182 films from the archive — every one, grouped by era.
182 films
2019-2021
Jon Warlick
A wedding videographer from North Carolina creates what may be the first AI-generated video. Using NVIDIA's GauGAN, a tool that turns simple color-coded sketches into photorealistic landscape images, Warlick paints a landscape in GIMP, animates a camera movement through it in After Effects, exports every single frame, feeds each one individually through GauGAN, then reassembles the AI-transformed frames back into video. The result is flickering, dreamlike, and unmistakably alien. Three years before CogVideo, one person with patience and ingenuity found a way to make AI move.
2022
Magdalena Bay
One of the earliest music videos built on AI image synthesis. Directed by visual artist Felix Green, 'Dreamcatching' runs real live-action footage of the duo — Mica Tenenbaum and Matthew Lewin — through VQGAN+CLIP and GAN networks, letting the algorithm interpret and smear the imagery to the song's lyrics. The result looks, deliberately, like a computer's interpretation of dreams: melting, hallucinatory, months ahead of the wave. Proof that before text-to-video existed, artists were already directing neural networks frame by frame.
enigmatic_e
It's hard finding longer-form AI generated video from 2022, but this curated compilation by enigmatic_e (one of the early pioneers) touches on some great examples from this era. A time capsule of the raw experimentation that what was possible before major generative AI video tools existed.
2023
Netflix Japan / WIT Studio
Netflix Japan and WIT Studio release a short anime with AI-generated background art, making it the first time a major studio openly credits AI alongside human artists in a commercial release. The backgrounds were AI-generated, but the craft was human. It sparked a heated debate in the anime community about where the line should be drawn, but it also offered a glimpse of what was coming: AI as a production tool, not a replacement.
Uncanny Harry
An early AI short film created in the raw days of consumer AI video, when the tools were limited and the aesthetic was defined by surreal artifacts and dreamlike distortion. Uncanny Harry crafted a piece that felt intentional rather than accidental — an early signal that creators were finding artistic voice inside the noise.
Framer
A viral AI comedy starring Binance CEO CZ and Arnold Schwarzenegger — stitched together from Midjourney stills, DALL-E 2 generations, and D-ID face animation, composited in After Effects. The Schwarzenegger voice is from voice actor @Govournator. Racked up 3.4 million views and became one of the earliest examples of AI-generated content breaking through to mainstream internet audiences.
Reddit user chaindrop / ModelScope
The video that made AI video generation mainstream. Horrifying, hilarious, impossible to look away from. Generated with ModelScope's text-to-video tool. Millions of views. The 'this changes everything' moment for the general public.
Timmy (IXITimmyIXI)
An early Runway Gen-2 era short. A quiet, atmospheric forest pass that circulated in the first wave of usable text-to-video — foliage that moved without melting, light that held its shape. Modest by current standards, foundational in context.
Shane Willet
A comedic AI rap video where dogs and cats don't see Red or Blue — just territory. Made with Runway Gen-2 in its earliest days, Property showed that generative video didn't have to be serious or cinematic to land. The humor works because the AI aesthetic adds to the absurdity rather than fighting it.
visiblemaker
One of the first recognized fictional AI movie trailers. Inspired by Baz Luhrmann's visual language, The Great Catspy is a deliberate statement to the film industry: the AI revolution is here, watch closely. Every shot was generated from text prompts using a beta version of Runway Gen-2 (with one iconic Gen-1 shot mixed in). About 500 generations went into the final cut. Edited in Adobe Premiere, voices generated with ElevenLabs (no cloning), script polished with ChatGPT. Made in two days from initial human idea to finish.
After contacting their PenPals, they have now awakened the Nemesis. There is no turning back — you can't escape the mafia bass beats of the apocalypse. A music video that leans hard into the visual language of the Gen-2 era: warped textures, hallucinatory motion, and a dark energy that feels native to the tools. The aesthetic is unmistakably of its moment — defining what early Runway Gen-2 art looked and felt like.
Abel Art
An early AI short film from Abel Art, crafted in the first months after consumer AI video tools became available. A haunting, atmospheric piece that demonstrated the emotional range possible even with the limited tools of mid-2023. Abel Art would go on to become a key contributor to community projects like the T2 Remake.
Dale Williams (The Reel Robot)
One of the earliest AI-generated anime shorts, created during the first wave of AI filmmaking in mid-2023. Made with the tools available at the time: Midjourney for imagery and early Runway Gen-2 for video generation. Williams would go on to become one of the most visible AI animators in the community, with over 7 million views and multiple festival awards. Lost Marbles is where that journey began.
A comedic AI short from Abel Art that leans into the absurdist potential of early generative video. Made weeks after THE OLD LADY, it shows a creator already experimenting with tone and genre — moving from atmospheric drama to irreverent humor. A reminder that the early AI film community wasn't all serious experimentation; some of the best work was just fun.
An early attempt at building a visually consistent world using generative AI — mechanical creatures, clockwork landscapes, and impossible machines stitched into a cohesive short. Mechaniverse captures the distinct aesthetic of this period: dreamlike, slightly uncanny, rich with details, and pushing for narrative consistency when the tools barely supported it.
Aze Alter
An early atmospheric short from Aze Alter, who would become one of the most distinctive voices in the AI filmmaking community — known for densely textured, dystopian imagery that helped define the visual language of the medium in its first wave.
JeffSynthesized
One of the first animated shorts that really showed the potential of AI generative animation. JeffSynthesized demonstrated that the medium could move beyond surreal experimentation toward genuine animated storytelling — a turning point for what creators believed the tools could do.
Dave Clark
Set in the neon-lit streets of Los Angeles in 2065, a cybernetic ninja seeks vengeance after his wife's murder by the Yakuza, uncovering an underworld where ancient powers collide with futuristic warfare. Made with Midjourney and Runway Gen-2, this early short became a landmark: it later grew into a hybrid live-action/generative AI feature film developed at Promise Studios and showcased at the Cannes Film Market in May 2025, with stunt choreographers from John Wick. One of the most important early AI films.
Matan Cohen-Grumi
One of the world's first real AI commercials. Matan Cohen-Grumi — who would later join Runway and lead its creative work — showed that generative video wasn't just experimental art; it could deliver a polished, brand-ready ad that holds up next to traditional commercial production. A signal moment for AI in advertising.
AIandDesign
A distillation of the Gen-2 aesthetic. Euphoria doesn't fight the tool's psychedelic drift — the way forms bloom and dissolve mid-shot, textures breathe, and subjects melt into the next. It leans into it and treats that instability as the visual language rather than an artifact to be hidden. AIandDesign built a sustained piece out of the drift itself, closer to a music-video reverie than a narrative short — the strongest surviving argument that the early models had an aesthetic of their own, not just a set of limitations.
"I want to believe." A ninety-second UFO reverie made in response to the July 2023 congressional whistleblower hearing on aliens — news commentary turned around in the same news cycle, which almost nobody could do before these tools. Midjourney stills driven through Runway Gen-2 and cut in CapCut. Gen-2's melting, half-formed quality is a liability in most subjects and an asset in this one.
Four minutes of pure Gen-2 id. AIandDesign typed the name of deoxy.org — a bizarre psychedelic website he'd stumbled onto in the early 1990s, still online today — into Runway Gen-2 as a text prompt and, in his words, let it rip with the motion cranked to 10. No storyboard, no stills, no intervention: a single phrase handed to the model as an instruction to hallucinate. He describes the result as an acid trip back to his all-nighter rave past, and it doubles as an accidental record of what Gen-2 did when nobody steered it.
A near-three-minute music video built from Midjourney stills animated in Runway Gen-2 and assembled in CapCut — the entire Gen-2-era toolchain, all of it consumer-priced. Cyberpunk is the era's most-attempted genre for a reason: neon, rain, and chrome are exactly what a model trained on the internet renders best, and the style's own grammar absorbs the drift and flicker that would sink a naturalistic shot.
Dan Hammill & Jeff Wood
Hilario Abad, Javi Lería & Esteban Diba
Shang Zhu
Emad Khalid Mughal
Andrea Aureli & Gemma Costa
Seif Abdalla
Gabe Michael
Daniel Scott Cribbs
Riccardo Fusetti
Áron Filkey & Joss Fong
Jordan Rosenbloom
Kyle Goodrich
Shan He
Jessey Nelson & Dani Barker
Sam Lawton
A short, AI-assisted film depicting a British gangster — and one of the first works to demonstrate real scene blocking and dialogue potential in AI filmmaking. Before this, AI shorts mostly leaned on montage, voiceover, or music to hide the medium's weaknesses. Uncanny Harry showed you could stage a conversation, hold on a character's face, and let an AI-generated performance carry a beat.
2024
FERA
A Cyberpunk 2077-inspired AI short that reimagines the game's world through generative video. FERA builds a moody, atmospheric narrative around the character Takemura, blending game aesthetics with AI-generated visuals. An early example of AI filmmaking drawing from gaming IP to create something that feels like a lost cutscene from a parallel universe.
Shy Kids
One of the first films made with OpenAI's Sora. A 1:21 short about a person with a balloon for a head. Goes massively viral as part of Sora's public debut. Toronto collective Shy Kids becomes the face of early Sora filmmaking.
UNVEIL
YZA Voku
Gurkan Atakan
Nelson Quimbay, Wilman Aguilera Charry & Juan Felipe Ortiz
Guillermo Miranda & Javier De La Chica
Evan Siegal
Mind Wank
Rich Klein
(this is one of my animated short films)
My Gen:48 2nd Edition finalist film and my first real experience working with generative video. I worked with my daughter on the story using Runway Gen-2 for all the video generations. The film explores themes of innocence and perception through a fairy tale (with a twist).
Sway Molina & Nem Perez / Storyblocker Studios
A feature-length AI parody remake of Terminator 2: Judgment Day, produced by over 50 AI artists and filmmakers. In this version, humanity must fight back against a world ruled by ChatGPT. Three months in the making, it premiered at the Landmark NuArt Theatre in Hollywood — one of the earliest community-driven, feature-length AI films to get a theatrical screening. A statement piece for the AI filmmaking community: proof that dozens of creators scattered across the internet could come together and ship a full-length movie.
Daniel Antebi
Léo Cannone
Samuel Schrag
Junie Lau
Hooroo Jackson
The first AI-animated feature film, and the first AI anime feature — 87 minutes, made by one person for $405. Betty Gray, a college sophomore steeped in gothic romance, falls for Duchamps De Ve, the leader of a banned occult society, and then has to defend him when a rival from his past accuses him of being a vampire and the town puts the charge on trial. Hooroo Jackson built it between February and July 2024, using AI for every layer — imagery, animation, performances, sound, music, editing — and treating the model as a writing partner rather than a shot generator. The finished film is assembled from more than 17,000 generated images. The economics are the part that's hard to argue with: $405 and six months against the several hundred thousand dollars and multi-person crews that every comparable project has needed since. It premiered July 3 and reached Prime Video on July 26, making it the first AI feature to land a major streaming license, with a Blu-ray following in August. Reviews were genuinely mixed — critics credited Jackson's direction while noting the film talks far more than it moves, and that a vampire trial contains remarkably few vampires. That's a fair verdict and beside the historical point: a feature-length animated film that one person generated end to end, distributed on a major platform, in the first half of 2024.
A visually rich AI short from Abel Art that follows a traveler through dreamlike landscapes. By mid-2024, the tools had matured enough to match Abel Art's ambition — the result is a piece with real cinematic scope, atmospheric depth, and a visual confidence that reflects over a year of working with generative video.
Various creators
AI-generated videos of cats performing Olympic-style dives and gymnastics routines, created using the Hailuo model by MiniMax. Racks up millions of views and gets remixed endlessly across TikTok, YouTube, and Twitter/X. Proves AI video generation can be pure, joyful entertainment — no narrative ambition required.
A 5-minute short telling the story of the 320th Barrage Balloon Battalion, the only all-Black unit to storm the beach during D-Day. Created 100% with image-to-video and text-to-video generative AI using Runway and Luma. Goes viral with millions of views and leads to feature film offers from major production houses. Clark proves AI can tell stories that Hollywood won't.
Sway Molina
A surreal, animated series by Sway Molina exploring beauty, horror, and the spaces between. The Yolky Universe was one of the most distinctive (and memorable to me) aesthetic signatures in early AI animation — weird, vivid, unmistakably handcrafted, even though it was AI-generated. As Molina put it: "The mind that's capable of creating beautiful worlds is the same mind that can create horrific ones for themselves. Always choose beauty and when in doubt — return to reality. It's not so bad here."
Waymark
Survivors in a frozen post-apocalyptic world follow a mysterious signal up a mountain in Antarctica. The complete 23-minute cut, merging Parts One and Two, premiered September 15, 2024 at the Creator RenAIssance Programme in Singapore — but the film's real significance is that its first half is a fossil of a moment that lasted about a year. Part One went out in June 2023 as a 12-minute short revealed exclusively through MIT Technology Review, and it was made without a video model, because there wasn't one worth using. Director Josh Rubin fed his script to DALL-E 2 and generated every single shot as a still, then animated those stills with D-ID — a tool that makes eyes blink and lips move. That's the whole pipeline. Waymark's Stephen Parker described the turn that made it work as giving up: they stopped fighting for photographic accuracy and started leaning into the weirdness DALL-E gave them. Rubin was blunt about the limits, saying it was a struggle to get emotional responses out of the faces. By the time Part Two was made, the tools had changed underneath the project, and the film wears that: Part One is uncanny and otherworldly, Part Two turns photorealistic. JeffSynthesized joined for the second half to work on character and landscape continuity. Watching them back to back is an unusually clean before-and-after of the gap between prompting stills into motion and generating motion outright.
AiMation Studios (Tom Paton)
Earth's last survivors send robots to farm a new world called Oracle. Cru, the newest model, has his purpose rewritten by a pod carrying the last human baby. Written, produced, and directed by Tom Paton, it runs 88 minutes, went up free on YouTube, and is the first release from AiMation Studios — a company built specifically to make features this way. The production numbers are the argument. Nine people and roughly $8,000 a minute, about $696,000 for the whole film — a rounding error against any animated feature that plays like one, and the reason Paton called it his Toy Story moment and insisted it wasn't hypothetical. What the film actually demonstrates is narrower and more honest than the pitch: robots render beautifully and humans don't. The human-presenting character POP is where the seams show, which tracks with every other generative project of 2024 and is exactly why a cast of machines was the right story to attempt. What sets it apart is the scale it was attempted at: a crew, a real budget, a voice cast, and a slate of ten more projects behind it. Most generative work of this period was one person proving a point. This was an attempt at building an industry.
OLD NEW RARE
Federico Papasodero, Emidio Battipaglia & Simone Adami
John Kelly
Holden Boyles
Neural Viz
Thomas Frenk & Clifny Alves de Lima
BLVCKL!GHT
Whispers in the Dark: Descent into Madness is a 30-episode series turned feature-length film that explores cosmic and psychological horror. At one hour and 43 minutes, it's an early proof of long-form consistency in AI filmmaking — sustaining characters, tone, and dread across a runtime longer than most theatrical features, at a time when most generative work still lived in ten-second clips. Soundtrack and score created with Udio, sound effects from Epidemic Sound, and voiceovers via ElevenLabs. Animation was handled with Luma Dream Machine and Hailuo AI, while imagery was generated with Midjourney and LetzAI.
One of my early animated shorts, made with Runway Gen-3. I wanted to explore themes of light, hope, and the lengths we'll go to preserve what matters most. A simple story, but one that felt meaningful to make.
Edmond Yang, Simon Valentine & Alex Holm
A poignant short film about an elderly inmate's final wish, created in just 48 hours as part of the Cinema Synthetica Norway workshop. Script by acclaimed novelist Erlend Loe, directed by Simon Valentine, with AI visuals and animation by Edmond Yang. Won 1st Place in the Narrative Category at Project Odyssey 2025 and was featured in Forbes. A powerful example of how AI tools can accelerate traditional filmmaking workflows without replacing the human elements of storytelling, performance, and craft.
Coca-Cola / Silverside AI / Secret Level / Wild Card
Coca-Cola remakes its iconic holiday ad using Gen-3 Alpha, Luma Dream Machine, and Kling. The result is polarizing: praised by some for ambition, called soulless by others. Sparks a major debate about brands using AI to replace traditional production. The backlash proves the AI filmmaking community cares deeply about craft and authenticity.
Ezra Li
Osvaldo Rivera Vazquez
Faith Cho
Prasad Gori, Anurag Tiwari & Sagar Chogale
Nicholas Flandro
Mahmoud Salah & Karim Moftah
Javier De La Chica & Guillermo Miranda
Matthew Berger, Kavin Kapoor & Ishan Modi
Alex Naghavi, Dan Taplin & Sara DeCou
Linus Mariscal Carulla, Max Ponce Pons & Dru Wu
Jim HuiHui
A 10-minute noir animation set in the neon-lit streets of Sichuan, China, following Xu Xia's struggle between conscience and survival. 242 shots, nearly three months of work, created largely by one person with AI. Midjourney generated 4,164 images, Kling produced 984 video clips, Pika added 693 more. Won Best Animated Short Gold at the Independent Shorts Awards. Featured at the AI Artist Festival and VACAT Awards. A landmark achievement in solo AI filmmaking: one creator, multiple AI tools orchestrated together, producing a film with the depth and craft of a studio production.
Jason Zada
Every shot generated text-to-video with Veo 2. A 1:56 bank heist and car chase. 1.1M views on the original X post, 10M+ across platforms. Proves that a single creator with taste and the right model can make something that looks like a studio production.
Sci-fi short based on Dave Clark's father's true story of being Black in 1960s America. Made with Midjourney, Runway, and ElevenLabs. 110K+ views on X. Goes viral and leads to feature film offers from major production houses. Personal storytelling meets AI tools.
Dustin Hollywood
Billed as the world's first 100% AI-crafted feature-length film. Set in a future where AI has surpassed human creativity, the protagonist Tanaka stands as the last artist wielding a brush in a world dominated by digital perfection. Written and directed by Dustin Hollywood, founder of NAKID Magazine.
PJ Accetturo
A cinematic AI concept film depicting a Japanese village preparing for war. Made from scratch in just 3 hours using only $50 in Kling credits. PJ Accetturo showed that cinematic quality, speed, and low cost no longer had to be trade-offs in AI filmmaking.
2025
Henry Daubrez
A tale of two lonely souls finding connection. Curated from 1,700+ AI-generated shots into a coherent narrative. Made with Veo 2. Named one of the 9 Best AI Animated Films of 2025 by Curious Refuge. A masterclass in AI film curation — the human taste is the tool.
An anime-style AI short from early 2025, showcasing Williams' signature style of character-driven narrative animation. Created during a highly productive period where Williams was collaborating with Promise AI on concept development and building his reputation as one of the leading AI anime filmmakers.
A feature-length AI-generated narrative, approximately one hour long, that transports viewers into another dimension through bizarre, surreal worldbuilding. Drawing from 1980s boundary-pushing shows like Ren and Stimpy and Pee-Wee's Playhouse, BLVCKL!GHT creates a sustained fever dream that proves AI filmmaking can work at feature length. Selected by Curious Refuge as one of the 9 Best AI Animated Films of 2025.
An AI anime short depicting two surviving tribes fighting over Earth's last heat source as global temperatures drop to deadly levels. Won 1st Place in Curious Refuge's AI Animation Contest. Described as 'easily one of the best AI anime shorts out there.' Williams' work has amassed over 7 million views across social media and earned multiple festival awards.
A surreal, darkly comedic piece set in an interdimensional water park where visitors encounter reality-bending experiences and may never truly leave. Part of BLVCKL!GHT's interconnected 'Route 47 Television' universe, an Adult Swim-inspired network of AI-generated shows. The concept grew into a recurring series with elaborate worldbuilding, its own website, and characters that cross between BLVCKL!GHT's other shows.
A twisted dating show parody where human contestants choose between grotesque cryptid bachelors like Mothman, a shapeshifting Skinwalker, and the recently divorced Jersey Devil. BLVCKL!GHT's breakout hit and most viral property, achieving tens of millions of views across TikTok and YouTube. The claymation-esque visual style blends horror, surrealism, and comedy. Proves that AI-generated serialized content can find massive audiences on short-form platforms.
Paul Trillo, Paul Flores & Cuco
A poetic, music-driven tribute to Los Angeles directed by Paul Trillo — one of the most respected names bridging traditional filmmaking and AI tools. The film weaves generative imagery into a love letter that feels both deeply personal and unmistakably AI-native. Trillo's work consistently demonstrates that the best AI films come from directors who already know how to tell stories.
The tri-state area's only combination taxidermy studio and ice cream parlor. Sad Steve preserves deceased pets while his manic assistant Sprinkles serves up questionably-sourced frozen treats with flavors like 'Rocky Road Kill.' Part of BLVCKL!GHT's Route 47 Television universe, explicitly referencing characters from Gorbo's Swim Hole and The Mysterious Forest. The cross-series worldbuilding is a distinctive achievement in AI-generated content.
Hal Watmough
First place winner at the Curious Refuge AI Animation Contest 2025. An unforgettable emotional narrative that showcases what generative animation can do when craft and feeling come first. Part of a growing wave of AI animation contests proving the medium has its own artistic identity.
A music-driven AI short by FERA that pairs striking generative visuals with a hypnotic soundtrack. Fluid, reflective imagery merges with rhythm to create a piece that feels more like a music video from the future than a traditional short film — an example of AI filmmaking finding its own native format.
Uno Particular
A lo-fi, VHS-flavored fever dream that plays like a lost transmission from a public access channel in another dimension. Uno Particular leans into the glitchy, uncanny qualities of AI video rather than fighting them. The result feels like outsider art filtered through a broken CRT.
@bigfootvlogs / Various creators
Just one day after Google launches Veo 3 with native audio and dialogue, TikToker @bigfootvlogs posts an AI-generated video of Bigfoot casually vlogging in the woods: introducing himself, showing what he had for breakfast, taking a dip in a river, foraging for berries. It racks up over 6 million views in weeks. The format explodes across TikTok, X, and YouTube, with creators building entire character universes around talking cryptids: Yeti cooking shows, Mothman relationship advice, Loch Ness Monster nature documentaries. The trend proves two things: Veo 3's native audio generation is convincing enough to make fictional characters feel present and alive, and the AI filmmaking community has a sense of humor.
A thriller-style AI short featuring a hitman entering a room, crafted to feel like authentic Hollywood cinema. Clark challenged himself to create a coherent cinematic narrative using Google Veo 3 and Google Flow, achieving a level of shot-to-shot visual consistency rarely seen in AI film at the time. The film showed what Veo 3 looked like in the hands of an experienced filmmaker: remarkably realistic, Hollywood-grade footage with coherent lighting and staging across cuts.
A nostalgic coming-of-age memory set to motion. Made with Google Imagen 3 and Veo 2 via Flow. Premiered at the 2025 Tribeca Festival. Created for Google's I/O 2025 developer event. AI film reaches a prestigious traditional film festival.
Runway
The third Runway AI Film Festival, screened at Lincoln Center in partnership with IMAX. Over 6,000 submissions judged by Gaspar Noé, Harmony Korine, and Jane Rosenthal. 'Total Pixel Space' by Jacob Adler wins Grand Prix. 'JAILBIRD' by Andrew Salter takes Gold. 'ONE' by Ricardo Villavicencio & Edward Saatchi earns Silver. 'Distance Between Two Points Of Me' by Illya Dutsyk, Bogdan Nesvit & Anatolii Kasianov receives Honoree. The remaining finalists, 'More Tears Than Harm' (Herinarivo Rakotomanana), 'Fragments Of Nowhere' (Vallée Duhamel), 'RŌHKI - A Million Trillion Pathways' (Hachi and IO), '6000 Lies' (Simon Rieth), 'Editorial' (Riccardo Fusetti), and 'Emergence' (Maddie Hong), represent the global breadth of AI filmmaking. The jump from 300 submissions in 2023 to 6,000+ in 2025 tells the whole story.
Jacob Adler
Andrew Salter
Ricardo Villavicencio & Edward Saatchi
Illya Dutsyk, Bogdan Nesvit & Anatolii Kasianov
Herinarivo Rakotomanana
Vallée Duhamel
Hachi and IO
Simon Rieth
Maddie Hong
Kartikye Kashyap
Ray Jaffer
Jaemin Yoo
Midday Visual House
Thomas Frenk, Eduardo Maluf & Yan Dotta
Ron Baranov
Jaehong An & Haeyun Shin
This one’s different — my first real attempt at an animated comedy. Steve is a sarcastic, sophisticated cat with big opinions and a refined palate. He drinks lattes, avoids litter boxes, and judges other cats. I wanted to see if generative AI could sustain a recurring character across episodes. Turns out, it can.
The Butcher's Brain
A stylish, high-energy AI short blending drifting cars, split-screen editing, and kinetic visuals. The Butcher's Brain combines multiple video models — Veo, Kling, and Wan — with ElevenLabs for audio, showcasing the multi-model workflow that became standard practice in late 2025: pick the best tool for each shot rather than committing to a single pipeline.
An AI short that Williams described as 'probably the best thing I've ever done.' Created during the late 2025 period when more advanced AI video tools were available. Williams' work has been exhibited internationally and he has spoken at events including Pacific Film as a featured AI filmmaker.
My experiment in psychological horror with Veo 3 (Twighlight Zone style). A man visits a local theater for an ordinary movie night, only to discover the eyes on screen might not be acting. As he returns home, reality begins to blur.
MetaPuppet
A sci-fi AI short from MetaPuppet, one of the prolific creators in the AI filmmaking community. KYRA showcases the cinematic quality achievable with late-2025 generative video tools — polished visuals, coherent characters, and atmospheric worldbuilding from a solo creator.
A character who fails at everything and just can't turn his life around. Hopeless Steve became the breakout character from Framer's Cartoon Hero workflow — a lovably pathetic everyman whose animated misadventures proved that AI-generated cartoons could carry real comedic timing and emotional resonance. It also demonstrated that consistent animated characters with personality could be created using generative AI.
A short film exploring what happens when the last traditional artist confronts a world that has moved on. The Butcher's Brain leans into Google's ecosystem — Veo 3 for video, Flow for editing, Gemini for planning — plus ElevenLabs for voice and Suno for music. A full AI production pipeline from a single creator, and a thematic mirror to the tool debate itself.
The sequel to 'The Cleaner,' continuing the cinematic hitman narrative. Clark returns to the character and world he established in the first film, building a longer and more complex story. The fact that an AI filmmaker is building sequels and recurring characters marks a maturation of the medium from one-off experiments to serialized storytelling.
2026
Darren Aronofsky / Primordial Soup
The Oscar-nominated director of Requiem for a Dream and Black Swan releases an AI-generated animated series on TIME's YouTube, chronicling the American Revolution on its 250th anniversary. Made with Google DeepMind. SAG-AFTRA voice cast and original score. Each episode drops on the exact anniversary of the historical event it depicts. The reception is brutal: viewers and filmmakers call it 'AI slop.' Significant not for quality but for what it represents — a major Hollywood director founding an AI studio (Primordial Soup) and going all-in on AI animation.
The Dor Brothers
Marketed as a '$200M movie made in 24 hours with AI.' A nearly 3-minute sci-fi short that racks up 50M+ views across platforms. Featured on Joe Rogan. Made with Seedance 2.0. Divisive reception: praised for visuals, criticized for prioritizing spectacle over story. The Berlin-based brothers become the most-viewed AI filmmakers in the world.
A meditation on chronic resilience and the beauty of survival. Made with Google Flow and other tools during Daubrez's residency at Google Labs. In consideration for 2026 Tribeca Festival. One of the longest and most personal AI-generated films, pushing the format toward true short film territory.
A concept trailer for a new project in Williams' signature anime-influenced style, showcasing the capabilities of early 2026 AI video tools. Williams has stated the industry is 'less than a year out from commercially viable AI films,' and concept trailers like this represent steps toward that goal.
An animated series pilot inspired by Arthurian legends in an 80s sci-fi setting. Made with Seedance 2.0, Google Flow, Gemini, Adobe Premiere/AE, ElevenLabs, and Suno. ~4 weeks of work, ~$1,000 in Seedance credits. Pioneered an 'omnireference' workflow. The future of serialized AI animation.
Léo Cannone & New Forest Film
A nostalgic masterpiece about childhood summers. Set in Corsica in the '90s, a young boy unboxes his first camera and documents a summer with his grandparents — what begins as play becomes a chronicle of small wonders and everyday magic. France-based Cannone proves AI film can be deeply emotional, and the festival circuit agreed: Costa Verde swept the AI International Film Festival in March 2026 (Best AI Film (Jury) plus three audience awards), then took Best AI Fantasy Film and Best WAiFF Film at the World AI Film Festival 2026, and was named an Honoree at Runway's AIFF 2026. One of the most decorated AI shorts of the year.
Theoretically Media (Tim Simmons)
A hyper-stylized action thriller about vengeance and katanas. Every shot generated with Seedance 2.0. Created by Tim Simmons of Theoretically Media, one of the most prominent AI filmmaking voices on YouTube and a former Hollywood insider (Sony Animation). Simmons has been decoding AI tools for creators since 2022, and Dragon Blue shows what happens when deep technical knowledge meets cinematic ambition.
Gossip Goblin
Gossip Goblin premieres The Patchwright, a moody AI-driven short that lands with the kind of polish and worldbuilding that's becoming the new baseline for independent AI filmmaking in 2026. PJ Ace publishes a detailed behind-the-scenes breakdown of how it was made, walking through the generation stack, prompting choices, and edit decisions — the kind of open-process post that's become a quiet hallmark of the community.
Robert Gaudette
The film that won the $50,000 Grand Prix at Runway's AIFF 2026. Robert Gaudette's 7:50 short follows Marcel Dupont — a man whose face only a mother could love, and whose mother told him every morning for thirty-eight years that he was beautiful. Marcel moves through Paris with complete confidence; the world moves away from him and he does not notice. Every evening he dances alone in his apartment, one hand raised for a partner who isn't there. One Halloween a woman sits down beside him in the park and does not leave — she has, it turns out, been listening to him dance through an open window for years, never ringing the bell. It's a film about two lonely people, one act of courage, and an accordion player who played exactly the right song at exactly the right moment. It's also a masterclass in the 2026 generative pipeline as a single authored vision: every frame prompted, character consistency held with AI character sheets, narration and dialogue voiced through ElevenLabs, a melancholic score composed in Suno, and AI-built sound design — AI as collaborator rather than shortcut. Beyond the AIFF Grand Prix, it also won the Best Director (Short Film) Award at the Portugal Indie Film Festival — a general festival and category not restricted to AI films, besting traditionally made shorts in a rare crossover recognition for a fully generative work.
Guillermo Miranda — Contanimation
Posted to X on the Big Pitch Contest's final day, Sincitium goes viral immediately — a Contanimation concept trailer (Javier De La Chica and Guillermo Miranda) that becomes the runaway favorite a full week before Runway names it the Grand Prize winner on May 12. The post itself, not just the prize, is the moment: a small Spanish animation studio drops a concept trailer that competes with anything coming out of LA, and the comments pile up with creators saying this is what the new economics of an AI-native pitch actually look like. Sincitium becomes the first $50,000 Big Pitch champion and quietly sets the visual vocabulary every copycat contest entry will reference next.
Kavan the Kid
Marko Slavnic
Two pigeons stand on a rooftop arguing over what an odd remote control device actually is, voice acting and all. Marko Slavnic makes the clip in Runway in a few hours from a handful of generations, posts it to X, and within days it's at 6.5 million views with people comparing it to Disney-level animation. Paige Bailey calls it "absolutely incredible" and says it "legitimately looks like a clip from a massive studio production." It also instantly polarizes — artist Twitter pushes back hard on the training-data ethics behind tools that can produce something this good this fast. The Pigeons Arguing moment becomes the spring 2026 distillation of the entire AI-vs-craft debate, all of it carried by one creator, one X post, and two funny birds.
Aiden Guo
An anime piece created in collaboration with Morphic Studio, featuring professional Japanese voice actors. Guo — a Sheridan College grad with credits on One Piece episode 1024 and Dandadan episode 4 — directs the latest entry in his Mars Series, co-written and produced with Jason W. and Polygon co-founder Jaynti Kanani, among others. Voice actors Jiena Gaim and Teru Ikuta bring the characters to life across sci-fi colony adventures — tense arrivals, daily triumphs, train sequences that rival traditional studio work. Fans pointed to it as proof that AI tools shine in expert hands, even with minor glitches: the craft of a working animator shows through.
A near-future short about a tattoo artist named Picasso, scraping by in a city that no longer has work for people like him — art, poetry, and music have been "fully automated for optimal productivity." When an emigration office offers him a one-way ticket to Mars on a special shuttle reserved for painters, poets, and musicians, he trades the ticket to settle a debt and stay. The other man takes it; the film ends at the Mars Embarcation Station with "Batch 472 processed." A film about artists being made obsolete by automation, made with AI — the irony is the point.
J Felipe Orozco
A harrowing story of loss at sea — two young boys facing impossible conditions and constant danger. Made by Orozco solo, in three weeks, and chosen as the headline film for Runway's Project Luxo. The behind-the-scenes thread is the part the community will pass around: a working filmmaker showing every seam of how the film was built, the kind of generosity that compresses a generation of learning into a single post.
LatentDiffusion
Wonder Studios
An extraterrestrial force has arrived to challenge the planet's greatest. A universe Olympics, where Earth's species must prove their strength and intelligence against an otherworldly presence. Directed by BAFTA winner Hal Watmough for Wonder Studios.
Dorian & Daniel
Lucas Levitan & Fabián Jiménez
Minju Choi
Taher Dhanerawala
A quiet six-and-a-half-minute short built around a lost tiffin and one small human gesture that turns out to carry the weight of an entire life. Directed by Taher Dhanerawala, written and edited by Michael McGruther, it's an AI-native film about kindness, memory, dignity, and the particular intimacy of sharing food — an Indian story told at a register most generative work doesn't attempt. The dominant mode in AI short film is spectacle: creatures, battles, impossible camera moves, the things the tools make easy. The Last Tiffin goes the other way and stakes everything on restraint and emotional legibility, trusting an everyday object to hold the memory of a person. That's a harder thing to land with generative tools, which reward the loud and the strange, and it's the reason the film is worth marking.
Ryan Phillips opens a British post-apocalyptic Lovecraftian horror-action series with a first episode that runs nearly seven minutes — a long hold for a genre that usually gets exactly one atmospheric shot before an AI short runs out of coherence. Phillips is one of the longest-serving filmmakers in this timeline: Digital Deity and The Cold Call are here from 2023, back when staging a conversation and holding on a character's face was itself the achievement. Three years on he's building a serialized world instead of a proof of concept. The craft background is the reason it holds: a BAFTA-nominated director with two decades in television, including lead director work for the BBC, who studied under Ken Russell before founding Uncanny Harry Productions to make AI cinema grounded in traditional filmmaking discipline. Horror is a useful stress test for the medium — it lives on dread, restraint, and what you withhold, none of which generative tools hand you for free.
At 21 minutes, Plume is Dave Clark's longest fully generative AI film to date — a sustained test of whether the AI-native pipeline he has been refining across shorts like Borrowing Time and Tairell Isn't Real can carry a single authored vision at genuine short-feature length. For a creator who first went viral with tightly edited two-minute pieces, stretching to 21 continuous minutes of fully generated footage marks a personal high-water mark for endurance and narrative control in the format.
A two-and-a-half-minute neo-noir about a hitman whose final contract is Death himself, written and directed by Tim Simmons and generated end to end with Seedance 2.5 on Dreamina — three days after the model went generally available. It's the sequel to Dragon Blue, Simmons's Seedance 2.0 short from March, and the claim he makes for it is narrower and more interesting than the usual capability-demo pitch: not the frames, the acting. Micro-expressions, pauses, the timing of a line reading. "No actors. No set. No cameras," as he puts it — and the thing he says finally happened is that he stopped scanning for glitches and just watched the performances. Performance has been the last honest holdout in generative video; motion, lighting, and consistency all fell first, while faces kept landing in the gap between a still and a person. Simmons, a former Sony Animation insider who has been decoding these tools for creators since 2022, is a credible witness on where that line currently sits. Worth reading with one caveat in view: the film was published as a showcase for Dreamina's Seedance 2.5 rollout, with campaign-tagged links to the platform.
Ryan Phillips built this scene specifically to test the thing that changes how a scene can be written: Seedance 2.5 generates 30-second shots and extends them near-seamlessly — with a little post blending — into far longer sequences. The result runs a full minute, and length is the whole point. Every generative model before this one imposed a five-to-ten-second grammar on its users, which is why so much AI film looks like a trailer: cut, cut, cut, because the tool couldn't hold a shot long enough to let anything develop. A thirty-second take is a different instrument. You can let a beat land, hold on a face after the line instead of cutting away from it, stage a scene in real time and let the pacing come from the performance rather than the edit. Phillips calls it the best synthetic performance he's gotten out of a video model — the same claim Tim Simmons made for Seedance 2.5 a day earlier, from a very different kind of scene. Coming from a BAFTA-nominated director with two decades of television behind him, the interesting part isn't that the shots are longer; it's that duration finally stopped dictating the cutting rhythm.
Higgsfield AI (dir. Aitore Zholdaskali)
A 95-minute action fantasy: four street thieves pull a heist, an ancient artifact drops one of them through a portal into the underworld, and the rest team with a covert military unit to go get her — a chase that runs through a Tibetan temple and feudal Japan. Directed by Aitore Zholdaskali, co-written with Adilkhan Yerzhanov. The production is the interesting document. Fifteen people, fourteen days, $500,000 — of which roughly 80% went to compute, not crew. It was built on Higgsfield's own Soul Cinema and Soul Cast products plus Seedance 2.0, generating and refining fifteen-second clips with prompts that averaged around three thousand words each, which is the part worth sitting with: holding a character's face steady across 95 minutes currently costs you an essay per shot. On August 4 Higgsfield open-sourced the whole thing — every prompt, character model, asset, and reference, free. That is a genuinely unusual act of disclosure in a field where workflow secrecy is the norm, and it is more valuable to other filmmakers than the film itself. Two claims around it don't survive a check. The banner is "World's First Ever AI Feature Film," which isn't true and wasn't true in 2024; the defensible version, which Higgsfield's own press used elsewhere, is the first feature made entirely on the Higgsfield platform. And the Cannes framing — the CEO wrote that it "premiered in Cannes" — describes a private industry preview on the Vieux Port on May 16 and a screening at the Cinéma Olympia, a working commercial cinema, on May 21. The Festival de Cannes publicly disputed coverage calling it a festival debut, confirming it was never in the Official Selection and had been shown at a third-party industry event during festival week. In Cannes, not at Cannes.
Higgsfield AI
A 110-minute action feature built with Seedance 2.5 on Higgsfield for $2,000,000, with a real cast performing alongside the generated filmmaking — and, six days after the studio's last open-sourced feature, released free with every prompt and asset public, so any shot in it can be rebuilt from scratch. The casting is the strategic move. N3on, Israel Adesanya, Matt Kiatipis, and Quinton "Rampage" Jackson — a streamer and two former UFC champions, not working actors. Higgsfield didn't buy acting pedigree; it bought audiences that arrive pre-assembled and are native to the platforms this will be watched on. Pair that with a hybrid pipeline where real performers front AI-generated everything-else, and the pitch stops being "look what the model can do" and starts being a distribution strategy. PJ Accetturo, who broke the workflow down publicly, called it the first AI feature with real celebrity actors — a narrower and far more defensible claim than the ones usually attached to films like this. What makes it worth studying is the disclosure, which went further than the film. Higgsfield published the full production, and Accetturo's threads note that the team distilled roughly $1M of R&D into a downloadable PDF of their process. Two AI features fully open-sourced inside a week, from the same studio, at four times the budget of the first — the interesting question isn't whether the films are good, it's that the workflow is now the product and the movie is the advertisement.
Horace Dodd
Izzy has a secret she doesn't want her boyfriend Marlon to know. He knows she goes out at night; what he doesn't know is where she goes, or what she does when she gets there. Tonight one of those secrets gets a little too close to home. Four minutes, 4K, and the most notable thing about it is the framing: not a short, not a test, but EP. 1 of an original animated series. That's a different kind of commitment. A one-off can coast on a striking look, while a pilot has to establish characters you'd follow into episode two — a premise with somewhere to go, and a world worth returning to. Dodd sets up a couple, a secret, and a nocturnal routine, then lets the title do the threatening. The craft is what sells it. Serialized generative animation has been the stated goal of a lot of creators for two years now and the hard part was never the first episode — it's holding a look and a character steady across a run of them. Episode one is a strong opening argument.
"An exploration in style, non-linear storytelling, and a mind plagued with end stage capitalism" — a public-access late-night kids' show for adults, running a full unbroken hour. It's BLVCKL!GHT's second hour-long piece after Rapture in early 2025, and another dispatch from the Route 47 Television universe he's been building across Gorbo's Swim Hole, Cryptid Dating Game, and Sad Steve's since 2024. The episode plays out of order, and his account of why is the funniest thing in a very funny description: they lost the tapes, put them back in the wrong sleeves, watched it that way, and decided it was better. The rest of the writeup is sustained deadpan — a $5,000,000 budget that "went into the fog," a fog machine they could not turn off, a fog now credited as a cast member, and Denise at the station who asked him to please stop. Underneath the bit is a real claim about performance: he says he has never gotten emotion this precisely dialed in, working with three lights, one of which is a lamp from his mother's house.
"First test with Seedance 2.5 at 1080p and Oh My!" Dave Clark posted just under two minutes of early-access footage, with the note that it was worth the late-night session. The resolution is the news. Seedance 2.5 arrived in June with the headline that mattered most — 30 seconds of continuous generation in a single pass — but it shipped capped below HD, and as recently as four days earlier LTX's public benchmark table listed 2.5 as topping out at 720p. That ceiling is what kept an otherwise remarkable model in the demo-and-social tier rather than a finishing pipeline. A native 1080p pass is the difference between footage you post and footage you can cut into something.