no one is going to save you, not god, not society, not love. learn to walk alone, or perish

Made this beauty ad with Wan 3.0 on @itsPolloAI @polloai_creator #polloai It took 4 generated to test different prompts and reference setups, costing an estimated 360 credits ($12) Definitely a learning process. With more precise prompts and better reference planning, I could probably get the result with fewer generations and spend fewer credits

made with API Seedance 2.5 via @CuddlerAI 14 clips × 15s. made music with Eleven v2.5, inspired by my own quote. ~$14 total. pretty worth it, if you ask me. might need another round of credits for the next one #cuddler #MusicVideo

#Malfoy Halloween outfit check, Which outfit suits her best? made with seedance 2.5 on @itsPolloAI @polloai_creator #polloai #MalfoyGirl

My first experiment creating a music video with kinetic typography on @AZ8official_JP #AZ8 Just exploring, learning, and figuring things out along the way. Turns out, it’s way harder than I thought. A first experiment, and definitely a lot more to learn

future does have flying cars made with seedance 2.5

Your AI slop entry here #AIMUSICVIDEOWAVE2026 @JACC_AI /FLOWER FEVER

What if you could go back in time and change just one moment? That question became the starting point of AIZENTO, an AI generated cinematic short film built around the concept of a time machine. I wanted to explore more than just time travel. The idea was to tell a story about memories, regret, and the moments we wish we could revisit. Because if you had the power to go back, would you really change the past, or would you learn to accept it? From the first concept to the final cut, I developed the story into six connected parts, each with its own visual direction, atmosphere, and emotional purpose. The goal was to make it feel like a cinematic film, not just a collection of AI generated clips. And getting there took 30+ generations. A lot of trial and error went into the process: testing prompts, refining camera movements, fixing character inconsistencies, maintaining the time machine's design, and regenerating scenes that didn't match the vision. Some attempts worked, others had to be scrapped entirely. Every generation was part of figuring out how to turn the idea in my head into something people could actually see and feel. This project reminded me that AI can create the images, but the concept, direction, storytelling, and persistence still matter. #AZ8 #AZ8AdTime @AZ8official_JP

Refining Yue Ning character details for visual consistency, polishing martial arts movements for smoother choreography, and enhancing expressions, physics, and scene continuity in a premium semi realistic 3D xianxia fantasy style Made on @CuddlerAI #cuddler

time to say alhamdulillah, because a few months ago, i was just someone trying to understand the world of ai video. today, i get to celebrate my first competition win. there were sacrifices, sleepless nights, and moments when i gave this craft almost everything i had. seeing it pay off feels surreal. grateful for how far i've come, and excited for how much more there is to learn

Honestly, I'm genuinely surprised. I stepped away for a while, but after coming back, @CuddlerAI didn't disappoint at all. This is what I've managed to create so far, and I'm actually impressed with the result Made on Cuddler #cuddler

You can try this one using wan3.0 Prompt bellow Create one 15-second UGC street vlog made of five chronological amateur moments with abrupt YouTube-era jump cuts, preserving the same young Japanese woman, face, long black bangs, white lightweight jacket, dark pleated skirt, silver headphones around her neck, black backpack, and wardrobe in every moment. Use @.Image1 strict identity and wardrobe reference as the strict identity and wardrobe reference. No redesign, hairstyle change, outfit change, missing headphones, missing backpack, or facial identity drift. Set it in Shibuya, Tokyo around 2010: Shibuya Scramble Crossing, Japanese storefronts, vintage Japanese ads, colorful period shop signs, Japanese pedestrians, taxis, bicycles, convenience stores, vending machines, and commercial side streets. No modern smartphones dominating, futuristic billboards, or 2020s aesthetic. Authentic early-2010s smartphone front-camera footage resembling an iPhone 4 or early Android YouTube upload: soft focus, washed-out colors, slightly overexposed daylight, digital noise, heavy compression, limited dynamic range, rolling-shutter wobble, autofocus hunting, exposure shifts toward signs and sky, crooked arm's-length selfie framing, natural walking bounce, no stabilization, no cinematic grading, no artificial film grain, no beauty filter, no glossy skin. Keep the low-resolution phone look inside the vertical canvas. 0–2s: Selfie while walking on a busy Shibuya sidewalk. Face fills frame; bangs move, headphones are around her neck, one backpack strap crosses the white jacket. She glances at the screen, gives a reserved smile, and says casually in English: Hey guys, I'm in Shibuya today! Natural bounce and muffled phone-mic voice. 2–4.5s: She physically flips the phone outward, not a digital transition. Reveal Shibuya Scramble Crossing: multidirectional crowds, Japanese signs, buses and taxis, dense buildings. Bright sky causes an exposure shift. Swing back too quickly to her face as she gives a small laugh; keep walking direction coherent. 4.5–7s: Stop beside a small Japanese street food shop. In selfie mode, raise a packaged Japanese snack close to the lens, briefly blocking part of her face; it is soft and blurry. Take a bite, chew naturally, and give a subtle thumbs-up with a faint smile. Pedestrians continue behind her. 7–9.5s: Walk to a colorful Japanese vending machine on a nearby side street. Show rows of canned drinks; insert coins, press one button, retrieve a cold can. Shaky tilt returns to her face as she holds it up and says: This one's cute! Open it; the can click is audible. Casual, slightly awkward gestures. 9.5–12s: Selfie approaching a pedestrian crossing. Crowd behind her; a cyclist passes at a safe distance. She follows it with her head, then looks back amused. Natural shake, motion blur, imperfect framing, and exposure adjustment. No collision. 12–15s: Turn into a narrower shopping street with Japanese accessory stores and small boutiques. Pause by a window of inexpensive hats and accessories, point at a funny hat, make a peace sign, and wave shyly. Say: Okay, bye! See you next time! Her thumb partly covers the front camera until an abrupt cut to black, as if she stopped recording. Audio: authentic early-2010s phone microphone with Shibuya crowd murmur, distant traffic, crossing sounds, footsteps, wind on mic, muffled casual speech, modest compression, mild laughter clipping, vending-machine beep, can-opening click, nearby shop ambience, and very faint Japanese pop music leaking from a storefront. No polished voice-over. Synchronize dialogue naturally. Preserve the same character, outfit, phone-camera aesthetic, and coherent Shibuya progression. The 2-second camera flip is physical. Keep imperfect cuts and framing mistakes intentional. @openart_ai #openart

ALL EYES TO KALIMANTAN The forests are burning. The skies are filled with toxic haze. The people suffer the consequences. This is more than an environmental crisis. It is a demand for justice, accountability, and action from those in power Please sue the Indonesian government made with wan 3.0 on @openart_ai #openart @PsyopAnime

I've been experimenting with developing my own character IP Aikoze, with Wan 3.0 on @openart_ai Turns out, the hard part isn't just generating the video, but keeping Aikoze face consistent across generations. There's always some tiny detail that changes lol Still in the trial and error phase, slowly figuring out a workflow that works #openart

An AI music video experiment Workflow OpenArt Director Mode + Sonnet 5 + MiniMax H3 Max Spent approximately 5,5k credits on this experiment. Testing the workflow, exploring visual direction, and learning how to bring a music concept into a full music video with AI Not perfect yet, but every experiment teaches me something new. More experiments coming soon #OpenArt @openart_ai

You can literally create another universe with AI, yet somehow you still don't know what to do with your weekend

Bikin ads AI buat smartphone fiktif AIZE, tapi treatmentnya dibikin kayak flagship commercial beneran. Fokusnya bukan ngomongin spek, tapi nunjukin "sejauh apa sih kamera HP bisa lihat?" lewat footage alam dari macro sampai wide landscape Workflow: Ide - storyboard - generate pakai Seedance 2.5 di ElevenLabs - pilih footage terbaik - edit #TheWorldsCatchiestAd #ElevenCreativeCompetitionVol1 @ElevenCreative @ElevenLabs

coba bikin video gaya UGC yang 100% AI generated. karakter, ekspresi, gerak kamera, sampe suaranya, semua digenerate, zero syuting. paling susah ternyata bukan visual, tapi bikin suaranya gak kedengeran kayak AI, walau gw ngerasa cukup bagus tapi masih banyak pr yang harus dibenerin bikinnya pake: az8.art/?source=cr_zeoaO4e5Y…

So the concept is an idol just walked off stage, still catching her breath, adrenaline still high, and she's just casually filming herself walking from the side of the stage back to her dressing room. She's talking about random stuff, complaining her ears are ringing, hyping about how good that moment on stage felt, then finally sitting down on a couch because her legs feel wrecked. Somewhere in the middle she sprays the perfume, but not in a "hey guys check this out" way, she just does it while talking about something else entirely, like it's already just a habit she has after every show. The thing I really wanted to avoid was that "ad pretending to be a vlog" feeling, so the dialogue is intentionally messy. She starts a sentence and switches topics, laughs for no clear reason, breathes mid sentence. Because that's how people actually talk when they're genuinely tired, not in neat, complete sentences. On the technical side, this is where the project gets a bit different. I generated it with Wan 3.0, which can do a single 30 second take, then chained two of those clips together using a last frame technique, so the final frame of clip 1 (her standing up right after spraying) becomes the exact starting point of clip 2. Even though it's technically two separate generations, it reads as one continuous 60 second shot with no visible cut. The walking route was also deliberately kept linear, from the stage wing, through a corridor, into the lounge, then out through a different path instead of backtracking, because if you make an AI regenerate the same room from a different angle, the geometry tends to glitch. Basically, the goal was for people to watch it and think "oh, this is just her vlogging" first, and only realize afterward "oh wait, this is AIZE Remnant" not the other way around. @OpenArt #OpenArtAdAwards @alibaba_cloud #Wan