It feels so much more sketchy when trying to do this with ChatGPT voice though. When dalle3 landed everyone was sharing its system prompt, but I haven’t seen the same emphasis on getting the conversational system prompt.
@fofrai
-
Achieving Sub-100ms Performance Through Optimization Techniques
By
–
Very. I reckon sub 100ms is doable with the right optimisations.
-
LCM Animation Model Now Available on Replicate
By
–
Here's a model on Replicate to play with this: https://
replicate.com/fofr/lcm-anima
tion
… -
LCM Animation Generation: 32 Frames in 10 Seconds
By
–
Playing around with LCM and animation – 10 seconds to produce a 32 frame animation at 512×512.
— fofr (@fofrAI) 26 octobre 2023
There's no control yet, just fixed seed, so it still jumps around a lot, but it's fast. pic.twitter.com/L03vp0szn6Playing around with LCM and animation – 10 seconds to produce a 32 frame animation at 512×512. There's no control yet, just fixed seed, so it still jumps around a lot, but it's fast.
-
Generate Images in One Second Locally on Mac
By
–
New blog post up – Generate images in one second on your Mac using a latent consistency model.
— fofr (@fofrAI) 25 octobre 2023
1 second. Locally. 🤯https://t.co/BWJqsaNy1s pic.twitter.com/bzIYl1v4lmNew blog post up – Generate images in one second on your Mac using a latent consistency model. 1 second. Locally. https://
replicate.com/blog/run-laten
t-consistency-model-on-mac
… -
Latent Consistency Model Pipeline Released on Diffusers
By
–
There was a pipeline in a diffusers PR, I think it’s landed on main now. I’ve also pushed this model: https://
replicate.com/fofr/latent-co
nsistency-model
… -
A40 GPU achieves 34.51 iterations per second performance
By
–
34.51it/s on A40
1 step That's like 29ms? -
Latent Consistency Model img2img: Single-Step Quality Generation
By
–
1 step img2img with a latent consistency model is wild. The quality and speed of the outputs is amazing. 1,3: Input image
2,4: img2img with 1 step (0.65 and 0.5 prompt strengths) -
LCM img2img model now available on Replicate platform
By
–
I've pushed an LCM img2img model to Replicate – needs to be faster, but you can explore img2img with it: https://
replicate.com/fofr/latent-co
nsistency-model
… -
Zero123Plus Advances in Multimodal AI Generation
By
–
I agree with @flngr
, Zero123Plus is doing amazing things!