This week the argument about AI in filmmaking changed shape, and it changed because of who showed up. Martin Scorsese, of all people, is now using AI to storyboard his next film. At the same time, the crown for "best AI video tool" changed hands again, and a couple of genuinely useful tools landed that you could put to work this afternoon. Put those together and the real story comes into focus. This stopped being a question of whether serious people use these tools and became a question of where and how.

In today’s AIography:

Read time: About 7 minutes

WHAT I’M THINKING ABOUT

If you wanted to pick the last director on Earth to publicly embrace AI, you might have picked Martin Scorsese. This is the man who has spent years warning that the theatrical experience is under threat, who treats film history like scripture. So when the New York Times reported on June 2 that Scorsese had signed on as a partner and adviser to an AI company, and that he used its tools to storyboard his next film, a lot of people did a double take. I did too.

Here is what actually happened, because the accurate version matters more than the headline. Scorsese partnered with Black Forest Labs, the company behind an image-generation tool called Flux, which is software that creates still images from a written description. He used it during pre-production on a new film, and he tested it on one specific job: storyboarding a single scene.

His own words, from the interview: "During the pre-production process, time costs money, and this allowed us to move faster without sacrificing quality or craft." That is the entire claim. Not "AI made my movie." Not "I replaced my crew." He used an image generator to help him show people the pictures already in his head.

It is worth sitting with how big that still is. In the same statement, he gave his reasoning: "Remember, cinema is a young medium, only around 125 years old, so we have to be open to how it can evolve." When a filmmaker of Scorsese's stature says, on the record, that this technology earned a place in his process, it moves the whole conversation. For a large group of talented people who have been standing on the fence, quietly experimenting but embarrassed to admit it, this is permission. The most respected living American director just told them it is allowed.

The backlash arrived within hours, the way it always does. Some of it is reflexive, the same crowd that calls everything AI-touched "slop" no matter how it was used. Some of it is worth taking seriously, and the way to take it seriously is to be precise about what Scorsese actually did. So let me be precise.

Previsualization, or previz, is the stage where a director works out what a scene will look like before anyone rolls a camera. Storyboards, rough sketches, animatics. It has always been the place where a director gets the movie in their head across to the people who help build it: the cinematographer, the production designer, the editor, the studio writing the checks. For a century that meant sketches on paper, or more recently rough 3D layouts built by hand.

What Scorsese did is use an image generator to do that faster. Instead of describing a shot to a storyboard artist and waiting days for boards to come back, he could generate the frame, look at it, adjust it, and put it in front of his team the same afternoon. That is the whole use. The generated pictures are not in the movie. The actors are not generated. The film still gets shot, in camera, with a real cast and a real crew.

This distinction is not a technicality. It is the difference between the thing people fear and the thing that actually happened. The fear is that AI replaces the artist and the worker. What Scorsese demonstrated is AI used as a sketchpad, a faster way to think out loud, with every human role on the production fully intact. One of those is a threat to your livelihood. The other is a pencil that draws quicker.

I will be honest about the part that is not free of cost. A storyboard artist is a real job, and a director generating his own boards is one fewer gig for that artist. That is a fair worry, and pretending otherwise would be dishonest. But notice the scale of it. We are talking about the sketch phase, not the shoot, not the cast, not the cut. The previz artist was always the most squeezed link in the chain, hired for a few weeks and let go. The bigger jobs, the ones that make a movie a movie, are exactly the ones Scorsese kept fully human. The line he drew is the line that matters.

I have watched a few of these moments from inside the edit bay, each one arriving dressed as a monster and turning out, after the dust settled, to be a better pencil. The nonlinear edit did not end editing. It ended rewinding tape and waiting for a film lab. The trick, every time, was telling the difference between a tool that does your job and a tool that does the boring part of your job so you can spend your attention on the part that matters. Previz is the boring part. The vision is the part that matters, and that is still entirely Scorsese's.

So when someone tells you Scorsese "is making AI movies now," they are wrong, and the correction is the most useful thing AIography can hand you this week. He is using AI to draw faster. That is a thing you can do too, today, without surrendering a single creative decision.

You do not need Scorsese's connections to do it, either. He happened to use Flux, but the same previz job runs on tools you can open right now: Midjourney, ChatGPT's image generator, or Flux itself. Describe the frame you see in your head, generate a few versions, pick the one closest to your intent, and you have a board to shoot from. The famous name got the headline, but the capability has been sitting on your laptop for a while.

While the culture was busy arguing about Scorsese, the tools kept moving, and they moved fast.

On May 31, xAI's Grok Imagine Video 1.5 took the number one spot on the Image-to-Video Arena. Two terms worth unpacking. "Image-to-video" means you feed the tool a still picture and it generates a moving clip from it, which is one of the most practical ways working creators actually use these tools. And the "Arena" is a public leaderboard where people watch two AI clips side by side without knowing which tool made which, then vote for the better one. Those blind votes produce a score, and a higher score means real people preferred that tool's output more often.

Grok's new version jumped 52 points in that scoring, enough to pass ByteDance's Seedance 2.0 and Google's Veo, the tools that had been holding the top. It generates clips up to 720p with sound built in, which most competitors still cannot do well. The catch, and it is a real one, is that you mostly cannot use it yet. It is open to developers through a code connection, with the consumer version still rolling out across X's paid tiers.

So the practical takeaway is not "switch to Grok today." You probably cannot, and by the time you can, something else may have taken the crown. The takeaway is the pace. The single best image-to-video tool is now changing month to month. That is the real lesson buried in the leaderboard, and it argues for a specific way of working: do not marry one tool.

In plain terms, a swap-friendly process means keeping your creative decisions in one place you control, your shot list, your reference images, your edit, and treating the generator as a part you can unplug and replace. Save your reference frames and prompts so you can run them through whatever tool is best this month. Do your assembly and finishing in an editor you own, not locked inside one generator's website. Then when a new king takes the leaderboard, you switch the engine without rebuilding the car. The creators who get burned are the ones who built their whole pipeline around a single tool that was number one in March and forgotten by June.

Underneath the headline drama, two tools landed this week that you can actually put to work, and they point at where the value is heading.

ElevenLabs released Dubbing v2, an upgrade to its tool for translating a video into other languages. A "dubbing model" is software that takes spoken audio and re-voices it in a different language. The thing that always made AI dubbing feel cheap was that it flattened the performance, turning an emotional line reading into a robotic one. Dubbing v2's pitch is that it listens to the original performance and carries the emotion, tone, and pacing into the new language, across more than 90 of them, with the translation, voice match, and timing handled automatically.

Picture what that unlocks. You make one video in English, and without a studio, a translator, or a voice actor, you have a version that plays naturally to a Spanish, Japanese, or Portuguese audience, with your delivery intact rather than ironed flat. For a creator trying to grow past a single-language audience, that is a usable-this-week tool, not a someday promise. Test the fit for your own work before you bet a client job on it, but localization is one of the most immediately practical things AI does for working creators right now.

Sondo AI, a smaller player in the AI music-video space, added a full editing layer to its generator, so you can restructure scenes and sync to audio without leaving the tool. On its own that is a niche story. But it rhymes with the bigger pattern, the same one behind PAI 2.0, the all-in-one platform that led our skipped Tuesday issue: the generators are growing finishing rooms. The race is shifting from "who makes the best clip" to "who lets you take a project all the way without exporting and re-importing through five different programs." The model is becoming a commodity. The workflow is becoming the product.

ACTION ITEM

Take the next thing you have to shoot, film, or even just pitch, and use an AI image generator to storyboard it before you commit. Generate the key frames: the establishing shot, the emotional beat, the moment you are unsure about. Not to put in the final piece, but to see it, adjust it, and show it to whoever needs to be on the same page as you. A client, a collaborator, or yourself at 2 a.m. when you cannot tell if the idea actually works.

You are not replacing your craft. You are doing exactly what Scorsese did, using a fast sketchpad to get the picture out of your head and in front of human eyes sooner. If it is good enough for a Scorsese film, it is good enough for your next spot. That is the whole permission slip, and it costs you nothing to use it.

SHORT TAKES

FINAL THOUGHTS

I keep coming back to the image of Scorsese at a screen, generating a storyboard frame, nudging it, generating it again. Not because it is shocking, but because it is so ordinary. That is what this looks like when it is healthy. Not a machine making the movie, but a master using a faster pencil to get his vision out of his head and onto the table where the real work happens.

The people who do well in the next few years will not be the ones chasing whichever tool is number one this month, because that crown will move again before you finish reading the announcement. They will be the ones who figure out where AI actually belongs in their process, the way Scorsese put it in previz and nowhere else, and who stay loose enough to swap the parts as everything keeps shifting. The tool that wins is the one that disappears into your work and lets you spend your attention on the thing only you can do.

The machinery changed again this week. It will change again next week. The taste, the eye, the judgment about what the story needs, that is still yours, and it is still the only part that was ever the point.

See you next week.

—Larry

ESSENTIAL TOOLS

Check out the Alpha version of our AI Tools Database. We will be adding to it on a regular basis.

Got a tip about a great new tool? Send it along to us at: [email protected]

If you have specific feedback or anything interesting you’d like to share, please let us know by replying to this email.

AIography may earn a commission for products purchased through some links in this newsletter. This doesn't affect our editorial independence or influence our recommendations—we're just keeping the AI lights on!