The most interesting thing about LTX-2 ↗ isn’t what it can do. It’s what Lightricks decided to give away.

In a space where OpenAI locks Sora behind a subscription, Google keeps Veo in a walled garden, and Runway has no public API at all. Lightricks shipped the full model weights. Apache 2.0. Free to use, fine-tune, run locally, build on top of. They published the training code, the inference code, the LoRA trainer, the ComfyUI nodes. Everything.

This is the Linux play for AI video. And I don’t think enough people are paying attention to it.

I’ve spent years watching the AI video space evolve from embarrassing 4-second loops into something that makes cinematographers nervous. I’ve played with most of the major models, Sora, Veo, Runway, Kling. They’re all impressive. They’re all locked. LTX-2 is the first model that makes me feel like the power is shifting somewhere more interesting.

What LTX-2 actually is

Think of LTX-2 as two models fused into one body. There’s a video brain (14 billion parameters) and an audio brain (5 billion parameters), and they talk to each other at every single step of generation. Not sequentially, not one slapped on after the other. The audio and video co-evolve during the diffusion process, which is why the lip-sync actually holds and why the foley sounds like it belongs to the image rather than being randomly assigned to it.

That’s not how most AI video works. Most pipelines generate video first, then add audio as an afterthought. LTX-2 bakes them together at the architectural level.

The other specs: 4K native output at up to 50fps, clips up to 20 seconds, trained on licensed data from Getty Images and Shutterstock (more on why that matters), and it runs on a consumer GPU. An RTX 3060 with 8GB of VRAM. Not an H100, not a data center, the card a student might have in their gaming PC.

LTX-2.3 and what just shipped

Three weeks ago, LTX-2.3 ↗ dropped alongside something that stopped me mid-scroll: LTX Desktop, a free, local, fully open-source video editor built on the same weights as their cloud platform.

No per-generation cost. No internet required after setup. Same model.

I’ve been messing around with the desktop app on my machine, and it’s genuinely usable in a way I didn’t expect from a beta. The LTX-2.3 upgrade itself also fixed the most annoying thing about the previous version, that Ken Burns effect where image-to-video would just slowly pan across your input frame and call it “animation.” Now it actually moves. The prompt adherence is tighter, the hair detail (historically a disaster in AI video) is noticeably better, and they added native portrait video for the first time, trained on actual vertical content, not landscape crops rotated sideways.

For TikTok, Reels, Shorts creators: this matters.

The platform play

Here’s where it gets strategically interesting, and where I think Lightricks is playing a longer game than most people realize.

LTX Studio ↗ isn’t just their own model wrapped in a nice UI. It’s a production platform that integrates Kling 3.0 Pro, Google Veo 3.1, Runway Gen-4.5, Flux, and a handful of image models alongside LTX-2. You pick your model per shot depending on what you need.

They’re not betting you’ll always want LTX-2. They’re betting you’ll want to work in their environment, from script to storyboard to generation to timeline editor to export. The model is the hook. The platform is the product.

The timeline editor exports .xml directly to Premiere Pro or DaVinci Resolve. The audio-to-video feature (launched with ElevenLabs in January) lets you upload a track and have the video driven by the sound. Timing, pacing, camera motion, all responding to the audio waveform. There’s an enterprise brand kit that keeps visual consistency across a whole team’s output.

As a designer and builder, the product craft here impresses me. These aren’t bolted-on features. They’re telling a coherent story about where the workflow lives.

The honest part

LTX-2 is ranked #36 overall on the Artificial Analysis leaderboard ↗. That’s not a typo. Kling, Runway, Veo, even a handful of models I’d never heard of before, beat it on pure video quality in blind evaluation.

The visual realism gap is real. Long-form temporal stability. Keeping a scene coherent beyond 8 seconds, is still a work in progress. If you need the most photorealistic output available right now, Kling 3.0 is probably your answer.

But here’s the thing: LTX-2 Pro costs $3.60 per minute of generated video. Sora 2 Pro costs $30.00.

That’s not a small gap. That’s an order of magnitude. And Sora doesn’t run locally, doesn’t give you the model weights, doesn’t let you fine-tune it on your brand’s visual identity, and can’t be deployed in an environment where your footage never leaves your building.

The licensed training data Getty Images and Shutterstock ↗ matters enormously for enterprise buyers. It’s the reason a Fortune 500 legal team might actually sign off on using this, while they’d never touch a model trained on scraped data of unknown provenance.

Five million downloads since the January open-weight release isn’t hype. That’s a community building on top of something. The LoRA ecosystem, the ComfyUI integrations, the speed optimizations, the community has already done meaningful work with this model in under three months.

What I actually think this means

The AI video market is on track to go from $32 billion to $133 billion by 2030. Every major player is sprinting. The walled gardens will keep getting better. Sora will keep improving. Veo will keep improving. Runway will keep raising money and building.

But there’s a version of this future where a fully open model, with a thriving developer ecosystem, a clean enterprise licensing story, and a platform that abstracts away which model you’re actually using, ends up being where most professional video work happens. Not because it’s the best model today, but because the developer community keeps making it faster, better, and more specialized. Because fine-tuned variants emerge that outperform the base model on specific tasks. Because it’s running on your hardware, not theirs.

That’s the Linux trajectory. Linux wasn’t the prettiest operating system in 2001 either.

I’m not saying LTX-2 wins. I’m saying the open-source bet is coherent and strategic, and Lightricks. A profitable company with $250M+ in annual mobile revenue and no existential funding pressure, is well-positioned to play a long game here. That’s a different situation than most open-source AI projects, which run on VC patience and hope.

If you haven’t tried it yet check out the LTX Desktop app ↗ is free, and it’s the fastest way to form your own opinion. And if you want the cloud version with access to every model, LTX Studio ↗ has a free tier worth poking around.

Let me know what you make of it. I’m genuinely curious what others in this community are seeing in their own testing.

Linus Ekenstam
31 March 2026

Almost 10 years ago now I first met David Okuniev — the then founder of the now legendary data collection company Typeform. I was new in Barcelona, had just sold my company, and made a soul searching road trip half way across the globe.

David, a man, obsessed with details, took me in, you see I was the first product designer to join the Typeform team back then. I learnt a lot working with David. How he tilts his head when looking at designs. How he relentlessly want to improve, built and contribute. He is by heart an IC.

Fast forward to today, and I have the honour to call him a friend. I would be lying if I did not mention he is a bit odd, but aren’t all founders a bit loose?

When stepping down from day-to-day ops at Typeform, David let me in on what him and Neil Kinnish was going to do. They had raised a bit of money, and had their target set on Screen Recordings.

Now after 6 months in stealth, they released to the world - supercut.video Loom I’m sorry but I’m now an EX-loom user, and I’m now all in on Supercut.

I’m not sure if you use Product Hunt, but if you do, go give them some love and share your thoughts directly with the team there.

Upvote on Product Hunt

Now this is a space I’m personally very attracted to. For one, I record my screen quite a lot, secondly, communication between individuals when using screen recording greatly improves. It’s a very powerful tool, that will make you stand out. Supercut is built with speed and quality from day 0.

So it does not matter if you are making discardable internal updates, or if you are making customer walkthroughs or sales pitches. Supercut handles all these cases with ease.

Another thing that really makes me bullish on Supercut, is that they are doing so much leg work behind the scenes. Someone or something crashing your screen recording party with a loud noise? Easy out of the box they clean the audio. Need chapters? Yeah they sort that too. Captions, yes. Download source video? Yes.

You see, maybe one of the most overlooked features, is that it just works. Speed is not an afterthought, its a core principle, as is quality and design.

I’m also very keen to see the team execute on future AI features, where instead of flashing this as AI powered screen recording. They simply use AI everywhere it makes sense and enhances the experience for the end customer, the supercut user and everyone between.

My final verdict is that Supercut is a tool that I’m almost certainly think you’ll love. It’s built by my friends, old colleagues and I know they have sunk a lot of their time into making this just right.

Go show them some love on Product Hunt, go try out Supercut for yourself, it’s free during their Open Beta.

Let me know what you like/dislike about screen recording tools.

Someone once said they would broker peace in two days if elected.
We all knew that was a sham.

Now that same person is pointing fingers, shifting blame, claiming others don’t want peace—anything to distract from the fact that their promise was a fantasy. This moment reveals exactly what kind of leader they are when faced with real, complex global problems: one who postures, deflects, and ultimately, fails.

It’s easy to wave the dictator’s pen around and sign executive orders. Don’t be fooled—EOs are just a modern shortcut to ruling by decree, a tool of convenience rather than governance. It doesn’t matter who uses them. They are a crutch for those who can’t negotiate, can’t build consensus, and can’t lead.

Now, if we step back—way back, from where I usually sit—this whole situation looks like a sharp thorn wedged in the world’s side. Everyone talks big. Plenty of hand-waving, grandstanding, and clever wordplay. But let’s not forget the fundamental truth: one country invaded another, with the explicit goal of wiping it off the map. That is the singular, undeniable cause of this war.

If the world really wants peace, the answer is painfully simple.

The invader leaves. Withdraw the troops. Sign a peace treaty. That’s it. No long negotiations, no half-measures, no spin. Anything less sets a catastrophic precedent—one that tells war-hungry governments everywhere that land grabs are back in style, that it’s “okay” in 2025 to redraw borders by force.

We are living through one of the most transformative periods in human history. AI is rewriting economies. Fusion energy is within reach. Space exploration is no longer just a billionaire’s playground. We are standing on the edge of an era that could redefine humanity—and yet, we remain trapped in the same dumb cycle of war, tribalism, and self-inflicted division.

It’s not just the war itself. It’s the way the world is reacting to it. The West, once defined by its ability to argue fiercely but respect boundaries, has become a global laughingstock—a schoolyard of name-calling and kindergarten theatrics. Worse, this brand of polarization is spreading, deliberately exported by those who want to fracture alliances, pit allies against each other, and create a world where no one trusts anyone.

Some people have reached out, applauding me for standing up for what I believe in. Others are angry, thinking I have something against their country. The truth? All of this sucks. Every side is losing in different ways, and no one is thinking ahead.

This isn’t a Cold War sequel. Or at least, it shouldn’t be. The worst mistake we could make right now is locking ourselves into opposing camps, shutting down trade, retreating inward. Yes, it might bring short-term security, but history is clear: isolationism doesn’t solve problems—it delays them, creating messes that take decades to clean up.

Each region of the world has its own reckoning ahead. The U.S. needs to stop running the world’s biggest payday loan scheme and get its $35 trillion deficit under control. Europe needs to stop treating security as an afterthought and start acting like a real geopolitical force—maybe even bring the UK back into the fold, because whether the paperwork says so or not, the UK is Europe.

And then there’s the East. It’s on track to dominate the next century, and that’s fine. Empires rise, empires fall. The torch gets passed. But what matters is how a nation carries its power. Will it lead in a way that builds bridges, creates trade, and advances civilization? Or will it repeat the same tired playbook of conquest and control?

Because in the end, it truly doesn’t matter who holds the title of “most powerful nation.” Not in the long run. Zoom out far enough, and you’ll see that we have way bigger fish to fry: artificial superintelligence, climate collapse, asteroid deflection, keeping this planet habitable for another thousand years. And yet, here we are, still bickering over borders and burning time fighting battles we should have outgrown centuries ago.

War is not the way. Free markets always win.
Taking care of humanity matters more than who holds the torch.

And if we can’t figure that out now, history will make us learn it the hard way. Again.

Linus Ekenstam
4 March 2025

First, thank you for your support, secondly I really need some help from you.

About 2 hours ago I was targeted by an Phishing Attack, and I’ve been sick for weeks, tired, fever, and everything in between. So I feel for it, I feel terrible.

The result is that someone took control over my x account @linusekenstam This is not great. It’s terrible.

The ask

I posted this video below on the Inside My Head X account, and I need your help to share it.

The link:
https://x.com/InsideMyHeadAI/status/1895061300796920168

Please, I never ask for much, but it would mean the world if you could help me with this. I’d be forever grateful.

Sincerely
Linus Ekenstam

👋 Hello! I’m Linus, and welcome to this 🔓 open edition 🔓 of my weekly newsletter. Each week, I delve into all things AI, answering your questions, writing op-eds, and creating AI tutorials.

This weeks newsletter is sponsored by Rendernet.ai – The absolute best place to create AI generated characters with consistency and deep control.

The TLDR;

Robot Names are linked to videos of each robot, and company names are linked to corresponding websites. To save you a bunch of time.

1. Atlas (new version) - Boston Dynamics

2. Figure 01 - Figure Robotics

3. Optimus - Tesla

4. CL-1 - LimX Dynamics

5. Kepler - Kepler Robotics

6. Digit - Agility Robotics

7. Apollo - Apptronik

8. Walker S - UBtech

9. Ameca - Engineered Arts

10. H1 - Unitree Robotics

11. GR-1 - Fourier Robots

12. MagicBot - Magic Labs

13. NEO - 1x technologies

14. Phoenix - SanctuaryAI

15. Aloha - Stanford University

16. Menteebot - Mentee Robotics

I hope you enjoy spending some time here and get a good baseline of whats currently going on in the land of (humanoid) robotics.

The Age of (humanoid) Robots

By no stretch of the imagination is this a new field. We have all grown accustom too men with hockey sticks trying to beat robots. This has all literally gone on for decades. I vividly remember how extremely amazed I was about the ASIMO robot from Honda a little over 13 years ago. But ASIMO is older than that. It started as a research project back in 1986.

To say that Honda was early to the humanoid game is an understatement. Actually it’s insane that the robotics industry seemingly have been standing still since ASIMO got unveiled 13 years ago.

When you look at the early videos of him running, walking, interacting, it was already super impressive back then. Not downplaying any of the current companies, but the somewhat older me is a bit disappointed in how long it has taken us to get here.

Today, with the wave of AI washing all over all faucets of our world, humanoid robots are getting better and better at understanding the world around them, interact with us humans in a more natural way and more importantly getting cheaper and faster to manufacture at scale. We’re simply at the precipice of something absolutely gigantic in proportions compared to anything that came before it.

A few days ago Boston Dynamics said goodbye to the Atlas HD robot, and thanked it for its services. Then just 1 day later they unveiled the new Atlas robot, and its one of a kind, omnidirectional upper torso, legs and head. If you playback the video below a few times, you can see that not only can it turn its torso 180 degrees, it can also invert its legs, and point them backwards or forwards, making it basically omni-directional with far more degrees of freedom than a human ever could.

It will be interesting to keep tabs on the new Atlas, and see if it sports the same abilities when it comes to parkour and lifting heavy gear as its predecessor.

Tesla Optimus

We also just got a new video from Tesla showcasing some of what Optimus has been up too. Showed it to my 4 year old, and she asked why it was moving so slow 😉 – the younger generation will not thing this is odd at all. For them this is as normal as television is to many of us.

What are you most excited about when it comes to robotics and more so even humanoids? Let me know in the comments or just reply to this email.

Startup Dad Pod

I was recently on the startup dad pod, where I spent an hour talking about male fertility, time delusion, the gameboy and artificial intelligence. Enjoy.

Still here?

If you are still here, thank you. If you have questions, feel free to reply to this email or reach out to me on X

If you read this in the substack app, please leave a like or a comment below it let me know if you enjoyed this or not.

Last but not least, thanks again to this weeks sponsor - rendernet.ai

👋 Hello! I’m Linus, and welcome to this 🔓 open edition 🔓 of my weekly newsletter. Each week, I delve into all things AI, answering your questions, writing op-eds, and creating AI tutorials.

First I’d like to thank this week’s sponsor, Ionic Commerce - Please check them out, it’s because of sponsors like them I can keep this newsletter going.

Monetize your AI Assistants

Ionic is the infrastructure that enables AI companies to embed magical shopping experiences within their chat agents. With Ionic, you can provide value to your users through product recommendations and you can monetize through sales commission when users make purchases

👉 Unlock E-commerce

SORA - A World model

Sora is an AI model that can create realistic and imaginative scenes from text instructions.

OpenAI is teaching AI to understand and simulate the physical world in motion, with the goal of training models that help people solve problems that require real-world interaction.

Sora is able to generate complex scenes with multiple characters, specific types of motion, and accurate details of the subject and background. The model understands not only what the user has asked for in the prompt, but also how those things exist in the physical world.

Sora serves as a foundation for models that can understand and simulate the real world, a capability we believe will be an important milestone for achieving AGI.

Full research release (OpenAI) ↗

Ipsos survey on public sentiment towards AI

Do products and services using AI have more benefits than drawbacks?

We’re seeing a trend that I’ve been very vocal about for over a year now. Asia and less GDP heavy countries are bullish on AI. In the bottom we find the west and in general richer countries.

This flipped equation could mean that many of the more bullish countries will see themselves super charged. People will figure out how to create a lot of value. I think this maps also shows why it will be almost impossible to unite globally around the question about AI. Also why the recent science fiction movie The Creator ↗ might get a few things right in terms of what the future holds.

Ipsos survey (visual capitalist) ↗

Futurist Ray Kurzweil Tells Joe Rogan AI Will Achieve Human-level Intelligence by 2029

"We're not quite there, but we will be there, and by 2029 it will match any person. I'm actually considered conservative. People think that will happen next year or the year after.

I actually said that in 1999. I said we would match any person by 2029. So 30 years, people thought that was totally crazy.

In fact, Stanford had a conference that invited several hundred people from around the world to talk about my prediction, and people thought that this would happen, but not by 2029. They thought it would take 100 years."

Joe Rogan Experience #2117 ↗

AI tools I’m all over this week

1. RenderNet - Create your own AI Influencers in less than 5 minutes.

2. Taplio — Use AI to build your personal brand on LinkedIn in just 10 minutes a day. ⭐️ Highly recommended, as it helped me grow to over 18k followers in just a few months.

3. Typefully – I really love writing twitter threads using Typefully, its my go to tool for drafts.

4. Bedtimestory.ai – I’m super biased, but since we added the new character feature I’m using it daily with the kids.

GTC 2024 - Keynote 18 march 1-3pm PDT

Potentially the most anticipated keynote so far this year. Jensen Huang live on-stage at the SAP Center in San Jose.

You can access the keynote virtually — Watch virtually here ↗

Apple Is in Talks to Let Google Gemini Power iPhone AI Features.

Apple Inc. is reportedly in discussions to integrate Google's Gemini, a generative AI engine, into the iPhone. This potential partnership aims to enhance upcoming iPhone software features. Apple has also explored options with OpenAI, indicating a broad interest in advanced AI technologies.

Full story on Bloomberg ↗

Studio Updates

There has been a flurry of activity, the upper office space is starting to take shape. Insulated and plasterboard is coming up. All the lower framing has been insulated and plasterboard is up.

I’ve been super busy with work and public speaking arrangements in February/early March. So progress has been relatively slow.

The backsplash framing and boards are up for the new kitchenette area. Sorted out some drawings for the new water and some new electricity lines.

I’ve been recording and trying to stitch together Episode 2 for YouTube.

Watch Episode 1 on YouTube to get up to speed on the build out.

Still here?

If you are still here, thank you. If you have questions, feel free to reply to this email or reach out to me on X

If you read this in the substack app, please leave a like or a comment below it let me know if you enjoyed this or not.

First of all, thank you for your relentless support during all of 2023. I’m not sure I’ve fully grasped this journey I’m on yet. It would not be possible if it was not for you. So truly appreciate you all.

Make sure you hit that subscribe button on YouTube. For all the paid subscribers on here there will be a little extra treat soon 🙏🏼.

👋 Hello! I’m Linus, and welcome to this 🔓 open edition 🔓 of my weekly newsletter. Each week, I delve into all things AI, answering your questions, writing op-eds, and creating AI tutorials.

First I’d like to thank this week’s sponsor, webAI.com- Please check them out, it’s because of sponsors like them I can keep this newsletter going.

webAI: Create your own AI powered apps with Navigator AI IDE.

Full code or drag & drop, anyone can build production grade AI in minutes or hours - not weeks. Get early access here

Midjourney V6 Alpha is here

Midjourney V6 is blurring the line between what's real and what's AI-generated. We’re seeing countless examples being posted all over the internet, mainly on X.

My take is that while V6 is extraordinary in quality, it also comes with a quite different art direction than previous models. It's almost as if we have a new hand at play with the general vibe of the model.

🔗 See more examples here

Suno AI - text to music generative AI

Honestly, this is one of those applications that are hard to believe is actually real. It’s still very early days. But I find it relatively easy to plot out a 12-month line and see rapid improvement across the board. If this follows the same trend as generative images have done, we’re in for one crazy 2024.

🔗 Make any song with Suno

AI tools I’m all over this week.

1. Bluedot: AI note-taker that records your meetings without a bot joining the calls. I love this, as a user of existing AI note takers, not having it join in the call. Install free extension.

2. Magnific.ai — Enhance any image for greater quality and details.

3. HeyGen – Effortlessly produce studio quality videos with AI-generated avatars and voices.

4. SunoAI - Make any song with Suno

Google VideoPoet

This one flew under the radar this week. VideoPoet is a large language model for zero-shot video generation. It has a few modalities: Text-to-video, Image-to-video, Video Editing, Stylization & Inpainting.

What's more impressive is that it also has Video-to-audio, so it can look at a generated video and apply the correct sounds to it.

Below is a short video demonstration of what VideoPoet is capable of.

🔗 VideoPoet.research.google

GPT Engineer App

Specify what you want. Get a published web app. Iterate in plain English. This will open up the possibilities for anyone to create apps without knowing how to code.

Get early access

YouTube and Studio Updates

We’ve completed the framing of the new storage room and the new second floor. Between managing the kids, work, and everything in between, I’m happy we managed to get this far this week.

The YouTube episode is not live yet. I’ll let you know when it is.

Partnerships.

I’ve continued to enjoy good reach on X. This has resulted in some great outcomes for some of the partnerships I’ve done. In the past 28 days, I’ve generated more than 137 link clicks on X.

Fully sold out in 2023, but really looking forward to building some long-term relationships with a select few brands/products during 2024.

If you are interested, feel free to reach out. Looking forward to talking. Check out partnership opportunities here: passionfroot.me/linus

Word of the year

I mean, could it be any better?

Still here?

If you are still here, thank you. If you have questions, feel free to reply to this email or reach out to me on X

If you read this in the substack app, please leave a like or a comment below it let me know if you enjoyed this or not.

👋 Hello! I’m Linus, and welcome to this 🔓 open edition 🔓 of my weekly newsletter. Each week, I delve into all things AI, answering your questions, writing op-eds, and creating AI tutorials.

Subscribe now

Ushering a new era for the sports industry

Introducing a new real-time data layer from Rizzle, named exar.live, which enables streaming vast amounts of match data in real time. This data can then be used to render the game on a tabletop in an augmented reality (spatial computing) environment.

These types of experiences would have required millions of dollars and take several weeks to complete.

I’m not a huge sports person, but I can really see this being the future for so many arena type sports, both classic and e-sports. Imagine spectating any game like this 😱

20% OFF Inside my Head until Christmas

Inside my Head is reader supported. By subscribing to a paid plan, you directly support my passion and work.. I appreciate all of you that supports me here and on X. By subscribing you also get access to some exclusive content. However I do try to keep much of the content i publish available to everyone. Going paid means you support me and my work. For that I’m ever grateful.

In the spirit of the holidays I offer 20% off forever for you or if you gift a subscription.

Thank you for your endless support

Get 20% off forever

GTA VI Trailer was leaked, then released.

We will have to wait until 2025 to get our hands on GTA VI. However the first official trailer was leaked. And then Rockstar released it on YouTube.

I must say that everything about the trailer was impressive. Yet, 2025 is a long way out. Given how rapidly AI-driven upscalers and enhancers are improving, I wouldn't be surprised if the graphics end up being even more impressive.

All we need is Nvidia DLSS with some creativity applied. So lets see where we are 18 months from now.

Here are a few enhanced screen grabs from the trailer.

AI tools I’m all over this week.

1. Taplio — Use AI to build your personal brand on LinkedIn in just 10 minutes a day. ⭐️ Highly recommended, as it helped me grow to over 17k followers in just a few months.

2. tldraw — Sketch anything and AI will turn it into a working website in seconds

3. Magnific.ai — Enhance any image for greater quality and details.

4. Leonardo LiveCanvas — Real-time AI design tool

5. Pika.art — Text to Video with greater controls, in-painting and much more.

Google released their new foundational Multi modal LLM – Gemini.

Googles release video of Gemini was… Impressive, but turns out relatively tempered with or lets just say, creatively edited. Despite that, I won't delve deeper into it here as it has already been extensively covered. If you are keen, Matt Wolfe has a good summary on it here.

If we focus on the important things, its a new foundational model from Google more specifically teams at DeepMind and all over the company.

The Gemini Ultra performance is on par or better than GPT4-turbo. But the jury is still out on real-world applications. OpenAI just have to much stuff out in the wild, and Google is not there yet.

Official website for Gemini

Use ChatGPT to help you with Interior Design.

There are many services that allow you to utilise GenAI to create interior images based on your own photos. Pieter Levels have one, interiorai.com

However I wanted to take things one step further, and actually get an interior design consultant, so I turned my gaze to ChatGPT.

I gave it a short half page brief, and uploaded a floor plan of our apartment and started chatting. It's a challenging task, as we are essentially four people living in a 35 sqm (350 sqf) space.

The output was quite impressive, what we had pondered on over countless of hours of inspiration and planning, and consulting with others was basically exactly the recommendations I got from ChatGPT after 15min of conversation. So now we could start from there and get to even better solutions. Read the full thing here.

Partnerships and X (formerly Twitter)

I’ve continued to enjoy good reach on X. This have resulted on some great outcomes for some of the partnerships I’ve done. Managed to generate more than 5000 signups for one partner 🤯 on one X thread alone. Total in the past 28 days I’ve generated more than 157k link clicks on X alone.

There are a few slots left for 2023, but really looking forward to build some long term relationships with a select few brands/products during 2024.

If you are interested feel free to reach out, looking forward to talk.
I’ve updated my partnership portal – passionfroot.me/linus

YouTube and Studio

I'm thrilled to announce that we now have 1.5k followers on YouTube. The rough edit of Studio Build Season 1, Episode 1, is complete and set for release this week. So make sure you are subscribed and click the little notification bell too 🙏🏼 🫢.

Here are some images from this weeks progress. Read full thread here

Still here?

If you are still here, thank you. If you have questions, feel free to reply to this email or reach out to me on X

And fram last weeks newsletter to this, I’ve managed to meet up in person with Bjørn Staal who made the incredible front-end demo that went viral everywhere on the internet.

If you read this in the substack app, please leave a like or a comment below it let me know if you enjoyed this or not.

👋 Hello! I’m Linus, and welcome to this 🔓 open edition 🔓 of my weekly newsletter. Each week, I delve into all things AI, answering your questions, writing op-eds, and creating AI tutorials.

Get 20% off forever

If you missed the BlackFriday Offer, I’m running a CyberMonday (late offer) until Friday, get 20% OFF for life. Click the button above.

In this video, you’ll see me showcase some of the insane outputs I’m able to get from Magnific AI. Some of the results are just mind bending. Make sure you also subscribe to my YouTube and click the notification bell so you don’t miss when I upload videos there.

Also make sure you read the last email, where I summarised last week in my AI land for you.

Leave a comment

I appreciate all of you, thanks for reading, and if you’ve enjoyed this please leave a little comment, or like this post.

See you again Friday.

/ Linus

👋 Hello! I’m Linus, and welcome to this 🔓 open edition 🔓 of my weekly newsletter. Each week, I delve into all things AI, answering your questions, writing op-eds, and creating AI tutorials.

Subscribe now

Entangled - A front end Art experiment.

This is an experiment by the talented Bjørn Staal - Bjørn is referencing quantum entanglement and is using some nifty front-end coding to achieve this effect. If you're wondering why, the answer is simply: because it’s art.

Bjørn has since released a github repo and a demo site test for yourself (desktop only) where you open multiple windows and can play around.

20% OFF Inside my Head – Black Friday Deal

This publication is reader supported, that means, by subscribing to a paid plan, you help me do what I love. I appreciate all that supports me here and on x. You also get access to some exclusive content. However I do try to keep much of the content available to everyone. That means going paid means you support me and my work. Plus you are helping me write and share to more readers.

So today because its Black Friday I offer 20% off forever.

Thank you for your endless support

Get 20% off forever

Magnific.ai - The ultimate AI enhancer

You can now enhance your AI generated images (or any image) using Magnific.ai. They are allowing in people at a steady pace. Built by below is a little example of whats possible. You can alos check out my full thread on x. This tool sits somewhere between upscaling and enhancing. It’s no short of magic.

Lifepaths (using GPT4) by Zachary

Imagine dropping your LinkedIn profile link and getting an entire potential career path laid out infront of you? Yeah, totally doable. This is Lifepaths (It will be live sometime this week) Get in line to try it out

Bonus: GPT4 Vision Skateboarding Coach.

This is another quick one from Zachary. I’ll urge you to look beyond skateboarding. Imagine a world where we will have unlimited visual coaching by machines.

If you want to understand good prompting paradigm, spend some time reading this prompt to better understand how to set up context in a general large language model. Here is the gist on Github.

We’re going to enter some really exciting times. When people are truly starting to understand the type of applications that can be built using AI. Hat off to Zachary

Personal – 28 million weekly impressions, new extreme heights on X

To everyone following me on x as well as here, thank you! Non of this would be possible without you. I had a stretch goal this year, to hit 100k followers on x. As of writing this I’m at 164.512 – I’m grateful for every single one 🙏🏼

I continually try to engage with as many of you as possible, its getting harder and harder, I won’t lie. I’m trying my very best.

Here is some crazy stats from my last 7 days on x. If you are a company looking to partner with me just reply or check out my open slots here currently taking on partnerships for the newsletter and some few spots on x.

Update on Realistic AI photography Guide

My guide on Realistic AI photography for midjourney has been delayed for far to long. There is no real excuse, life has happened, we moved countries, every time I felt ready to release it I got scared and Midjourney released new features.

After talking to many of you, I think the best thing is to finish the current version, release it, and offer perpetual updates to all existing pre-orders.

I’ve been carving out time these last few weeks to get to a state where I’m proud of the work. To all pre-orders, Thank you so much for your patience.

Selected Midjourney Tutorials

🧇 ☕️ Quick Prompt: Kitchen Design

✨ 🧠 Tutorial: Dynamic Prompting

👩‍🌾 👀 Tutorial: Create Consistent Characters in Midjourney

💄 👁️ Quick and Beautiful Portrait Prompts

💬 🛑 Stop using ChatGPT the wrong way

Still here?

If you are still here, thank you. If you have questions, feel free to reply to this email or reach out to me on X and subscribe to my channel on YouTube.

Easter Egg - New Studio Update

The new studio is coming along, slow but steady. Will drop weekly updates over on YouTube. So make sure you are subscribed and click the little notification bell too 🙏🏼 🫢

Build a website in 30 seconds with AI

The AI website builder and platform that makes running a business easier than having a job. Get any business idea online in 30 seconds. Then, drive traffic with powerful AI tools, book jobs, and get paid — all with one AI-driven platform. Free for 90 days using the link below. Thank you Durable.

→ Build your website ←

Tools that change how we build software

We’re currently in the midst of a global AI revolution. One that is touching everything that we are interacting with. Because of this everything is about to change. Street advertising from large brands are using AI to generate A/B/C tests on the fly. In mainstream media shorts on social platforms, voices are cloned from the anchors and mass-produced. These and many more things are enabled by this wave of new tools. Let’s look at a few that are currently creating ripples.

HeyGen

HeyGen is a video tool that allows you to do an array of powerful things. You can create Avatars, Voices, Translate and dub your own content (see below). Plus a lot more. Go and check it out.

→HeyGen.com

Fine

Now this tool is geared towards developers. It’s probably one of the first agent type apps I’ve seen that actually works really well. It works by you focusing on what needs to get done and documenting that. Then the AI gets to work. Laying out a plan, and executing, then you as the human driver gets to review. I think this is the future. Powerful, and brings down barriers to enter.

→ Fine.dev

Project IDX from Google

This is a sneak peek at what Google is working on for developers and the IDE(IDX) platform they are building. Think VS code for the UI, ChatGPT style chat for the AI all running in the cloud.

→ idx.dev

Cursor

From the Anysphere team’s website:

“Humans should focus on bigger problems. Human history is a story of capabilities and tools. New capabilities are rare. They're often discovered rather than invented. In the early ages, we got fire. In the 1800s, we got electricity. In the 1900s, we got computing. And in the 2000s, we are getting artificial intelligence.”

Okay That was a look at some engineering-centric tools this week. I firmly believe that we are living through a paradigm shift at the moment. But it’s not just about developer tools. Next week we will have a look at a few other tools that are turning heads.

OpenAI previews DALL-E 3

Yes, it’s finally happening. DALL-E 3 from OpenAI is getting embedded into ChatGPT and there will be API support. That will happen already next month for enterprise customers and later in November via the labs playground.

What is enchanting is that they are building it straight into ChatGPT. So you’ll be able to ask for images in any conversations at any point in time. Very powerful.

The demo is also cute, they are doing a bedtime story, and I feel like they hit a nerve given that we built and run bedtimestory.ai

Quick Midjourney Light Tips

{Light source} portrait a woman over 30 years old with hair with an even parting of chestnut color with light makeup and light ruddy skin, woman with hair removed from the face, woman with brown eyes --ar 2:3

Here is a quick little tip for light tokens you can use in Midjourney.
Above are 3 tips on light, “golden hour light”, “Light Misty” and “Sunbeams”.

This is also part of the E-book, that’s finally getting ready to ship. So sorry for the delays.

Prompt share

Here are some explorations of prompts I’ve been playing with the last few weeks.

A real car overgrown with long fluffy hair and a man with long fluffy hair --ar 16:9 --v 5.1

Side view, a tribal maskot wearing a crochet wool animal mask, next to a stick, in the style of alessio albi, depiction of rural life, grzegorz domaradzki, restrained forms, shot on 70mm, medieval, photographic weavings --style raw --ar 3:4

<https://s.mj.run/gHwfX8VTMNg> <https://s.mj.run/ZxUEKmU3PBE> Technological, 90's walkman, black and red color, nike air shoe, Nike product photo, white background, studio photography --s 300

candid street photography, close-up analog film photo of a black woman in a mirror-kaleidoscope art installation:: analog film photo, reflections on glass, selective focus, light and shadow::0.5 kaleidoscope::0.5 monochrome::-0.5 --v 5.1 --style raw --s 750 --ar 4:5

side-view of a black off-road vehicle by Dieter Rams --ar 4:5 --s 50

Thanks for reading to the end

If you're still here, I extend my gratitude. Thanks for being a subscriber. Should you have any questions, feel free to reply to this email or get in touch on Twitter

Lastly I’d like to say a huge thank you to this week’s sponsor, durable.co - Please check them out!

PS. Here is a playlist made from all the fine folks over on X sending me their best tunes.

Welcome to this weeks edition of Inside My Head, a member & sponsor supported publication. Thank you for supporting my writing, and I’d like to say an extra thank you to todays sponsor Salad – They offer the lowest price GPUs for AI/ML inference.

Sponsored message

The lowest GPU prices for AI/ML inference at scale

Tired of GPU shortages, expensive cloud bills and waiting for AI-focused chips like the A100?

Access 10k+ consumer-grade GPUs starting at $0.10/hour.
Save up to 90% of your cloud cost on SaladCloud.

Our recent Stable Diffusion benchmark generated 9M+ images in 24 hours for just $1872 - the lowest price anywhere in the market.

Get a demo today to see how you can get more inferences per dollar for your use case.

Thank you Salad for sponsoring this newsletter 🙏🏻

Beyond Smart: Devices That Think and Remember.

Soon, you might find deeper conversations with your toaster than with some friends. Intriguing, right?

The empire of the future belongs to whoever perfects the best Large-Language Models (LLM) combined with memory API.

Imagine, with just a few lines of code, integrating your toaster, fridge, smart speaker, robo-vacuum, or front door lock with an intelligence like GPT-4, or even more advanced.

Suddenly, the inanimate objects around you would have voice, ears, and memory.

Many emerging experiments and apps allow us to converse with data, be it a book, a YouTuber, a famous personality, a departed loved one, or just an object.

Now, imagine that capability compacted into a user-friendly API, applicable to everyday items: your smart mirror, coffee maker, and more. These devices could recognize you, remember your unique preferences, and communicate with each other, forming a unified system - reminiscent of Jarvis from fiction.

Previously, expecting a toaster manufacturer to be proficient in AI development was absurd. Now, they won't need to be.

In a future where virtually anything can possess intelligence and voice, how will that reshape our lives?

I believe this reality is closer than we think. And this LLM+Memory API concept won't just redefine physical objects, but software products as well.

Currently, our interactions are fragmented between isolated AIs. These AIs lack continuity; they don't remember us nor recognize each other's existence. This fragmentation is a hurdle we must overcome.

After nearly ten months of daily interactions, it's frustrating that ChatGPT lacks the ability to recall previous exchanges. It does not know my name, it does not know my favorite colour and It does not know what I do for a living. It's high time AI remembered who we are.

The fusion of short and long-term memory will be a game-changer for LLMs.

Wrap this in an intuitive API, make it accessible for all, and the potential is boundless. Let users interconnect their data across platforms using this shared layer, and the digital realm will truly come alive.

Examples of AI Apps that allow you to chat with Data

• ChatCVS

• Knibble

• BrainyPDF

• Eesel

• MindPal

• ChatWithData

Future of AI assistants

Justin Alvey hacks his Google Nest Mini to run any LLM and custom voice model, demo below is using OnBeeper

Indie makers all around the world are tinkering

Workshop nation turned a boring Alexa Echo device into an animatronic robot! Isn't this what we really want from our technology?

Still here?

If you're still here, I extend my gratitude. Should you have any questions, feel free to reply to this email or get in touch on Twitter

Right now, it feels like it's tough to be certain about, well, anything. From the nuances of material science to the mystery of UFOs, not to mention the blistering pace at which Machine Learning is weaving AI into our lives.

Then, there's the James Webb Space Telescope. With it soaring in orbit, some scientists are scratching their heads, wondering if the Universe might actually be 27 billion years old instead of the previous estimate of 13.77 billion.

So, here's a thought: what if we're just part of some colossal experiment orchestrated by a higher intelligence? Imagine if AGI isn't just a future concept, but our reality—and we're living in its petri dish.

Honestly, I doubt there's a straightforward answer. Much like how we can't quite grasp the 4th or 5th dimension, hunting for evidence might lead us nowhere. If this hypothetical proof were organic in nature, it might be all around us, hiding in plain sight. Heck, maybe the mycelia network is the universe's way of keeping tabs, reporting back to some grand entity in ways our brains can't fathom.

When I listen to voices like Sam Harris or Eliezer Yudkowsky, I get where they're coming from. But I can't shake the feeling that our knowledge barely scratches the surface. Perhaps we've only uncovered a measly 0.000001% about the real nature of our existence.

Life is fleeting. Our cells degrade as we age, and many things about our world just don't seem to add up. There are quirks in how life unfolds and how minuscule we are in the grand tapestry of the cosmos. These might just be breadcrumbs hinting at a bigger story.

Think about it. Earth's been around for a staggering 4.543 billion years. That’s 56.7 million human lifetimes, back to back. And us humans? We've been on the scene for just a blip. It makes me wonder if entities, who perceive time differently than we do, are running cosmic tests on a grand scale. Or maybe, just maybe, we're an outcome of the famed Drake equation.

Frank Drake once speculated that the Milky Way alone might host between 1,000 and 100,000,000 civilizations. Given that hopping to other galaxies might remain a dream, let's keep our sights on our own Milky Way backyard.

So, if otherworldly life does exist, where the heck is it? Aren't we scanning the skies? True, but humans have only been hanging out in LEO for less than 50 years. We're still babes when it comes to space exploration. Until our own solar system becomes as bustling as Earth's airports, we might just be waiting a bit longer to bump into other intelligent beings.

However, if the latest UFO whispers hold any weight, maybe we're already on someone's radar. These beings might know more about us than we do, or they've just journeyed farther down the cosmic path. Who knows? They might be observing a hands-off policy thanks to some space-age treaty we're blissfully ignorant of.

Pondering the vastness of it all is mind-bending. We often see ourselves at the top of the Earthly food chain. But on the cosmic scale? We might be as inconsequential as ants or even viruses. Regardless, from our vantage point, we matter. Maybe we're a once-in-a-universe anomaly.

Though I'd bet against those odds.

/ Linus

Still here?

Still reading? Thank you. I hope you enjoyed this, if you have questions, feel free to reply to this email or reach out to me on Twitter

🔒 This is exclusively for subscribers. 🔒

The inaugural office hours happening now!

🌮 Join the Office Hours (Zoom meeting)

Read more

This simple midjourney prompt allows you to create really fun photos that are incredible fun to share in that family group chat of yours. See how many of your family members you can prank. Let’s dive in.

“Hey dad, remember that time the bears came into camp and you just chilled with them while eating your dinner?”

Below is the prompt used for this, you can manipulate it however you want, easiest thing is just to change the [Scene Description] part to what ever you want. You can also play around with the aspect ratio.

[SCENE DESCRIPTION], found photo album photo, photo scan, taken on disposable camera, night flash photo --v 5 --ar 3:2

As we can see it’s really easy to generate random images that could potentially be in the family album.

THIS WEEKENDS SPONSOR IS HOWARD, THANKS FOR THE SUPPORT

Free, instant access to the video lesson and cheat sheet that helped us launch two AI products in four weeks.

→ Get Free, Instant Access

Let’s try creating some more photos

“Dad rescuing me from the tree I accidentally climbed into when I was 4 years old”

Personally I don’t remember all the strange things that happened during my childhood, and looking through old family albums often have me shaking my head. Some images are even hard to see if they are actually photos of me or my siblings or some random neighbourhood kids.

“That time that dad accidentally used the wrong candles in my birthday cake”

Here is an hypothetical thing that my dad most certainly could have done. So I found it really fun to make this into an example.

“Grandma, Grandpa & Mom dressed for carnival, wtf mom?!”

If this photo came up in my family album, I’d never second guess its authenticity. Grandma and grandpa did a lot of crazy things. So I’d buy into this straight away.

Face swap to make it more real.

The next step to make these images realistic for your family is to use face swap on your Midjourney generations. Below is a great short tutorial on how to use Faceswap in Midjourney.

Still here?

If you're still here, I extend my gratitude. Should you have any questions, feel free to reply to this email or get in touch on Twitter

PS: Thanks for all the feedback on the last email, starting to understand more what it that you want from this newsletter.

For all subscribers here and on Twitter, I'm thrilled to announce my first-ever office hours. I envision this as a recurring bi-weekly event and will be testing it out over the next couple of months. I'm excited to interact with many of you in this new format.

🔒 This is exclusively for subscribers. 🔒
The link on how to join is provided below.

The inaugu…

Read more

The MSG Sphere

This is the $2.2bn Sphere in Las Vegas, and it’s not CGI. It’s covered in LED lights that together make up this massive Sphere that can display any digital content on it.

Personally I love this type of mix between architecture and digital. The Sphere is one of the first venues in the world that are built at this scale thats 100% is built for the experience of sound.

Apple Music has made an engaging interview with U2 at the Sphere, discussing their upcoming performance there. Well worth your time.

Adobe Generative Fill to Creating Memes

I’ve been using Generative fill now since it got out, and I never stop being amazed by all the crazy things people do with it. This was no exception.

In my opinion, Adobe Generative Fill is one of the best-positioned genAI tools on the market. Its integration into Photoshop and easy accessibility give it a competitive advantage.

Frank Sinatra's AI Rendition of Gangsta's Paradise

Upon first hearing this, I was floored. My mind spiraled into a flurry of thoughts - if we can do this, what else can we do?

Will my kids have the technology to digitally recreate me once I'm gone? Soon, we might be able to recreate any song by any artist on platforms like Spotify. What does this mean for the artists? Incredible nonetheless.

In my research, I discovered Lace Editing, the creative source behind this project. They've shared a tutorial on how to create these AI covers – and it's surprisingly straightforward, “It’s very easy”.

An Update on the Newsletter

For over two months, Substack, the platform hosting these newsletters, has been facing a constant shadow ban on Twitter.

This has significantly hindered my ability to manage the newsletter as intended, and for that, I apologize. I'm exploring potential solutions to address this.

Some alternatives include switching platforms, setting up a custom domain, or even attempting to persuade Elon Musk to lift the ban.

Regardless, I find myself here, asking for your opinion. I've tried several times to elicit your feedback, but haven't quite cracked it. So, I'm reaching out again.

What content do you primarily wish to see? News, Tutorials, random ideas or thoughts, interviews with AI/Tech leaders?

As I draft my plans for the rest of the year, I'd love to incorporate as many of your suggestions as possible.

Still here?

If you're still here, I extend my gratitude. Should you have any questions, feel free to reply to this email or get in touch on Twitter

PS: To all my paid subscribers, stay tuned for another email coming your way.

👋 Hello! I’m Linus, and welcome to this 🔓 open edition 🔓 of my weekly newsletter. Each week, I delve into all things AI, answering your questions, writing op-eds, and creating AI tutorials.

Subscribe now

A brief intro to Midjourney 5.2

Discover an all-new Aesthetic System.
Midjourney have honed the visuals for enriched aesthetics and sharper images. Revel in the understated enhancements in text understanding and coherence, and plunge into a broader diversity of outcomes.

You might need to roll more than once to find your desired result, but they assure you it's part of the thrill!

The --stylize command has been remodelled to significantly impact the stylization of your image, much like in V3. The command parameters extend from --stylize 0 to --stylize 1000, with --stylize 100 being the default setting.

High Variation Mode

Introducing the "High Variation Mode".
Activated by default, this mode notably amplifies the variety of all variation tasks. To adjust settings, simply type /settings and select an alternative variation mode. Additionally, you now possess the capability to dictate the strength of variation for each upscale. This is pure fire.

/shorten command

Presenting the innovative /shorten command.
With this feature at your disposal, you can 'inspect' a prompt and receive suggestions on potentially unnecessary words and those that might be crucial.

Zoom-out, outpainting

Say hello to the "Zoom Out" Feature.
Reframing the image has become effortless with the newly introduced 'zoom out' buttons located beneath all upscales. Options [Zoom Out 1.5x] and [Zoom Out 2x] retreat the camera, enriching the details all around. [Make Square] empowers you to augment details on two sides, transforming a non-square image into a square one. For advanced users, [Custom Zoom] offers a popup text box that lets you alter the prompt as you zoom out and adjust the aspect ratio or specific zoom. Set --zoom 1 and tweak the --ar to your liking to modify the aspect ratio. This feature is designed to eradicate black bars from your prior images.

Join me tomorrow for a live workshop, a few spots left.

🔗 Live Workshop on June 27th at 7pm GMT+1

You are creative. This course helps you quickly bring ideas to life.

This high-energy course will teach you about AI by helping you use AI. You will learn to use AI, specifically Midjourney, to improve the creative visual processes of your entire team, from photography to graphic design to logo design and ideation.

Curated Tutorials

🧇 ☕️ Quick Prompt: Kitchen Design

✨ 🧠 Tutorial: Dynamic Prompting

👩‍🌾 👀 Tutorial: Create Consistent Characters in Midjourney

💄 👁️ Quick and Beautiful Portrait Prompts

💬 🛑 Stop using ChatGPT the wrong way

Still here?

If you are still here, thank you. I hope you enjoyed this little dive into Midjourney version 5.2. If you have questions, feel free to reply to this email or reach out to me on Twitter and subscribe to my channel on YouTube.