Today, we’re launching Reve 2.0, the best 4K image model in the world.
We invented a new way to generate and edit any image using precise layouts. For the first time, it’s possible to create images you can touch.
Reve's new text-to-image model is here. Really proud of the team to rank at #3 with the big labs. To my knowledge, we are the first lab to use native pixel space diffusion without latent autoencoder at 4k (16MP) resolution for production level image generation.
It was very fun to pull this editor together for the 1.0 launch with @dkrajews123.
We're shipping bugfixes & improvements already based on feedback. Let us know what works for you & what doesn't!
Discovered an insanely cool feature in the new @reve image model.
You can now edit specific objects within an image - move, delete, or prompt changes while keeping the rest consistent.
Watch me manipulate this scene ⬇️
@emollick@reveimage Nailing it with the simple prompt box in the public preview would involve lots of dice rolling! I've got more knobs to turn; @m_gharbi:
"Our vision is to build a new semantic intermediate representation that both a human and a machine can understand, reason about, and operate on"
Typography in generative models has been a hard problem to solve. Reve Image 1 does not only generate long text, but also it is very good at stylized typography.
You can generate designs like this, it follows the prompt faithfully.
Anyone else try Reve image generator yet? @reveimage Discovered it earlier today and it's DAMN GOOD, very nice surprise! Aesthetics are excellent! Solid prompt adherence as well. Censorship is also quite relaxed. Results can look very cinematic! Highly recommend trying it when you can, it seems like their servers are overloaded right now. https://t.co/vmOncEMtgX
Brought one of the frames I generated with Reve to life using @Kling_ai
The Halfmoon 🌓 reveal: Congratulations to @reveimage on creating the world’s leading image generation model with Reve Image!
Reve Image has been in the Artificial Analysis Image Arena over the past week and is the clear leader, beating strong competition including Recraft V3, Google's Imagen 3, Midjourney v6.1, and Black Forest Lab's FLUX.1.1 [pro].
The model stands out for its impressive text rendering, prompt adherence, and aesthetics. Currently, the model is accessible through Reve’s website (link below), an API is not yet available.
See below for example images and a link to try the model out for yourself 🧵
Help me test websockets across timezones? I just need you to wiggle your cursor / drag your finger from wherever you are right now: https://t.co/9ycqAgk6zC
@persingerscott The capabilities of my Google home mesh network are unparalleled. Speed, reliability, simplicity, even whole-house music.
But having to use the Google home app to control it all is miserable.