Pinned Tweet
Introducing LTX-2.5, the world model the world builds on.
One of the biggest upgrades yet to the model already powering film, robotics, and real-time workflows. Higher pixel fidelity, multishot scenes that hold together across cuts, and a pretrained foundation built to be finetuned across domains.
Plus: Diffusion Fidelity Rendering, a new rendering approach built for pixel quality that holds up frame by frame, even on a cinema screen.
This isn't a tool you rent. It's a foundation you build on, open and yours.
A major LTX Desktop update is here.
Built-in LoRAs, IC-LoRAs for deblur, colorization, day-to-night and more, a batch queue, and remote monitoring from your phone.
Download LTX Desktop: ltx.io/ltx-desktop
We teamed up with @_doubleAI_ to push our already-optimized pipeline further with their Maxwell model. A 5-second 4K clip on 2xB200s: 250s → 134s.
Same quality, no quantization. The update is open source.
1/ Together with @ltx_io, we used Maxwell, our Code Execution Model, to make video generation in LTX-2.5 1.87x faster.
With Maxwell, generating a five-second 4K clip on 2xB200s went from 250 seconds to only 134. No quantization or loss of quality at all. The code is open source.
LTX is proud to serve as executive producer on Mutilator. Eric Fogel created this world as a student at NYU. Now he’s bringing it back with Wonder Studios.
Full film Oct 22.
From MTV's Celebrity Deathmatch to DreamWorks' Megamind, Eric Fogel is now returning to the world he first created as a student at NYU.
MUTILATOR.
Wonder Studios × @ltx_io. Full film coming Oct 22.
@Deathmatch_Guy
Here are five things the community built with LTX-2.5 this week. 🧵
@rapidmlx made a 30-second sci-fi short on two Macs. Qwen Image made the keyframes. LTX-2.5 generated the motion. Nothing ran in the cloud.
xitter.cf/rapidmlx/status/210966…
A robot has 1% power left. A seedling is freezing beside it.
"The Last Shift" is a 30-second sci-fi short made entirely on two Macs. No cloud.
🖼 Keyframes: Qwen Image 2.1
🎬 Motion: LTX-2.5
🔧 Pipeline: ComfyUI
⚡ Both served locally by Rapid-MLX
How we made it 👇
rapidmlx.com/blog/ai-short-f…
@adsilva264 built ReAngle, an open source re-shoot app for ComfyUI. Draw a camera path in 3D. Preview the move. LTX-2.5 renders the new angle on a 16GB GPU.
xitter.cf/adsilva264/status/2111…
ReAngle is my open-source re-shoot app for ComfyUI. It's built on LTX-2.5 with depth-based reprojection, and you can draw your own camera path in a 3D preview before anything renders. Orbit, crane, pull-back or a freeze, all on a 16 GB RTX 5080, with Comfy Cloud for bigger jobs.
Try it in your browser: tekholms-lw.github.io/ReAngl…
Code and workflows: github.com/Tekholms-LW/ReAng…
#OpenSourceHiggsfield
Open-source AI moves faster when builders have the tools to push it further.
Join the challenge on the Comfy Developer Platform for a chance to win $25K in prizes.
Higgsfield promised to open-source their entire platform. They didn’t.
Let's help them.
For the next two weeks, Comfy and our partners are putting up $25,000 in prizes for the best open-source builds on the Comfy Developer Platform.
Partners: @NVIDIAAI, @runpod, @MiniMax_AI, @ltx_io, @bfl_ai
→ $10K grand prize
→ RTX 5090: Most Innovative
→ RTX 5090: Most Viral
→ RTX 5090: Open Weights Only
→ 10 runners-up: $250 in Comfy credits
At LTX, we strive to empower partners like Asteria, who are pushing cinematic storytelling forward while keeping artists at the center of every workflow and honoring the craft of traditional filmmaking.
Great filmmaking demands control and LTX’s open weights model makes this possible; from securely training on your own IP to our full suite of professional capabilities.
Together with Asteria, we’ll keep driving the community forward, turning feedback from filmmakers like them into tangible capabilities that can be deployed in professional workflows.
Plan the camera move in 3D.
Layout to Render renders the final shot with perspective and context kept intact.
Download the LoRA here: huggingface.co/Lightricks/LT…
It is the simplest stories that tell us the most about life.
The Very Next Day: a short film that ruminates on the relationships that connect us and shows what we can accomplish when traditional filmmaking and technology work hand in hand.
Written and Directed by Paul Trillo, created by the team of 2D and 3D animators at Asteria and built on the backbone of LTX’s open weights.
Coming soon to a screen near you.
Bridging the gap between generation and reconstruction quality in diffusion models.
Meet APEX: our latest research brings them together by building structure before refining detail, helping world models make better use of fewer tokens.
Join @ZeevFarbman, our Cofounder and CEO, at #NVIDIAGTC Berlin on October 22.
He joins @bfl_ai, @vesoma_ai, and NVIDIA to discuss the future of world models.
Open world models will define physical AI. Hear why.
What will it take for AI to move beyond generating pixels and start understanding worlds?
At #NVIDIAGTC Berlin, hear from @bfl_ai, @LTX, and @vesoma_ai for a conversation about the future of world models.
Explore whether traditional scaling laws still apply, where foundation models end and robotics begins, and what the open ecosystem must standardize to accelerate progress.
🕒 October 22 @ 9 AM CEST
📅 Add to calendar: nvidia.com/en-eu/gtc/session…
An easier way to use the LTX Trainer is now inside LTX Explore.
Train a Character or Style LoRA on your own footage and bring your creative direction into new shots. No local GPU required.
Test in Explore here: app.ltx.io/loras/new
Here are three things the community built with LTX-2.5 this week. 🧵
@linoy_tsaban replaced video backgrounds with no green screen. Alpha Gen pulls the matte. H3 inpainting builds the new scene.
xitter.cf/linoy_tsaban/status/21…
Doing video background replacement with H3 Inpainting + LTX Alpha Matte LoRA, and it's surprisingly good 🔥
> LTX 2.5 LoRA to extract matte
> H3 Ref inpainting (+ Turbo LoRA) to generate a new background
> Mask drop knob to reduce flickering and improve lighting match - drops the inpainting mask at step t
I added the workflow to 🧨diffusers modular + built Spaces for it 👇
@rapidmlx previewed video continuation on Mac. Generate a clip. Describe what happens next. The model extends it locally. It ships in their next release.
xitter.cf/rapidmlx/status/210762…
Our community built something really fun: video continuation using LTX-2.5 from @ltx_io 🎬
Generate a clip you love, then tell it what happens next—and keep the story going.
Here’s a sneak peek: an anime rooftop run, extended into a leap to the next rooftop. Generated entirely locally with Rapid-MLX.
Coming in our next release. 👀