, 512x512 T2V)

24 7 2 years ago

GaggiX2• 2 years ago

Sora in this case really has nothing to do with the model, as OpenAI hasn't released any specific details about the model, we only know that it's a diffusion model using a transformer architecture and it was trained on videos of different dimensions and lengths, this is just an open source text-to-video.

j45• 2 years ago

It's OK. The text-to-video realm won't be solved by a single solution. There's simply too many applications. Exciting times.

devit• 2 years ago

Seems a bad idea to call it Sora given that it has nothing to do with OpenAI Sora (other than achieving a similar goal). Also a trademark violation probably.

amenhotep• 2 years ago

Also just a really bad word to combine with "open". Lots of emphasis necessary on that final "a", lest only the first syllable registers.

geor9e• 2 years ago

I wonder if it would have even been posted, if not for the misleading name.

55555• 2 years ago

is this the open source state-of-the-art for video generation?

GaggiX2• 2 years ago

I imagine it would be the Stable Diffusion Video model.

OpenSora Releases its first trained checkpoints (2-5 SEC, 512x512 T2V)

Related Stories