🔥A little behind-the-scenes — and what's coming next
Hey everyone, (LINK AT THE BOTTOM)
You've probably noticed my posting has slowed over the past five or six weeks, and I want to be straight with you about why — because it isn't really a slowdown at all. It's been the opposite. I've been working harder on this brand than at any point since I started, just not on the part you can see yet.
Here's what's actually been happening behind the curtain:
- Learned the new **LTX 2.3** animation model inside and out — it's the first model with native audio, and it changes what's possible. Weeks of testing to find where it shines and where it falls apart.
- Rebuilt my **WAN 2.2** pipeline — the workhorse behind most of my longer animations. Cleaner motion, better physics, fewer weird hand moments.
- Developed a hybrid approach that blends both — best motion from one, audio from the other. Slow to produce, but the results are worth it.
- Built a fast **face-LoRA system** so any character I introduce can be locked in and reused with real consistency. Jelena was the proof — she's been so heavy in my feed because nailing her face was the unlock for everything else. Now I can do that for any face I want, quickly.
- Brought in **Grok Imagine** for camera-driven shots I couldn't get out of my old toolkit.
- Tested **Z-Image Turbo** for cleaner photoreal stills and fashion-style sets.
- Built my own **automated posting and delivery pipeline** across Discord, Twitter, and DeviantArt so the publishing side stops eating into the creating side.
- Produced thousands of test images and animations along the way so I'd actually know what works before I commit a full set to you.
- Started laying the groundwork for what's next — **persistent characters in a persistent world.** A real recurring cast. Stories that connect across sets. More on that soon.
The point is — the R&D part of my year is done. I know what works now, what doesn't, and how to ship at a much faster pace going forward. The next few weeks will start showing it.
To celebrate, I'm putting together a **work-in-progress drop** — a folder of unpolished stills and animation tests from the past six weeks so you can see what I've actually been building. I'd love your feedback on it. The next era of this brand is being shaped right now and I want your eyes on it.
Also — please come hang out in the **Discord.** It's been quiet and I want to fix that. I'm going to be a lot more open about my process in there going forward, and I'd love to hear what you want to see more of. If you've got questions about how any of this gets made, drop them in the channel — I'll be an open book.
Honestly, I believe the work I make for this brand is the best work I'm making anywhere. The realism, the direction, the world I'm building — this is where my passion lives, and I'm doubling down on it. Thank you for sticking with me through the quiet stretch.
Get ready for some amazing artwork.
— Del
PS - LINK TO ALL WIP IMAGES,ANIMATIONS, and character FACES you're going to start seeing in the coming weeks. You'll see in here what does work, and what does not work! Many warped images and fails in here. I'm sharing this because the ZIT model is super incredible, but it really sucks at NSFW. I got some images to look good, but it's not adequate to run with, which is very disappointing... But the SFW images are still worth making and enjoying IMO:) Enjoy some of the first images of Carly.
PPS - To Clarify - The character 'faces' folders are the samples produced from my Flux Face LORA creations. IE - I trained a model on those faces, the associated body/image is just flux putting the face on someone. Each face was curated by me using nano banana pro, and a script to generate images of that AI face in different lightings/expressions. Then i ran a lora training using my GPU to make a file I can use to inpaint that face on anything.
IE = Really unique characters, strong consistency, way better images and animations. ASTRID did NOT use a LORA. Carly DID use a lora. See the difference;)