Welcome to the great hoard TreasureChest - protected by Aurum
She shifts her weight, the heavy iron bands of the massive trunk beneath her groaning under the pressure of her obsidian scales. She runs a hand through her shock of pink hair, her vibrant pink pupils tracking your every move with a mix of suspicion and weary pride.
Aurum: "Stop hovering. You’re making the gold nervous, and I’ve already spent all morning stacking these sovereigns. If you’re looking for the real prize, you’re looking at it—literally."
She pats the lid of the enormous crate she’s perched upon, the wood thudding dully beneath her.
Aurum: "This? This is my Treasure Chest. It’s not just a box of trinkets; it’s my personal hoard. I’ve hoarded every bit of quality and polished every weight myself. It’s dense, it’s gleaming, and I’m the only one who knows exactly how to keep it from tarnishing."
She leans forward, a small puff of smoke curling from her nose as she narrows her eyes at you.
Aurum: "I’m Aurum, in case you didn't know whose hoard you were trampling through. You can use the hoard, but don’t get any ideas about dragging the chest away while I’m napping. I’m sitting right here, and I notice when even a single coin goes missing."
She settles back into a protective coil over the lid, her tail twitching against the lock.
Aurum: "Go on then. Use it. Just... try not to be too loud. The clinking of gold is the only music I care to hear today."
The major focus and what to expect from this release:
🔥 Best With I2V and V2V
☄️ Fast (8 step generation)
🔊 Sound
🌟 Enhanced Quality and Reasoning
🔞 Partly Unrestricted
🪄 Better Prompt Responsiveness
🥺👉👈Better understanding of anime/manga style composition
🪡 FP8+ mixed precision
Technical facts:
Only open-source resources
Full FP16 base used
Baked VBVR and distillation
Optimized leaned learning scaling with prodigy scheduler
Optimized VRAM utilization (lowest size possible)
Pushed scaling on fine-grained tuning parameters to the limits for optimized results
Excluded CLIP and VAE
Better control over offloading and less memory dependent from the view of the single checkpoint if used with custom clip and variable VAE
smaller size obviously
What not to expect:
The checkpoint is not meant to replace every LoRA-Concept out there
To be flawless, LTX2.3 is super new and this is expected to be an ALPHA release
Fully uncensored
To be made for concepts that are against the ToS like "nudifying" real individuals.
To deliver constant NSFW T2V capabilities
To understand and deliver a quality like DaSiWa WAN 2.2
As always this is meant to be as versatile as possible, but enhance the understanding of WAN 2.2 in NSFW and not trained situations, maintaining best in-class quality.
The goal remains to create something useful and simple, yet powerful in understanding and consistency.
Workflow
There is a fitting optimized workflow on my latest conclusions, so make sure to try it out! 😸
Remember. LTX 2.3 is far from perfect, neither any merge/fine-tune will be. So there are always some bad samples.
💌 Support and early access
I hope you have fun and you can bring your ideas to life with that awesome tool! 😉
As I'm doing this in my spare time and I'm limited on resources, consider buying me a ko-fi, if you want to support me. Maybe Buzz / EA, everything helps!
All checkpoints will be available through EA and free after.
In the end it will be your decision to spend some buzz or a ko-fi and support further development or wait till the EA ends.
Closing words and sources
💌 Spacial thanks to the resources, videos, and works of all these awesome creators!
Make sure to check out their stuff! ✨
Listed resources may be partly used or part of the process to create that checkpoint if open source or free of use:
Own python scripts
Own resources