1628852404157353986

@LumaLabsAI @camenduru It would be good to explain what this means in terms of Before vs After to people that have never used this tech and don’t understand if or why this is a big deal.

1628836036464852993

Just the begining. The progress we are making with #GenerativeA Ifor voices will allow @YouTube to auto-generate dub voices in any language for any video automatically. >

YouTube launches a multi-language audio feature for dubbing videos https://t.co/VKIRUjKCoG

1628832992008323072

Dear @VentureBeat, we have been running Stable Diffusion on a smartphone for a long time, thanks to @drawthingsapp. And performance thx to support for CoreML is good. >

Can Stable Diffusion generative AI run on a smartphone? Qualcomm says — Yes it can https://t.co/SoBLwBnayC https://t.co/kbaGNuGcn0

1628821812040679424

https://t.co/RJiFvc2HTd.amazing.

Install #StableDiffusion and #AUTOMATIC1111 inside your Google Drive (no more downloading of the models every time) and run it on Colab compute instances.

YES.

@camenduru is the person to thank for this gift. https://t.co/ssY8Zr3J7X

1628823998971756549

Tomorrow, I send out Issue #1 of Synthetic Work, my new newsletter on applied #AI. Both the Free Edition and the Splendid Edition.

It’s crazy how many people signed up without even reading one email*. Thank you.

*Boy, how many will be disappointed over the weekend.

@wintonARK Speaking from past experience training a LLM for an enterprise use case: if you have found a mechanism that yields more accurate & consistent output from a LLM (especially if open source, eg CodeGen) you’ll be inclined to keep it secret as it’s one of your competitive advantages.