PJ Accetturo's Recent LinkedIn Posts

AI is now 100% photorealistic with the launch of Kling 3.0 In just two days, I created the opening sequence from The Way of Kings by Brandon Sanderson. You have to try this new Multi-Shot technique that makes making films much faster and cheaper. Multi-shot lets give you near-perfect continuity in a sequence. This is one continuous output from Kling based on one image. You do this by uploading a single image, click "Multi-cam" and then you can set the length for each shot and describe the differences in each shot. My prompts were pretty simple and for this one I used a Start Frame and End Frame so it would understand both characters. "alternate angle: handheld camera: two warriors fight. A bald warrior in white with a 6 foot long glowing sword fights an armored soldier in blue armor with a 7 foot long steel large sword. They fight with master swordsmanship." Speaking is also GREATLY improved. When you combine it with multi-cam, it really looks nearly indistinguishable from a real film. (You can generate up to 15 seconds per clip now) The Multi-cam is cool, but the biggest game-changer for me is how realistic the speaking is. This is what we've been waiting for. We're ready for 90min AI feature films. I made this 90-second sequence in two days. Does that mean I can make a 90-minute film in 180 days? Maybe. I'm a huge fan of Brandon Sanderson and the Cosmere and really excited that Apple+ is greenlighting Mistborn and Way of Kings. Brandon spent 10 years trying to get his books adapted and I'm sure Apple will spend nine figures on this, but AI now lets smaller IP get greenlit at a far lower cost. Thanks to Kling for letting me have early access! (Not paid post) If you want to see a full breakdown of how I made this film, follow my newsletter: https://pjace.beehiiv.com/ Anyone else a fan of The Way of Kings?

Fast Company just nominated our Kalshi ad as one of the top 7 ads of the year. I literally made this in my underwear in two days. You can just prompt things 😂

I just found a Sora 2 cheat code This will let you make a million-dollar full anime episode with CONSISTENT CHARACTERS right now. Homemade anime will be the biggest trend of 2026. Millions of new shows will be born next year and here's how they'll do it: Sora only outputs 10-15 second clips, so you need to have a way to hack the system for consistency. Screenshot a collage of images (see reference image in the comments) as the uploaded image for Sora to reference, so that it knows how both characters look. Then you need to prompt each 10-15 second scene for what happens with the action. Describe the characters in the same way each time. Voices: In order to get consistent voices, at the moment, you're likely going to need to run it through Elevenlabs voice changer or lip sync it yourself and voice change yourself. (sucks) Music: Try and prompt it to avoid music so you can have control over it later. Is the tech there yet? No. But this is the basis for uploading episode script, adding character photos and letting Sora 2 just rip it out in a setting. (we're probably 3 months away from this) Check out the Mage Quest YT series here. I didn't make it, but this will be the technique to build on for full TV shows. https://lnkd.in/eTztR34F Curious to see who else can find Sora cheat codes like this. Tag me in it! I'll be posting lots more Sora 2 hacks and tips here: Joining my newsletter is free and takes 5 seconds 👇 https://pjace.beehiiv.com/

We’ve now hit 95% photorealism with AI video. Within a few months, it will be nearly indistinguishable from real life. But what does that actually mean? Within 12 to 18 months, live-action productions will drop to near zero, except for the highest-end Hollywood projects, the Vinyls floating in a digital sea. Brands will put out an ad per month, then an ad per week, then an ad per day, and eventually they’ll personalize every single ad for the viewer. Right now, making ads like this is still manual and time-intensive, but platforms like Sora show us that some level of automation is inevitable. For the next year, AI ads will still be a co-pilot workflow between humans and AI, but after that, fully automated and personalized ads are basically unavoidable. So what actually breaks through when a customer is hit with an ad a day from every brand on the planet? Stories. Narrative. Stories are the brain’s natural language. We’ve made sense of the world through them for thousands of years around the campfire. Brands will start experimenting with episodic weekly narrative content. Imagine if Dunder Mifflin was a real company and The Office counted as a content marketing expense. This is the endgame. Creators who never loved making ads in the first place, but did it because that’s where the money and opportunities were. Brands win either way, because character-driven stories create emotional attachment to their product. Think about how Pixar is basically a toy company disguised as a media company. If you’re a creator, ride the advertising wave for the next 6 to 12 months, then start pivoting into episodic, 3 to 10 minute narrative series funded by brands. That’s the real opportunity coming. The video below was made by my friend Seb Shueti for the Up Summit. Seb is one of the most underrated AI filmmakers out there, and he only drops a few films a year, so it’s always a pleasure to watch him push these models to their absolute limits! What do you think? How will the next two years of AI video unfold?

OMG, our new commercial for Ownwell is insane 😂 —- Pro tip 👇🏼 We had to “reshoot” it a few times because lip sync models weren’t that good when we started. The secret was to measure the bpm and then give Kling 3.0 and Veo the bpm and lyrics and then sync it up manually! Massive props to Dan Packer on directing this and the rest of our team at Genre.ai bringing it to life!

“We need the most degenerate ad possible to launch our startup.” PJ: “Can it include a coked-out grandma in a bedazzled bikini riding an alpaca and ripping bongs with college kids?” Kled: “...bet” Here’s how we made our most insane ad ever 👇 --- Let's start with the legendary Veo 3 prompt. This is my pièce de résistance, learned from a lifetime (3 months) of prompting: A cinematic handheld close-up shaky camera shot at night, following a 90-year-old grandma riding on the back of a galloping alpaca through a suburban residential neighborhood. The grandma is wearing a sparkling, bejeweled USA flag bikini and a thick gold chain around her neck. Her wrinkled face is full of wild joy, gray hair flying wildly behind her. She is screaming at the top of her lungs, mouth wide open: “I GOT PAID $400 FOR UPLOADING PICS OF MY FEET!” Behind her, two college girls also wearing bikinis ride on their own alpacas, laughing and yelling, joining the wild chase. Police car lights flash red and blue in the distance, illuminating the neighborhood houses, mailboxes, and front lawns in chaotic bursts. The shaky handheld camera captures the intense energy, close on grandma’s face and upper body as the background blurs with speed. Streetlights and cop lights create harsh flickering highlights, casting deep shadows and chaotic streaks of light, emphasizing the reckless, rebellious midnight adventure. • Time of Day: Night, suburban street lit by streetlights and flashing police lights • Camera Type: Handheld digital cinema camera • Shot Type: Close-up shaky tracking shot, focusing on grandma’s face and upper body • Focus: Sharp on grandma, motion blur in background • Lighting: Mixed harsh artificial lighting from street lamps and cop car lights, creating dramatic highlights and shadows • Atmosphere: Wild, comedic, chaotic, rebellious • Audio (implied): Grandma’s loud screaming, college girls yelling, distant sirens, hooves pounding pavement -- Linkedin would cancel me if I listed the entire B-roll list (Old people riding Giant D*ldo shaped fireworks...etc.) But I do a full deep dive on my X account here: https://lnkd.in/esSg9KDD -- In my newsletter I also do a full breakdown of the prompts and process! Follow along here! https://pjace.beehiiv.com/

We made an ad for a major airline in 14 hours at 30,000 feet! Here's how we pulled off this insane video for Qatar Airways with the help of Google and Starlink. Full prompts & process below👇🧵 A month back, Google MENA’s Andreas Toscano Mielenhausen hit me up with a wild idea for Qatar Airways. They wanted to show off how fast their Starlink Wi-Fi is and how business isn't interruppted, even though you're traveling. Qatar Airways's Daniel Andrade loved the concept and booked an entire business class suite for our team. So I pulled in top director Torey K. for a head-to-head, with Roberto Nickson hosting the challenge. (Note, if you read the rest of this post on X, it's 1,000 times better because each example has videos and photos. Here's the linK: https://lnkd.in/efeyfq4m Creative Brief: Andreas provided us with a stellar creative brief and storyboards for both commercials which made this MUCH EASIER. Creative concepting can take 15+ hours alone so though we were free to make tweaks, it really helped us have a framework to tell the story. First thing we did when we got to the airport was take as many reference photos as we could. The Qatar team also supplied us with a number of reference photos. This was CRITICAL in getting near perfect details to the airport and the plane. Generating two full spots in 14 hours would have been ABSOULTELY impossible without my team at Genre.ai helping dive in in real time. With Starlink we were all able to jump into Figma and use Gemini to create all the shots together. Thank you to Dan Packer Olga Baranova, Julie Wieland , Marco Isle, Matthew Warter, and Austin Cook. Once all the images were finished we threw them in Veo 3.1 for animation. Simple prompts work best: A photorealistic video. The camera slowly orbits left. The man on the left sips his coffee. The man in the middle grins, looking around, excited. The girl on the right continues to look at her phone. push in on closeup angle of man smiling, looking up Music: We worked with Austin Cook and @TUPACABRA2 for the music. My video's track was done by Tupacabra using Suno and I'm blown away at how good it sounds. The Edit: After 4-5 hours of sleep over the last 48 hours (and then 12-13 hours of prompting and animating...Torey and I were cooked. But we pushed through with the help of the team, providing some amazing After Effects help with some of my favorite transitions ever! And with less 15 minutes before landing, we completed the two commercials! Go to Qatar's Youtube Channel to vote (just comment/like) on which commercial you liked best! Big thanks to Qatar and Google for having us out! If you liked this, I'll be doing a deeper dive into our process on my newsletter: https://pjace.beehiiv.com/ Also huge thanks to everyone at my agency Genre.ai for making this possible 30,000 feet in the air over 14 hours! Most insane project yet!

We are about to see a massive explosion of books turned into cinema. Previously, a series like Red Rising would've needed a $200M+ greenlight. Not anymore. I used Luma AI's new agent to build this teaser in less than a week. Bookmark my full workflow + Seedance 2.0 prompts 🧵👇 Full thread on X with images (this is 10x easier to read there) https://lnkd.in/gNXFqsQ8 Red Rising is my favorite sci-fi series (2M+ sold). As AI ads automate, my company Genre.ai is taking big swings at epic IP that would otherwise cost $200M+ to greenlight. We're able to create AI-hybrid series with real SAG talent for a fraction of the cost. Luma gave me beta access a week ago to test their agent out, and this tool is insane. I was able to generate THOUSANDS of images and do some extremely complicated sequences in record time. This is a force multiplier for creative work. Story background:   This is actually a passage from book 2, but I’ve always felt like this sequence is the best microcosm of the series as a whole (yet has minimal spoilers). Darrow (main character) and his crew are escaping in a small ship and have to run an entire armada that is blockading their escape back to friendly territory. Darrow hatches a crazy plan to launch himself as a missile onto the 5km long capital ship’s bridge, and start a revolt from within the ship, having the lower classes revolt against their Gold overlords. This is where Darrow meets Ragnar, who is an absolute titan of carnage. Born into his own type of slavery, Ragnar relishes the chance to strike back at his captors and join a leader that may free his people. The Red Rising series is an epic tale of rebellion against an oppressive caste system, and this scene encapsulates that meta-story. Script: This is a 20min sequence in the audiobook that I had to condense down to 3min.    I worked with an LLM to strip out just the dialogue and then I further condensed that into the most essential lines. At the start of this process, I gave Luma's agent: The 15-page book passage My script/shotlist My 2x2 prompt structure I had it generation 30 options for each character's look and then began integrating them into the scenes. My chat with the agent was: "Use this description of the scene and character references to generate 30 2x2 grids of them in the scene." It then gives me back more cinematic shots which I then use as references for more cinematic versions of the scenes. This also keeps the location consistent when you're giving it reference images of the characters in the location. --- Linkedin is cutting off my article length here. To see this with images that show you every step in the process, check it out here: https://lnkd.in/gNXFqsQ8

Almost exactly a year ago, I made my first AI film and woke up to 10M+ views and hundreds of death threats. I almost quit. Thank God I didn't. Best 12 months of my life. "First they ignore you, then they laugh at you, then they fight you, then you win." Keep creating 👍🏼 If you want to see the viral film "Live Action Princess Mononke", here's the link: https://lnkd.in/eSCMwnkw

I took 6 top Hollywood cinematographers with ZERO AI experience and taught them a simple framework to master the tools in minutes. Here's all the secrets of how we made this film for the $1,000,000 AI film festival. Exact tips, prompts, and framework we used 👇 This thread is 100x easier to understand with visual examples, take the 2min to read this on X by clicking the link below so you can see the pictures referenced: https://lnkd.in/g5tB4DpK Shotlist: I asked my friend Dean Israelite (director of the Power Rangers movie) if he would direct this film while I played more of a showrunner and AI orchestrator. He started the process by creating a shotlist. Pro tip: If you're new to shotlisting, upload your script to Gemini and ask it to create a shot list for you as if Gemini is an award winning director. Then tweak the shotlist to match your vision but it can do a pretty good first pass at the shots. — Character Creation: Before generating scenes, generate your main characters for the film. Prompt them on an off-white background with soft lighting. First find a face that you like, then use Nano Banana Pro to generate a wide shot or two of them. — Character Sheet: You don't need to do the 3x3 grid for a character sheet. It actually hurts your consistency because it's too low-resolution. Just generate two shots, a closeup and a wide and stitch them together as one image. This is what you'll use as a char ref later. — Shot Organization: If you learn nothing else from this thread, this is the most important thing I can teach to you today. By having a framework like this in Milanote (below), Figma or Miro, can divide your films into scenes. — Shot Org 2: I first generate plate shots of the scene, and then I drop actors into the scene (see images on left side) I then use those as references for a lot of the shots in the scene for consistency (and place them on the right side next to each shot). — Image Generation Overview: To generate the images, we want to use a platform that supports Nano Banana Pro. @freepik is the best for this. Notice how i'm easily able to add my character sheet as a reference and also a style reference? You don't need a fancy prompt. — Style Reference: I took references from my original film and created a collage image so that it gets a cohesive aesthetic style but also some variation in the lighting and time of day. Always use a style ref for consistency in a project! -- The rest of this thread is continued on my X account, check it out here: https://lnkd.in/g5tB4DpK -- Credits: Created by: PJ Ace Director: Dean Israelite Writing: PJ Ace / Mike Ryan Editing: Martin Bernfeld Cinematography and Animation: Olga Baranova Matthew Warter Evan Zissimopulos Gabriel Olsen Luke Palmer Sina Dolati Andres Aquino Nick Mahar Alex Duloz Blvckl!ght Check this out on X for the full experience!

Luma giving away $1M! Also, I'm a huge fan of Jesus in advertising. Back in the Renaissance period, religious iconography was the highest form of art. There's something iconoclastic about mixing Jesus, AI, and advertising that gets people's attention. Is it sacrilegious? It depends on how it's portrayed. I've gotten millions of views on my "Bible influencer" videos, but I think if you treat the characters with respect, it's a conversation starter. A lack of Jesus in public discourse feels like a worse world than one where he's inserted playfully. Looking forward to see the first Cannes Lion won by an AI filmmaker! The next few years of advertising are going to be fun :)

Sora shows us the future of advertising. Here's how it will go down over the next 6 months 👇 They'll launch a "pro mode" within 2 months. (no watermark, ability to tweak individual clips) Brands will need to make an ad every other day to stay relevant. Production companies will be cooked within 6 months. Agencies will fire nearly everyone except writers/directors who can make repeatable, "on brand" and funny content. Comedy = 👑 I'm putting together a gameplan for brands and agencies in my next newsletter, make sure you're subscribed 👇 https://pjace.beehiiv.com/ (Video by Andy Orslow)

Seedance 2.0 is about to launch in the US and I'm sharing my ENTIRE WORKFLOW GUIDE below! This will change your life 👇 I recently used it to create an ad for AI on the Lot in Dreamina AI record time—literally 1/10th of the usual workflow. Here is the 7-step framework I used to move from a "vibe" to a finished cinematic promo: Linkedin is terrible for these guides so take 2min to read it on X with images nad examples: https://lnkd.in/gSPgMYSr 1. The Brief Keep it simple. For this project, the goal was a "wild and fun love letter to LA." No complex scripts—just a core energy. 2. The Shot List (AI-Powered) I used Luma's AI agent to brainstorm hundreds of "larger-than-life" LA stereotypes in iconic locations (like a nun on a motorcycle or a gorilla on a skateboard). 3. Image Curation I generated images in 2x2 grids to see volume. When I found a "hero" shot (like the skateboarding nun), I’d expand that specific vibe into 40+ variations to get the perfect coverage. 4. Animation with Seedance 2.0 Upload your reference images directly into Dreamina. Instead of trying to nail one long sequence, focus on generating multiple 2–4 second clips. This gives you the flexibility you need for a professional edit. Check out my X post to see the full prompts: https://lnkd.in/gSPgMYSr 5. Prompting for Movement Use high-octane language in your animation prompts. Focus on: Camera Work: "Whip pans," "snap zooms," and "ultra-low tracking shots." Physics: Describe the interaction (e.g., "wheels rattling," "dust clouds rising"). 6. The "No Music" Hack Always add "No music" to the end of your animation prompts. This keeps the generated sound effects clean, allowing you to layer your own custom track later without audio "mud." 7. Accuracy by Volume Generate 5–10 variations per scene. This allows you to cherry-pick the most realistic physics and most dynamic movements during the final edit. Mastering this workflow over the next few months will change how you approach production. Join us at AI on the Lot (May 27-28) at The Culver Studios! Get 15% off your tickets using the code "SLOP" and see these workflows in action. 🎬 Thanks! Todd Terrazas and Zoey Zhao! Register for the event: https://lnkd.in/gK7CgGRC Try Dreamina Seedance 2.0: https://lnkd.in/gMuEVX6t

I just woke up to 500 new Linkedin connections randomly overnight. How are you even supposed to sort through all this? It takes forever to mark "ignore" (or even "accept" at scale. I'm pretty sure they're all bots. This platform is so trash at sorting real potential clients from random people spam-selling services.

Here's the other ad we made for Qatar Airways in 14 hours, 30,000 feet in the air. These videos already have 5 million views on YouTube. Mass AI adoption isn't about to happen; it's already here. (Full breakdown of how we made it on my previous post or on my X)

My ad was just selected as a winner for the Grok Super Bowl ad contest! I think this counts as a Super Bowl ad? Kinda crazy! I only had 6 hours to make this and was limited to just using Grok Imagine for images and video. https://lnkd.in/gDWU256Z Congrats to Dave Clark and Andy Orsow for winning the other top spots! The last 12 months have been INSANE and I can't imagine what the rest of 2026 will look like. As always, I will keep you all in the loop with the latest tips and tricks. Make sure you're subscribed to my newsletter; there are some crazy new models coming out soon that I'm about to share with you! https://pjace.beehiiv.com/

Nintendo took 40 years to give us a Legend of Zelda movie. I made this in 5 days on a $300 budget. It looks like a $300M blockbuster. Let me show you how I made this in 5 simple steps inside of Freepik Please do yourself a favor and read this with images on X, it is infinitely easier to understand with reference images for the steps below: https://lnkd.in/grswUSz4 1 - Story I ditched the bright Nintendo look for a gritty, grounded reality. Zelda isn't just waiting in a castle; she's seeing her home get destroyed and has to act. I mapped the specific beats and scenes: Fear → Ruin → Rage → Capture → Confrontation 1.5 - Location Scouting I then mapped out all the iconic Hyrule locations that I wanted. I wanted them to be cinematic and chaotic. Generating a quick "mood board" helps you get consistent results when you dive into each scene next. 2: Image Generation The secret weapon is Nano Banana Pro inside @freepik. I use a specific prompt structure to generate a 2x2 grid. This gives me consistent lighting and multiple angles of the same scene instantly. Do not use 3x3 grids, they lose too much detail. Look closely and see how Freepik lets you add a character reference (Zelda), a style reference (I used a cinematic grid of shots) and other references (in this case, a specific wooden doll) This is the secret to consistent, cohesive visuals! You can get the exact 3000-character prompt that I use for consistent scenes. I’m sharing it in my newsletter tomorrow, so make sure you sign up (link at the end of the thread). Tip: Always set the aspect ratio to 21:9 for that cinematic feel. 🎬 4: Selection & Organization I generated about 10-20 grids per scene to get the perfect shots. Download grids Crop individual shots Once you have a bunch of cropped shots, arrange them as cohesively as possible. Generate way more than you need! Once you have your "selects", take the cropped shots back into Freepik and upscale using Nano Banana Pro. I have a special super long upscaler prompt that really works well (also in my newsletter) Step 5: Animation In Freepik, I used @Kling_ai  2.6 Prompt Tip: Describe exactly what physical action you want (e.g., "looking back in fear") and SPAM camera shake. Music & Sound I found an epic orchestral version of the Zelda theme by Epic Trailer Music UK. It made the trailer go from "good" to "great". -- My friend Matheus did the sound effects. DM me if you want his contact info, he's fantastic to work with! Be sure to give @freepik a try! If you want to see more breakdowns like this, check out my free newsletter. I break down the exact prompts I used for this video in this week's edition: https://pjace.beehiiv.com/ And that's it! If you enjoyed this thread, let me know what you think, give me a follow and let's stay in touch! I loved playing the Zelda games growing up, and can't wait for the real movie!

Ad Age ad of the week! Qatar Airways + Google + Starlink <3 14M views in a few days! So fun to make but we're gonna go back to taking 6-8 weeks to make ads from here on out 😅 Still catching up on sleep from that trip!

Let me show you how I just got 9M views on my viral Zelda trailer using Freepik • First, I wrote a simple story. Zelda sees the devastation firsthand. She’s scared, but she can’t ignore the call to adventure. • Then, I broke the story into locations and emotional beats for each scene. • After that, I generated cinematic 21:9 2x2 image grids in Freepik using Nano Banana Pro, with style and character references to keep everything consistent. • Next, I sorted my favorite shots in Figma and upscaled the selects back in Freepik. • Finally, I animated everything in Kling 2.6, spammed camera shake, and cut it to an epic Zelda theme. Comment “AI” and all send you a link to my special prompt technique!

So I got access to Seedance 2.0 today, and it's crazy. Unbelievably good. But it also represents a problem. Making AI films is easy now. I actually miss when it was hard. A couple months ago, it was easy to go viral with AI video. The tools were fragmented, nobody knew how to make them work. Now? Someone can type a cinematic prompt into Seedance and get an incredible action scene. The problem is differentiation. And this brings me to what I'm calling the new law of AI physics: The more inputs you put in correlates directly to the amount of inbound demand you'll get. The fewer inputs you use (one sentence, one prompt, maybe a reference image) the fewer opportunities you'll get. Because there's no you in that work. This is where I'm starting to side with the AI skeptics: if there's no unique act of creation, if you're not orchestrating something to your specific vision beyond just a reference image and a prompt, it's slop. It can be cool, but you won't get hired. You can't build a career on simple things that have no defensibility. Remember all those viral Sora creators with 10 million views? 99.9% never translated that into actual work. Clients didn't reach out because, if you can do it in one sentence, anyone can. How to Stand Out (When viral filmmaking is "easy") I'm pretty damn good at going viral and this is terrifying, even for me. But we can all adapt. We just need to pivot in two directions. Pivot #1: Branded Work Commercials are easier than ever to make. If you're doing funny viral videos that are brand-centric, you will consistently get paid. You can still work on short films. But if 70% of what you're creating is ad-related, I highly recommend it for building a solid financial foundation. Pivot #2: Narrative Storytelling (If You're a Masochist) If you love AI narrative filmmaking (fantasy, sci-fi) then level up. Don't just post little action clips. Post full sequences. Add storytelling. Add emotion. I'd recommend two-to-three-minute sequences that show character progression. There's a big opportunity in narrative, weekly episodic content, especially if you can incorporate brands. Creators like @NeuralViz and Gossip Goblin, Inc. have been building their universe on a weekly basis for the last year. That's very defensible. The Adaptation Strategy Here's a big strategy I'm employing: go for popular books and video games that aren't too popular. Make content where you can make a splash, go viral, then leverage that to talk to the authors/creators and say, "Hey, let me adapt your book or video game into an episodic series done with AI." The Bottom Line When using Seedream (or any tool), don't just yolo a prompt and post it. Ask yourself: Could anyone else make this but me? Does it reflect my voice? My style? My unique point of view? If so, then let it rip! Have fun and make some A5 Wagyu slop 🤣

Veo 3.1 + Nano Banana is insane Google’s new models let us make million-dollar looking ads for brands like Wander. Copy our entire process for making this ad below 👇 If you want to see visuals for each step, this is better viewed on X: https://lnkd.in/eYHjkjpN This ad was directed by Theo Dudley Concept: Tim Burton was a big inspiration for Theo growing up. Beetlejuice was a huge reference for this ad. He wanted a high-energy romp that wasn’t scary but never too safe either. Treatment: Writer Nate Dern and our clients at Wander had a great script and fun ideas, which we refined into three distinct characters we could continue to revisit and play with. Treatments help us align on tone and visuals before the greenlight. Image Generation: For start frames Theo collaborated with the super talented DDPs Olga Baranova and Julie Wieland to bring our characters and worlds together. We use Figma to organize shots and get a bird's eye view. Sometimes it takes hours just to get a single frame perfect. We used Nano banana for consistent characters and the blue ghost-like effect as a starting frame. Animation: We used all the tools: Veo, Seedance, 11 Labs, Kling, Nano-Banana, Seadream. Theo personally contributed performances via Runway, which a voice actor friend dubbed over. The girlfriend's line needed the nuanced performance only Act 2 could deliver. Revisions: Kevin Fermini edited this. Our first cut was pretty unhinged (lots of explosions and screaming family members) so we aligned back on tone with our client and found a middle ground that kept the fun but stayed appropriate Keep your eyes out for a possible director's cut! Want more BTS breakdowns? Join my newsletter: http://pjace.beehiiv.com Want an insane ad? Work with our agency, Genre.ai

"Live-action anime BTS" is blowing up right now on IG/TT. Tens of millions of views a day. Here's the prompt and process to make one yourself and get millions of views 👇 Pick your favorite anime and work with Chatgpt to create a series of prompts like this for 5-7 shots. Then plug it into Midjourney A behind-the-scenes photo on a live-action Dragon Ball Z film set, featuring a muscular actor in a highly detailed Goku costume, shirtless with battle damage makeup, spiked black hair, orange torn pants and blue wristbands, speaking with crew members on a blue soundstage floor, surrounded by practical rock set pieces and green screen walls, studio lighting rigs in the background, cinematic lighting, handheld DSLR-style photo, candid moment, high detail, natural color grading -- Midjourney I think is the most cinematic, though I LOVE Reve AI's chat interface -- it gives you three angles per prompt and really feels more conversational. You can then ask it to generate more characters from the show in different poses. -- Lastly, animate in your program of choice. Honestly, they're all pretty similar. Kling AI is a personal fav for this kind of content. Here's an example of the video in action: https://lnkd.in/ehdNgx7T Enjoy!

Underdressed old ladies are my thing? Here's my favorite section from our latest campaign with Levels Protein / Blake Niemann. 3 months back, Theo Dudley started working with our team on this campaign for Levels Protein (full 60sec ad in the comments) Here's how we made it (and AI copyright tips): 🧵👇 Concept: Levels came to us wanting to launch a new spokesperson in the form of Sir Bovine Cowley, a gentlemanly minotaur passionate about their all-natural protein. Theo Dudley and I both lift (pics available upon request), and protein is very important to us, so from the start, we knew this couldn’t feel like a gimmick. Cowley needed to feel iconic. Confident but warm. Noble, articulate, and memorable in just the right way. Levels as a brand is bold and unapologetic about their product, so the character had to carry that same clarity. Nate Dern wrote the script, and Theo brought a strong point of view early on about how far we could push the world and the tone without losing credibility. Copyrighting the character: We brought in concept artist Bart Bus at the very beginning to properly envision Cowley. Starting with a human-led illustration process gives the Levels team real creative ownership of their IP and allows us to be ultra-specific with facial structure, wardrobe, posture, and expression. We went through dozens of iterations before landing on the version you see in the ad. This Cowley struck the perfect balance of strength, wisdom, charm, and authority. World Building: Cowley’s home and office were essential. It needed to feel lived-in, intellectual, and distinctly his. Theo used @theworldlabs to create a 3D visualization of the office, generating a Gaussian splat environment we could explore. Screenshots from that world became reference points for precise camera angles and shot composition. From there, he collaborated with three digital DPs @olgabrnv, Evan Zissimopulos, and Ben Munoz, reviewing hundreds of frames until we had something that looked remarkably close to a ripomatic before animation even began. Voice: The voice of Cowley comes from the wickedly talented Rob Ricotta. You might remember him from our nationally televised Defy the Odds Kalshi spot. Product: For the product shots, we worked with an incredible team, Eduardo Pricinato and Elton Felix, from efxlab.xyz, who had a bespoke, proprietary AI workflow to make labels and packaging look picture-perfect. The results were immaculate, no garbled texts or melting logos, and the client approved those shots on first look. Shoutout to our producer Rod Orozco for tying everything together! Full breakdown with images on my X account: https://lnkd.in/gwuDb2TR