Does FLUX Schnell benefit from detailed prompts? I made a glif to compare - left is the simple prompt, right shows the same prompt enhanced by Llama 70B (for which I used a Claude 3.5 prompt writer glif that generated Llama's system prompt) for paintings, a clear improvement:

🐙 Stage Visuals: KREAM x IAMNEUBERT for Drumsheds London 48 x 4 meters. A 12:1 aspect ratio. The largest club in the world, with one of the widest stage screens. This collaboration pushed @runwayml to the limits creating generations at 24000px. I'm happy to share some of the 150 bespoke generations for @wearekream's legendary Drumshed's show last Friday. These visuals debut their new visual brand for the upcoming summer. Pushing AI to the limits working with such an extreme aspect ratio and an equally insane screen size has been exciting. A huge thank you to Daniel and Markus for believing in this project! 🙏🏼🤍 Live recording dropping soon!

Meow 😻

we loved it! @rakelle @studioDao

YOUR LUCKY DAY was rejected by Sundance, SXSW, Toronto, Tribeca, and Cannes. CAA, WME, UTA and more all passed on representing the movie because it was “too small”. It’s now in the Top 10 of all movies streaming on Netflix for the second week in a row. No one knows anything.

First Person View photography using Midjourney Here is how you can use it: 💬 first person perspective photo ... → Think of it as you being the subject, therefore no need to mention it. Instead focus on the action and the environment. 💬 35mm first person perspective photo on a BMX in the crowded streets of Tokyo As you can see in the example above, I did not mention any subject, this helps Midjourney's understand the key token. Otherwise, you will most likely end up with 1 shot out of 10 that is a true "first person view"

Whodunit? 1/10 #powerdalle #magnific #photoshop #aiart

“What is the root of suffering?” Tokenization What if that’s not an ai joke & actually correct? Like if Tanha and 90% of human suffering is caused by compression errors, and glitches in human inference & perception there’s some compelling neuroscience theories

"MOSTRO" Fellowship artist @MowglyLee recently directed a new short film in collaboration with A$AP Rocky for @PUMA. Also, a new collection is on its way to Fellowship 😶

Sora's video quality seems impossible so I dug into how it works under the hood it uses both diffusion (starting with noise, refining towards a desired video) and transformer architectures (handling sequential video frames) read on 🧵

New work coming from Backlot, mysterious dog story. Woof. Unconcerned with perfection, here is our first rough cut, no audio. youtube.com/watch?v=0TclUKz0…

So good

@rakelle and I are super excited to share our first AI short from Backlot. A labor of love, and a peek into the future. Made possible with the support of @juiceboxETH and @studioDao youtube.com/watch?v=nadVorZx…

Do you guys ever think about how, despite paying for six different streaming platforms, the film you want to watch is never available on any of them?

“SWEETNESS OF THE NEW” (sound on) Made with Stable Diffusion, Runway, and After Effects. Danced & Edited by me. Music by Gray. A tribute to one of the greatest artists to live.