Diego Rodriguez (@asciidiego)

Diego Rodriguez reposted

Rhys

@RhysSullivan

Jul 3

Horrifying to be listening to a human being speak and realize they’re just a vessel for Claude

229

281

5,444

529,082

Diego Rodriguez reposted

Tinker

@tinkerapi

Jul 2

Training image models requires a surprising amount of Tinkering: prototyping reward models, training a prompt expander, and creating the RL environment. We love the creativity from @krea_ai!

Krea

@krea_ai

Jun 23

our technical report is out. deep dive on the data, architecture, and training techniques used to create Krea 2. krea.ai/blog/krea-2-technica…

6

10

179

40,618

Diego Rodriguez reposted

NYRE

@sleenyre

Jun 30

In the magnet link version of K2, I embedded in a small easter egg in the linear layer which doesn't affect the output of the model. Kudos to all the people who have found this!

1

3

38

1,741

Diego Rodriguez

@asciidiego

Jun 27

CG

@cgtwts

Jun 27

Anthropic CEO: “open source is kind of a distraction you still can’t really see what’s happening inside the model, so it’s not truly “free.” People should be paying a lot more attention to what Anthropic is doing.

1:41

1

9

1,728

Diego Rodriguez

@asciidiego

Jun 26

trying to seem “cool” has unironically held me back way more than being inefficient or being ignorant inefficient and ignorant people at least get things done - and that’s the first step to doing things faster

8

318

Diego Rodriguez

@asciidiego

Jun 25

yall keep crying about anthropic fable 5 while some of us are rebuilding america inside @AmazonAlexa on prime day 👊

1

12

1,040

Diego Rodriguez reposted

Naga/Abhi

@NagaSaiAbhinay

Jun 23

retroanime lora for Krea 2: huggingface.co/krea/Krea-2-L…

7

4

60

2,326

Diego Rodriguez reposted

rob - comfyui

@hellorob

Jun 25

Krea 2 trained a style I couldn't get into any other model. Midjourney included. It's about time an open source model got there. I've been sitting on this dataset forever. It's of full-page plates from an early-1900s book of watercolor illustrations (via Public Domain Review). The texture and faded tonal gradients with soft bleeding edges tripped up every model I tried... none could get the aesthetic right with decent prompt adherence. Krea 2 nailed it.

30

33

634

67,042

Diego Rodriguez reposted

gokaygokay

@gokayfem

Jun 25

playing with krea 2 turbo all day, i feel like im inside the heads of the old masters

5

4

100

4,314

Diego Rodriguez reposted

Gorkem Yurtseven

@gorkem

Jun 24

new krea model is incredible at lora training

Gorkem Yurtseven

@gorkem

Jun 24

"lora training is back on the menu" @gokayfem

3

52

7,429

Diego Rodriguez reposted

ilker

@ailker

Jun 24

When I first tried Flux Kontext, I had a feeling it was going to be a major breakthrough for image editing. I also said that Klein 4B/9B LoRAs would take it to the next level. Now I can confidently say that Krea 2 is the closest thing I've tried to MJ when it comes to training style LoRAs. I'll be sharing a few examples soon.

fal

@fal

Jun 23

Krea 2 Open-Source is now available on fal The most aesthetic open-source image model, released with open weights Turbo delivers fast, polished text-to-image at low step counts RAW serves as the malleable base checkpoint for LoRA training

0:59

5

8

194

15,134

Diego Rodriguez reposted

LMSYS Org

@lmsysorg

Jun 23

🎉 Meet Krea 2 from @krea_ai, an aesthetic open-source image model ranked #1 text-to-image from an independent lab on @ArtificialAnlys. Day-0 support is now live in SGLang! Krea 2 ships as two models built to work together: 1️⃣ RAW: undistilled base checkpoint: diverse & malleable, made for fine-tuning, post-training & LoRA 2️⃣ Turbo: 8-step distilled checkpoint: fast, high-quality text-to-image Train LoRAs on RAW, run them on Turbo: base for training, Turbo for fast inference on your own hardware Run it now with SGLang!

0:58

Krea

@krea_ai

Jun 23

today, we release the open weights of Krea 2. welcome Krea 2 Raw and Krea 2 Turbo, an undistilled model from mid-training meant to be fine-tuned, and a fast distilled version with a wide aesthetic diversity. read the details below 👇

0:24

2

18

49

10,915

Diego Rodriguez reposted

Design Arena

@Designarena

Jun 22

With an input cost of $0.015, Krea 2 Turbo establishes a new Pareto frontier in Image Preference vs. Price, alongside Krea 2 Medium, GPT Image models by @OpenAI, and Z-Image Turbo by @Alibaba_Qwen.

1

2

14

2,394

Diego Rodriguez reposted

gokaygokay

@gokayfem

Jun 23

i am a krea 2 addict

fal

@fal

Jun 23

Krea 2 Open-Source is now available on fal The most aesthetic open-source image model, released with open weights Turbo delivers fast, polished text-to-image at low step counts RAW serves as the malleable base checkpoint for LoRA training

0:59

11

5

116

10,333

Diego Rodriguez reposted

青龍聖者

@bdsqlsz

Jun 23

Compared to open model weights, I am even more impressed by Krea2's detailed technical report. They generously shared all the techniques they used and the training details. Thank Krea team! I will talk about the parts I think are valuable.↓

Krea

@krea_ai

Jun 23

today, we release the open weights of Krea 2. welcome Krea 2 Raw and Krea 2 Turbo, an undistilled model from mid-training meant to be fine-tuned, and a fast distilled version with a wide aesthetic diversity. read the details below 👇

0:24

5

17

165

17,633

Diego Rodriguez reposted

NYRE

@sleenyre

Jun 23

Spend a lot of time writing this over the past week. If you want juicy details on our training stack and infra. Go read it! krea.ai/blog/krea-2-technica…

Krea 2 Technical Report

Technical report for Krea 2 (K2), our open-weights text-to-image foundation models for creative exploration. Release includes K2 Raw and K2 Turbo open weights.

krea.ai

9

28

163

20,500

Diego Rodriguez

@asciidiego

Jun 23

it's a joke people -- the underlying principle is correct

2

147

22,629

Diego Rodriguez

@asciidiego

Jun 23

remember the guy who used to write tiny c files and train gpt-2 without a gpu? he's shilling slack integrations now

Andrej Karpathy

@karpathy

Jun 23

This is a new paradigm for interacting with Claude that is significantly more "inline" with all the other human activity org-wide. Once you do all of the under the hood engineering work to make this "just work" (e.g. across tools, integrations, compute environments, memory, security, etc.), Claude basically joins the team in a seamless way - you can talk to it as you would talk to a person and it can help with a very large variety of workloads. Imo this is the 3rd major redesign of LLM UIUX. The first paradigm was that the LLM is a website you go to, the second was that it is an app you download to your computer. This third one is that it is a self-contained, persistent, asynchronous entity with org-wide tools and context, working alongside teams of humans. It really takes a while to wrap your head around it, but it works and it is awesome.

70

110

4,823

431,595

Diego Rodriguez reposted

Wildminder

@wildmindai

Jun 23

Krea-2 is now in ComfyUI. Already looks good. - 4K in 50 seconds - Clean details and textures - Realistic output - Strong prompt following All Base/Turbo quants are already available (fp8, mxfp8, nvfp4, int8, convrot-int8) huggingface.co/Winnougan/Kre…

4

32

301

18,560

Diego Rodriguez reposted

fal

@fal

Jun 23

Krea 2 Open-Source is now available on fal The most aesthetic open-source image model, released with open weights Turbo delivers fast, polished text-to-image at low step counts RAW serves as the malleable base checkpoint for LoRA training

0:59

6

21

186

46,842