Welcome, humans.

We go live in five minutes.

Want to make better AI videos but have absolutely no idea what you’re doing?

Perfect. This livestream is built for TOTAL beginners.

In five minutes, we’re going live with Daniel Berkovitz and Alon Yaar, LTX’s VP of Product, for a beginner-friendly lesson on video prompting plus a hands-on demo of the brand-new LTX-2.5.

We’ll start with the model itself: what changed in LTX-2.5, and where teams are already using this kind of video generation in studios, robotics, and real-time applications.

Then we’ll get practical. Daniel and Alon will walk through how to describe a scene so a video model understands the shot, subject, motion, and action you actually want.

After that, we’ll put LTX-2.5 to work live and leave the back half open for questions and audience demos.

Here’s what we’re covering live:

• What’s new in LTX-2.5: Sharper visuals, cleaner motion, synchronized audio, and native multishot generation.

• Video prompting 101: Why prompting video is different from prompting an image.

• Your first good prompt: How to describe a scene without writing a screenplay.

• Controlling the shot: How to steer subjects, camera movement, action, and scene changes.

• Live LTX-2.5 demo: Daniel and Alon will build with the new model in real time.

• Your questions: We’ll tackle Q&A and audience demos in the back half.

No previous AI video experience required. If you can describe the video you want to make, we’ll show you how to turn that idea into a much better prompt.

LTX-2.5 is LTX’s newest open-weights video model, built for film, real-time generation, and physical AI workflows, with weights available on Hugging Face.

You’ll also get the LTX team’s view of what video models understand today, where they still break, and how people are using the technology outside of one-off clips.

So yes, this is a TOTAL beginners class. You do not need a film degree, a ComfyUI flowchart, or seventeen tabs of prompt theory open beside you.

Open the stream now, say hello in the chat, and bring the video idea you’ve been trying to make actually work.

We’ll introduce LTX-2.5 first, teach the prompting fundamentals next, then get into the live demo and your questions.

Learn more about LTX-2.5: Model page | Announcement | Hugging Face

P.S: If you arrive late, the same YouTube link will have the full replay afterward.

New from The Neuron: AI can write DNA now

Radical Numerics CEO Eric Nguyen helped build Evo, a genomic AI that treats DNA like language. Researchers have used it to design CRISPR systems and generate a complete bacteriophage genome.

The part that stuck with us was his warning about “deepfake viruses”: as biological AI gets better at generating plausible sequences, sequence-matching defenses may become less reliable. The same AI designing new biology may eventually be needed to help spot and defend against these new threats.

Watch / listen: YouTube | Spotify | Apple Podcasts

New from The Neuron: We could have Fable level AI on our laptops “within two years”

Intel’s Dr. Olena Zhu showed us the version of local AI that feels useful right now. They just released SuperClaw’s public beta, which includes email, coding, and deep-research agents that can combine outside research with private company data while keeping the sensitive material local. Parts of the stack are open, so teams can inspect the architecture and build on it.

The bigger idea behind Intel’s approach is wild: local models have been inheriting frontier-level capability after an average of roughly 24.8 months. If that trend holds, a Fable-class intelligence model could run on a high-end laptop by 2028. See above re: problems with datacenters for why that’s actually really important.

Watch / listen: YouTube | Spotify | Apple Podcasts

Shout out to Nerdio who sponsored this video!

What should we learn next? 🤔

Lets us know below!

Stay curious,

The Neuron Team

P.P.S: Love the newsletter, but don’t want these podcast and livestream announcement emails? Don’t unsubscribe. Adjust your preferences to opt out of them here instead.