SKB retweetledi
SKB
1.4K posts
SKB
@skbzz7
the cake is a lie https://t.co/Q0Au1TdIED https://t.co/xc4GP93ZIv
Katılım Ağustos 2021
264 Takip Edilen682 Takipçiler
Stupid question: what if real art starts when we stop writing prettier prompts and start messing with the activations until the model hallucinates on purpose? I don’t mean the hallucination after it becomes an output. I mean the weird internal moment before the model remembers how to behave. I’d write more, but my English ran out.
Anthropic@AnthropicAI
New Anthropic research: A global workspace in language models. Of everything happening in your brain right now, only a tiny fraction is consciously accessible—thoughts you can describe, hold in mind, and reason with. We found a strikingly similar divide inside Claude.
English
@creator_kachun Not yet. I've got quite a few nodes waiting to be cleaned up and released.
English
Nothing special, just a small radio player for ComfyUI.
Supports YouTube/live links and direct radio streams. That’s it.
github.com/SKBv0/ComfyUI_…
English
Dropped GreyZone. For Krea 2.
Just that strange early 3D atmosphere.
Not clean, not modern, not pretty in the normal way. Just old 3D mood and weird places.
civitai.com/models/2735938…
SKB@skbzz7
Also, here’s a little sneak peek at the next LoRa I’m working on.
English
@icantevenfilms It’s more of a sequencing/testing toolkit for automating the settings you’d normally have to change manually every run.
English
@skbulous Will this let me use two Lora characters in the same scene. Can never get that to work properly. One always starts to morph into the other
English
SKB retweetledi
🎨 I trained 999 Style LoRAs with the @fal Krea 2 Trainer.
Every single image in this video is a different LoRA.
That's 999 unique Style LoRAs, each trained for just 100 steps at 0.00035 LR, taking an average of ~30 seconds to train.
As I said before, @krea_ai 2 continues to prove that it's an incredible model for open-source Style LoRAs. Next week I'll be sharing even more LoRAs, along with a few surprises.
English
Dropped V0id for Krea 2.
Krea 2 picks up multiple style signals in a single LoRA much more strongly than Flux.
It took the dataset more seriously than I expected, which is great honestly; the model is doing the right thing, my data just needs a different approach. For v1 I had to rebalance the atmosphere/form mix, because unlike the Flux version, I can’t push the atmosphere as hard here without pulling the forms too much. So yeah, v1 is the slightly calmer one. Already reworking the data from scratch for v2.
Link: civitai.com/models/2559636…
SKB@skbzz7
Dropped a new LoRa for Flux Klein 9b: V0id. Made it mostly for my own taste. It makes subtle stylistic adjustments to areas like lighting, shadows, blur, flare, and overall atmosphere. Link 👇
English
SKB retweetledi
I’m very excited about this article from @OpenAI on my attempt to use GPT-5 Pro to understand the results of an experiment we did several years ago. Remarkably, GPT-5 Pro figured out the key mechanisms, as explained in the article & infographic below! openai.com/index/gpt-5-im…
English
It also seems surprisingly good at handling multiple styles in one LoRA. I’ve started training my v0id LoRA for Krea 2 too. It still needs more testing, but I’ll probably release it soon.
Ostris@ostrisai
Ostris AI Toolkit has day 0 support for training LoRAs on Krea 2. The model has beautiful aesthetics and finetunes very well. 🧵
English
Nobody needed this, which is usually my strongest product signal.
Bob Ross Preview for ComfyUI.
Live preview, but happier.
github.com/SKBv0/ComfyUI_…
SKB@skbzz7
There are no mistakes, just happy accidents.
English
