search

nateraw / stable-diffusion-videos

Create 🔥 videos with Stable Diffusion by exploring the latent space and morphing between text prompts
Apache License 2.0
4.4k stars 421 forks source link

Determine if best default is lerp or slerp for text embeddings #46

Closed nateraw closed 1 year ago

nateraw commented 1 year ago

The use_lerp_for_text param lets you use lerp instead of slerp for text embedding interpolation. There was a comment on my original gist suggesting lerp was better, but I never got around to doing exhaustive comparison.

Would be nice if we could do this + set the value (or even get rid of the option to change it) based on what's best.

nateraw commented 1 year ago

@0x1355 have you experimented with this by any chance?

0x1355 commented 1 year ago

Not yet but it is now priority one on my experiment board. Will let you know soon.

0x1355 commented 1 year ago

Experiment done. Writing up. Taking longer because I am changing something in my pipeline. Will share result tomorrow 🔆

nateraw commented 1 year ago

Thank you!!! Looking forward to it. Also running a few experiments separately

0x1355 commented 1 year ago

Putting my thoughts in a thread and tagged some AI artists I follow. Hope we can get more people to compare notes.

https://twitter.com/DreamLatent/status/1573166763268767744