Polka dots break AI video models.
Especially silk ones. In motion.
Small repeating patterns are where every video model loses control - by second three the dots drift, by second five the border warps, by second seven the brand wordmark scrambles into nonsense.
Here is the thing.
Lumoo just dropped their first creator contest for ELOÈ. The brief was specific - the exact polka-dot scarf must remain identical across the film. Pattern not modifiable. That was the disqualifier.
This is also my first editorial fashion film. Eighteen years directing character animation - the Kung Fu Panda days at DreamWorks taught me one thing. The character sheet decides everything. Editorial is a different room entirely - different silence, different breath. But the discipline carries.
So I treated it like an animation pipeline. Locked the scarf character sheet first. Storyboarded ten panels in ChatGPT Image. Then translated to Seedance 2.0 with the scarf as the anchor reference across every shot.
Five women. Five cities. One silk square.
Lagos. Tokyo. Copenhagen. Mumbai. Mexico City.
The film holds the pattern across all fifteen seconds. That is the win.
To Lumoo team - thank you for opening this door. More coming.
Comment LUMOO and I will share the full Seedance workflow right here 👇
#aifilmmaker #adaivision #editorialfilm
First Bicycle, Summer '91
A wordless short film. 141 seconds. Rural Hokkaido, August 1991.
A 16-year-old girl learning to ride her brother's bicycle through golden rice fields — to wave at a boy three villages away.
Directed entirely inside OpenArt Director — a new agentic filmmaking system that lets you shape a film through conversation. No stitching. No clip assembly. Just directing.
Huge thanks to Ruiyan Hu and the @openart_ai team for the early access. This tool is going to change how AI films get made.
Watch on YouTube - link in the comments. 🎬
AI is the camera. You still need the director.
#AIFilmmaking #OpenArtDirector #ADVision
Anyone can prompt motion.
Stillness is what you direct.
You watch enough action sequences and they all blur into one. Punches. Kicks. Quick cuts. Dust. Motion blur. The audience eyes glaze over - they have seen this a thousand times.
Then someone holds a beat.
The fighter lands the final strike and just stops. Fist out. Breath steady. The dust settles around him. The petals come back down. Your chest tightens because suddenly you can feel the room.
That is the shot. Not the strike. The pause after.
Here is the thing - AI video tools are obsessed with motion. They want to move every frame. Every petal. Every camera angle. Pushing forward, pushing forward, pushing forward. I learnt this at DreamWorks before I ever touched AI - the cinematographer's job is to know when to stop pushing.
Trust me guys, the same rule applies here.
This is 靜 JÌNG - Stillness. A Wing Chun study. 15 seconds.
No edits. No cuts. No color grade. One Seedance 2.0 generation - flawless straight out of the engine.
The film exists for the last 3 seconds where he goes still and the world catches up to him.
Made this with @floraai Seedance 2.0. Day 1 of a 3-part study - same film on Kling next, full director's comparison after. Going to show you where each engine wins and loses.
Comment STILL and I will send you the full breakdown on Substack 👇
#aifilmmaker #adaivision #florafauna #cinematography #flora
Skill beats power. But only if you survive the mess first.
Everyone thinks AI video is type a prompt, get a film. It is not. Not even close.
I made OVERTAKE - a white GT car racing a dark beast on a wet track. The drift overtake. The finish line. Looks clean now. It was not clean.
The cars flipped direction between shots. Two drivers showed up in one seat. The black car was winning when the story said it should lose. Every generation broke something new. Trust me guys, that part nobody talks about.
So I stopped fighting the prompts and started directing. I described the speed. The motion blur. The expressions through the helmet visors - eyes only, because that is where the acting lives.
Here is the thing - 18 years of animation taught me one skill that no AI tool can replace. You have to SEE the whole film before you make it. The pacing. The tension between two drivers. The camera shake. The raw racing movies playing in my head. The AI did not give me that. I brought that. The AI just rendered it.
And honestly speaking - not one clip came back perfect. You keep the 3 seconds that work. You throw the rest. That is not prompting. That is editing.
Comment "OVERTAKE" and I will share the full breakdown right here 👇
Made with @dreamina_ai 🙌
#aifilmmaker #adaivision #dreaminacpp
You know what makes a fight scene boring?
When both characters feel equally dangerous from the start. There is no tension in that. No reason to lean in.
The best action sequences in cinema - the ones that make your chest tighten and your jaw clench - they all follow the same invisible rule. You need to feel like someone is going to lose. You need to feel that hopelessness in your body before the comeback means anything.
That is what I built with ROAR.
Two fighters. A ruined temple. Fifteen seconds. And one arm that changes everything.
The tattooed warrior dominates - every punch pushes the hero backward. Every kick sends dust flying. And for a moment you genuinely believe the hero is done.
Then the arm ignites.
I shot this entirely in Seedance 2.0 using a Director Mode storyboard with Asura's Wrath camera dynamics - crash zooms, whip-pans, reaction cuts, freeze frames. Every single panel has a different camera angle. No two consecutive shots share the same lens distance.
Honestly speaking - I spent more time on the storyboard architecture than the actual generation. That is the part nobody talks about. The prompts are the last 10%. The directorial decisions are the other 90%.
Two versions. One anime. One photorealistic with Interstellar colour grading.
Made with @dreamina_ai Seedance 2.0 🙌
#aifilmmaker #adaivision #dreaminacpp #aianimation
Ever watched a race and felt it in your chest before the finish?
That tightness when the underdog falls behind. The breath you hold when they refuse to give up. The exhale when skill beats raw power.
That is what I tried to build. Frame by frame. Shot by shot. An AI racing short film - white GT car vs a dark beast of a machine. Wet track. Overcast sky. Two drivers with completely different philosophies.
One driver slams the gear with aggression. Dominance. The other sits behind in the spray - calm, reading the track, waiting for the corner that changes everything.
Here is the thing - I spent 18 years at studios like DreamWorks and Rockstar animating moments exactly like this. The drift. The overtake. That split second where skill catches brute force off guard. Now I am doing it entirely with AI.
Every shot was directed using Seedance 2.0 on Dreamina. Character sheets for both cars and drivers. Environment reference plates for colour grading. Expression beats through helmet visors - because you only see the eyes and trust me guys that is where all the acting lives.
This is not a tech demo. This is filmmaking. The camera is AI. The director is still human.
Comment RACE and I will share the full breakdown right here 👇
Made with @dreamina_ai
#aifilmmaker #adaivision #dreaminacpp #aianimation
Your character sheet is only doing half the job. And nobody is talking about the other half.
Here is what I mean. You generate a character - beautiful face, great outfit, perfect lighting. Then you need three more shots of the same person. Different angle. Different pose. Maybe a close-up of the face. And suddenly the AI gives you a completely different human being. Same prompt. Different person. Every single time.
The problem is not your prompt. The problem is you are treating every shot as a new generation instead of building from a locked reference sheet.
I built a system that takes any character - photoreal, painterly, stylized, animated - and generates a full character sheet. Three-view turnaround. Face panel. Colour palette. Same face, same lighting, same render style across every panel 🎬
Two versions of the prompt. One for ChatGPT. One for NanoBanana 2. Both lock seven variables so the character holds together no matter how many shots you generate after.
18 years in animation at DreamWorks, Marvel, Netflix taught me this - consistency is not a prompting problem. It is a reference problem. On real productions we never animated a single frame without a character sheet pinned to the wall. Same principle applies to AI. Better say, it is even more critical in AI because the model has no memory between generations 😉
The full prompt PDF is free. Comment SHEET and I will share the IG link right here 👇
Drop a 🔥 if character consistency has been killing your workflow
Workflow built on @ImagineArt_X 🙌
#aifilmmaker #characterdesign #imagineart #adaivision
This thought suddenly came into my mind: what if I can make something like Star Wars kind of clip in the theme of Anime?
Picture this - a girl in a hooded cape stands alone in a moonlit field. She ignites a silver energy blade and starts moving. Fast slashes. Dynamic cuts. Cape whipping through every frame.
But there is no enemy.
The camera pulls back slowly and then you see it - the trails she left in the air are not random. They form the Chinese character 道. The Way.
She was not fighting. She was writing.
The lightsaber was not a weapon but a brush! What if combat and calligraphy were the same thing?
That is it. That is the micro clip about.
Comment DAO if this one landed for you 👇
A micro film × @dreamina_ai
I wish you all a happy weekend 🩵
#aifilmmaker #adaivision #dreaminacpp #animestyle
Your AI dance videos look beautiful but nobody saves them.
Let me clarify what is actually happening.
Most AI creators right now are obsessed with smooth fluid dance content. Slow motion. Elegant motion blur. Beautiful flowing movement. Looks impressive on the first watch.
And then your eyes have nowhere to land.
The movement keeps flowing past. Your brain doesn't get a single frame to anchor onto. You watch it once. You think "nice." You scroll. You forget within 30 seconds.
Here is the thing - the MJ moves you remember most clearly are not the smooth parts. It is the freeze. The anti-gravity lean. The locked-in pose. The exact moment the body should not be at that angle.
Your brain registers it as a mistake before it registers it as genius. That half-second of confusion is where the save button gets pressed.
I have 18 years directing animation across Kung Fu Panda, GTA V, Marvel - and this is the same principle every single time. The freeze is the most expensive shot in any sequence. Because it is the one the audience remembers.
So when I built BEAT. - a 15-second AI breakdance film in @dreamina_ai Seedance 2.0 🙌 - I designed every shot around a freeze. Kick freeze. Baby freeze. Anti-gravity lean. Hollow back. Each shot under 2 seconds. Each shot ending or beginning on a locked-in pose.
I chose this approach because it needed to feel like a moment your scroll thumb actually stops on - not a video that washes over you.
Comment "BEAT" below and I will send you the full production breakdown - the failed character version, both prompts, the shot-by-shot Seedance prompts, and the DaVinci stitching workflow 🎬
Save this one if you build AI character content. The principle applies way beyond dance.
minimum a follow I deserve right? See you next time 👋
#aifilmmaker #adaivision #dreaminacpp #aianimation #breakdance #seedance #aishortfilm #characterdesign
Nobody teaches you the part where the character doesn't move.
That is actually where the whole film lives.
Let me clarify.
I spent 18 years animating characters for Kung Fu Panda, GTA V, Marvel - and the one thing every senior director drilled into me was this: the stillness before the move is where you feel the dancer.
Not the move itself. The stillness before it.
So when I built BEAT. - a 15 second AI micro film - I started from that principle. Shot one is just a hand on a hat brim. Face hidden. Total silence. Your eyes go to the sequins catching the light. Something in your chest waits. You don't know what is about to happen but you feel it is coming.
Then he drops into a breakdance freeze and the whole stage explodes.
Do you feel the difference between - "a dancer on stage" and "a boy holding completely still while the beat builds underneath him"?
One is a description. The other is a film.
I built this entire thing using @dreamina_ai Seedance 2.0 🙌 - character design, storyboard, shot by shot breakdance freeze prompting. Every shot under 2 seconds. Every move locked and precise.
The workflow took mistakes. A lot of them. Prompts that failed. Freezes that looked wrong. Character drift I had to fix three times.
Comment "BEAT" below and I will share the entire step by step breakdown - with the mistakes included - in a detailed video and a full Substack post. No filtered highlights. The actual process.
This one took time. I think it was worth it 🎬
Minimum a follow I deserve right? See you next time 👋
#aifilmmaker #adaivision #dreaminacpp #aianimation #breakdance #seedance #aishortfilm
He never noticed. Not once.
But here is the painful part - she will never know that the one time she looked away, he finally looked up.
That is the whole film. 30 seconds. No dialogue. No text on screen. No narration. Just two people in a coffee shop who almost connected but didn't.
Let me tell you why this hurts more than it should.
Because your brain fills in everything I did not show you. The backstory, the what-ifs, the "why didn't she just walk over." I gave you silence and your nervous system turned it into a full love story.
Better say - I gave you NOTHING and you felt EVERYTHING.
That is directing. Not showing more. Showing less and letting your audience's imagination do the heavy lifting.
I chose to keep this entirely in black and white because color would have been too generous. The only warmth in the entire film lives inside her fantasy - a golden daydream that lasts 10 seconds before a coffee cup slams her back to grey reality.
do you feel that slam? That exact moment when the warmth disappears and the cafe noise floods back in?
That is not a sound effect. That is emotional architecture. The fantasy exists ONLY so the reality hits harder. Remove the dream and the ending is just a girl drinking coffee. Keep the dream and suddenly she is heartbroken.
Same girl. Same coffee shop. But now something in your chest tightens.
Ask yourself this - when was the last time your AI film made someone feel something in their stomach? Not "cool visuals." Not impressed. Actually feel something uncomfortable that stayed with them?
Does not matter how cinematic you make it - if there is no emotional trap, your audience scrolls in 2 seconds. 100% Guaranteed.
Made this entirely with Seedance 2.0 on Dreamina. One character sheet. One master establishing image. Three prompts. The camera, the pacing, the golden fantasy shift - all directed through text. But the feeling? That is 18 years of knowing exactly where to put the silence 🎬
AI is the camera. You still need the director.
I share my directing secrets every week, minimum a follow I deserve right? 🔥
🙌 #DreaminaCPP @dreamina_ai
She wasn't supposed to win.
Everything about the matchup said no.
Kira walks into the cage at 173 cm against a heavyweight twice her size. Bald, moustached, yellow shorts, fury packed into every muscle. She has speed. He has weight. On paper this fight is over before the bell rings.
Then the bell rings.
She lands a clean three-hit combo on his face. He grins under his moustache and does not move. He throws one straight punch and she slides back into the cage fence. Then comes the body blow - in slow motion, his hook lands in her stomach, her body folds, hands dropping to her ribs, face twisting in a pain that has no language. She stumbles backward, grabs the chain-link to stay standing, head down, breathing through the kind of hurt that decides whether you stay in the fight or whether you go home 🥊
This is "Grit" - a micro film about the moment when the only thing you have left is the thing nobody can take.
Here is what 18 years inside DreamWorks, Marvel, Netflix taught me. Underdog stories do not work because the underdog wins. They work because the underdog refuses to leave. The win is bonus. The staying is the whole film. Everything else is just craft serving that single emotional beat 🎬
Built inside @floraai using Seedance 2.0 - five clips, one visual language, electric cyan rim light cutting through every frame. The kind of color consistency that traditionally took a colorist a full week. Flora held it across all five generations.
Comment "GRIT" and I will share the complete workflow in comment box below👇
#aifilmmaker #floracpp #flora #adaivision
Saw a random pic online and knew I had to make a movie out of it. 🎬
Meet "ONE FISH" — a 30-sec AI short of pure street market chaos, a stolen tuna, and one incredibly smug cat. 🐟🐈💨
Workflow dropping soon, stay tuned! 👀👇
@floraai CPP
#aivideo#seedance#flora
He stopped fighting. And the ocean did not notice.
"Still" — a 45 second AI micro film.
Built with Higgsfield Soul Cinema + @floraai Seedance 2.0.
"Workflow" in the replies 👇
#aifilmmaker#floracpp#adaivision
Thanks @dreamina_ai for the creative problem-solving workout. Every wall taught me something real. 🙏
I named the micro film - "The Catch"
Produced by Chat GPT Image 2 and Seedream 2.0
#Seedance2#DreaminaCPP#aifilmmaking