Degeniv

If you're interested in this kind of thing, I'd recommend picking a lane for it first before getting into the details. It has a UI made by the community, it has a Gradio UI, OR it can be loaded into ComfyUI as nodes. Comfy's the most powerful, but the more complex.

Ace-Step 1.5 effectively gives you cloud-based payment model music AI diffusion quality, completely locally and completely free, pretty neat. Any AI tinkerers or creatives that don't wanna pay cost to screw around, there's an option now. No need for SUNO github.com/ace-step/ACE-S…

@stephane @LFabien Tbh as someone that's used agentic workflow stuff extensively, if I had to remember which agent did which thing based on actual names I'd lose my mind. You would need an index for agent name to task, or ask them what their role is. It would be fine for a few you can keep track of

As I was building an app around agents with @LFabien, I wondered whether we should stay generic (Design, PM, etc.) or give our agents names. I did some research, and here's what I found.

Now for a lighter note: any time a series, game, anime, movie, story, whatever, kills a child, it's generally instantly a 7/10 at the very least, I challenge you to find any examples proving otherwise. I think it's a pretty strong metric for audience score.

Just ask first: Is the cat already being used in this way.

If you can code yourself, or have the funds to work on more complex projects and pay costs for your solution you have a tangible benefit there. You have an advantage in being able to use things like X's own API.

My grandmother loved the book "101 Uses for a Dead Cat" by Simon Bond, and I've just been thinking about how many simple problems people are still having and they still don't realize you can just go to grok, chatgpt, claude, whatever, and just ask.

Sweaty Faye Valentine..

I'm trying to focus more on scenario poses that are interesting, rather than pinups/straight to the point. I don't mind doing both, but I find coming up with more unique scenarios better, personally. It's like the difference between implicit and raw explicit. Both have draw.

Orihime being surprised out of reading her book.

Some Yukari Takeba I used as a test run for newer models, prompt adherence and some post-processing. It'll be nice when AI models can keep subject details cleaner from further away naturally, most generations still need to be center focused.

Guys will see an eldritch lovecraftian abomination waifu and still say "Would"

I tried recreating old AI art of this spider-esque waifu to current-level advancements. What a shift for degeneracy.