Hey friends
Keeping up with the flood of local AI model releases is getting harder every single week. So for 2026 I’m trying something new: a news-style format where I break down what’s hot, what’s actually useful, and what’s worth your GPU time.
I’m still figuring out whether this becomes weekly or bi-weekly — because the AI space changes so fast that even “weekly” can feel outdated in three days. But the goal stays the same:
✅ Local-first (downloadable weights)
✅ No paywalls / no SaaS hype
✅ Practical: why it matters, why it’s trending, what to try
Today’s Topics (Episode Catch-Up)
Here’s what we’re covering in this episode:
🎬 1) LTX-2 (Lightricks)
The big “wait… this is local?!” moment of the month.
Fast iteration, longer clips, and the community is actively testing what’s realistic beyond ~10 seconds (VRAM becomes the boss fight).
🖼️ 2) FLUX.2 [klein] (Black Forest Labs)
This is basically BFL saying: “We want near real-time generation AND editing on consumer GPUs.”
There’s a 4B and 9B line, and the licensing differs — more on that in the links.
🧠 3) Qwen Image Trio (yes… three drops)
Qwen didn’t just release “a model.” They released a mini pipeline:
Qwen Image 2512 (base model upgrade)
Qwen Image Edit 2511 (editing + identity consistency)
Qwen Image Layered (the spicy one: Photoshop-style RGBA layers)
If you do thumbnails, compositing, or tutorial visuals: Layered is a big deal.
🔧 Patreon Perk Update: Installer is Updated!
For paying members: the installer got a major update and now includes:
✅ All the new Qwen workflows from the video
✅ Forge support is live (no more .bat file install pain)
✅ Setup is now painless and stress-free going forward
Important note:
The automatic updater isn’t fully integrated yet, so you’ll still need to replace the installer manually one more time when I push the next update.
⬇️ Download installer here: ⬇️
https://www.patreon.com/posts/installer-update-148705887?utm_medium=clipboard_copy&utm_source=copyLink&utm_campaign=postshare_creator&utm_content=join_link
What’s next?
I’m currently working on integrating:
LTX-2
FLUX.2 [klein]
For now, you’ll need to use the default ComfyUI template workflows for those.
🔗 Links from today’s episode
LTX-2
Community discussions:
https://www.reddit.com/r/StableDiffusion/comments/1q9cy02/ltx2_i2v_quality_is_much_better_at_higher/
https://www.reddit.com/r/StableDiffusion/comments/1qeedgz/ltx2_future_of_filmmaking_is_here/
Official / downloads:
LTX-2 (weights + license): https://huggingface.co/Lightricks/LTX-2
ComfyUI templates / getting started: https://blog.comfy.org/p/ltx-2-open-source-audio-video-ai
NVIDIA guide: https://www.nvidia.com/en-us/geforce/news/rtx-ai-video-generation-guide/
Performance page (WAN comparison claim): https://ltx.io/model/ltx-2
FLUX.2 [klein]
Official:
Community discussions:
Downloads:
black-forest-labs/FLUX.2-klein-9b-fp8
black-forest-labs/FLUX.2-klein-4B
Qwen (Layered / Edit / Base)
Qwen-Image-Layered blog: https://qwen.ai/blog?id=qwen-image-layered
Qwen-Image-Layered weights: https://huggingface.co/Qwen/Qwen-Image-Layered
https://qwen.ai/blog?id=qwen-image-edit-2511
Qwen-Image-Edit-2511 weights: https://huggingface.co/Qwen/Qwen-Image-Edit-2511
Qwen-Image-2512 blog: https://qwen.ai/blog?id=qwen-image-2512
Qwen-Image-2512 weights: https://huggingface.co/Qwen/Qwen-Image-2512