
43%of prompts lock the camera
Jaime Yao, director of In Our Time
Start from a character sheet, then write every shot around one face, one emotional change and one sound choice: close-ups for the moment a feeling changes, pauses in the dialogue, a locked camera, no score.
This guide is built from:

Most AI drama short films break at the face. A character looks different from one shot to the next, the model adds music nobody asked for, and the moment that should land plays as a blank stare. We read the 1,527 video prompts behind 17 films made in Flick, by the Flick team and the Summer 2026 Flick Residency, including The Herder, shown at the Cinequest Film Festival. This guide is what their makers do, in seven steps, each with a real shot and the exact prompt that made it.
Step 1
Karan Archana made an expression sheet for Evelyn, one of the two leads in Drop of Water.
@Vivi Lovell Character Sheet.
Professional character expression reference sheet for Evelyn (same woman from the turnaround sheet reference images — mid-30s South Asian woman, olive skin, sharp jawline, high cheekbones, black hair in a tight ponytail, brown penetrating eyes, wearing fitted earth-tone trousers, tucked shirt with sleeves rolled up, sleeveless light jacket). Create ONE wide image showing SIX close-up portrait panels of her face, each with a different expression: neutral calm, determined focused, fearful scared, sad grieving, intense fierce, and tender emotional. Same face and features in every panel, 100% identical likeness to the reference, plain pure white seamless background, even flat lighting, chest-up framing for each panel. No scene, no props, no other people.
Step 2
The turn of One More Day, the moment a character's feelings change, is one face, one emotion and one line.
00:05In the setting of @Hospital Room, Vivi Unconscious, Vivi @Vivi Lovell Character Sheet stands at the side of the hospital room. Extreme close up on Vivi, the hospital wall behind her, out of focus. Hope floods Vivi's face. She nods and says in voice of @Maya's Warning Label : "Yes. Please."
Step 3
This Drop of Water shot spends more words on silence than on the line.
Start from the last frame. Ken has just said, "We can't leave her out there." He slowly turns his head toward Evelyn and waits. Evelyn doesn't answer immediately. She holds his gaze in complete silence for two long beats, her face unreadable, exhausted, resolute. No camera movement. No smoke, dust effects, or visual changes. The room remains perfectly still. Finally, without anger or hesitation, Evelyn quietly says, "She made her choice... when she drank the toxic water." She continues looking at Ken after speaking. Hold the silence for another beat. Natural cinematic performance, restrained emotion, realistic facial micro-expressions, no dramatic gestures, no additional movement.
Step 4
The family car scene in One More Day is a single shot.


VIVI (in the back with black hair in two messy buns) looks at her father who is driving: "Easy for you to say. You've ordered the exact same thing for... approximately four hundred years." FATHER: "When something's perfect, you don't change it." Everyone laughs. For just a moment, everything is perfect.
Step 5
Hannah L. animated a giant with one instruction: her eyes.


camera locked. her eyes look to the right, then to the left, then to the right again. keep the analog film grain texture.
Step 6
Tell the model to drop the music, or it may add some.
00:05Audio: No music. Foley only — room ambient, a rising tension hum barely present under the scene. Characters: Elise — @Elise Exact face, hairstyle, outfit, body proportions. Visual Style: Photoreal cinematic. Widescreen. Close coverage — tight singles, slow push-ins as the revelation deepens. Minimal camera movement. The stillness amplifies the weight. Scene Duration: 6 seconds Shot Breakdown: Shot 1 Scene starts from @Elise Shot Medium — Elise. Something shifts behind her eyes. She holds up a hand, stopping him. ELISE: "Wait — wait, wait, wait." Her voice remains controlled but the pace breaks. ELISE: "Are you saying all of this wasn't initiated by a human — but by the Anger itself?" Foley Sound Design: - Elise's "Wait — wait, wait, wait": the first crack in her composure — small, controlled, but present
Step 7
Grain and a handheld camera make a shot feel remembered, not produced.


heavy film grain. handheld camera. quiet room. she looks up and around the room

Jaime Yao, director of In Our Time
Owen Doerr, director of Lifeguard

Marius Jensen, director of Still Time

Raj Rishi More, director of Seasons
We matched keywords in the video prompts saved on each film's public canvas on 2026-10-17. A prompt can count toward more than one move.
M.E. OwensSummer 2026 Residency
After a car crash leaves her in a coma, Vivi meets Death, who offers her one more day.
183 generated in Flick
Seedance 2.0Seedance 2.0 MiniGPT Image 2Nano Banana 2Eleven v3
Karan ArchanaSummer 2026 Residency
In a drought-cracked land, Ken and Evelyn argue over whether to save Maria, who drank the toxic water, with what little clean water they have left.
150 generated in Flick
Kling 3.0 4KKling O3 ProGPT Image 2Nano Banana ProEleven v3
Marius JensenSummer 2026 Residency
A small-town family story about Caleb, Jules and their Ma that turns into a robbery and a standoff with the deputy on Main Street.
539 generated in Flick
Seedance 2.5Seedance 2.0Nano Banana 2
Owen DoerrSummer 2026 Residency
Brothers Joe and Benji sit out a giant sandstorm in a sealed metal chamber, while a memory of the lifeguard from Benji's childhood pulls at him.
183 generated in Flick
Seedance 2.0MiniMax H3GPT Image 2Nano Banana Pro
Dr. DreamsSummer 2026 Residency
Three young drifters, Casey, Patches and Gumby, ride freight trains through the Midwest with their dog Thor, a railroad officer on their trail.
95 generated in Flick
Seedance 2.0Seedance 2.5Nano Banana ProGPT Image 2Eleven v3
Raj Rishi MoreSummer 2026 Residency
An elderly couple, Tenzin and Pema, in a stone Himalayan house, told through shared meals and weather that turns from summer storm to blizzard.
166 generated in Flick
Kling O3 ProKling O3 Pro MultishotGPT Image 2Nano Banana Pro
Jaime YaoFlick team
A lone traveler crosses a storm-beaten world shot like tilt-shift miniature photography.
617 generated in Flick
Kling 3.0 ProSeedance ProNano Banana ProMidjourneyEleven v3
Jaime YaoFlick team
Inspired by a writer's years after the war, drawn as two-tone woodcut animation: a lamplit desk, rain over the city, a giant bird, a white stag.
131 generated in Flick
Seedance ProWan 2.6Nano BananaNano Banana Pro
Zoey Z.Flick team
A true story.
45 generated in Flick
Seedance 2.5MidjourneyNano Banana ProEleven v3
Simay Ezer PekcanSummer 2026 Residency
In a small town, people bring strange objects to the Tailor to turn into garments that prove their worth, while she slowly fades into the background.
365 generated in Flick
Seedance 2.0Nano Banana ProNano Banana 2Eleven v3
Sam WassermanSummer 2026 Residency
In a dying world, the last Runners enter the final running of the Gauntlet, a trial that no one in its history has survived.
118 generated in Flick, plus mostly uploaded footage
Seedance 2.5Seedance 2.0MidjourneyGPT Image 2
Marco ZenkiSummer 2026 Residency
An analyst, Elise, traces a city riot back to a gas released from a warehouse vat, and starts to suspect the Anger itself is behind it.
45 generated in Flick
Seedance 2.0MiniMax H3
Walter MauroSummer 2026 Residency
Duro, an eight-year-old girl, in hand-painted 2D animation: her home, her mother's kitchen and an immersion in the river of Oxum.
94 generated in Flick
Seedance 2.5MiniMax H3GPT Image 2
Adrien BodsonSummer 2026 Residency
A discarded toy robot down to 1% battery crawls across a bedroom floor, past the house cat, to reach the charger.
23 generated in Flick, plus mostly uploaded footage
Seedance 2.5Eleven Music
Hannah L.Flick team · Experiment
A 1940s prizefighter faces his opponent in a gilded opera hall, shot as grainy 35mm with a slow-motion stare before the punch.
62 generated in Flick
Seedance 2.5GPT Image 2Nano Banana Pro
Hannah L.Flick team · Experiment
A young woman grows into a giant in a black and white park and lies down for a nap while the crowd below scatters.
53 generated in Flick
Seedance 2.0Kling O3GPT Image 2Midjourney
Hannah L.Flick team · Experiment
A woman alone in a quiet sunlit room, shot handheld on heavy grain.
23 generated in Flick
Kling O3Kling 3.0Nano Banana Pro
| Film | Length | Generations | Per minute |
|---|---|---|---|
| The Herder | 4:27 | 617 | 139 |
| Still Time | 18:04 | 539 | 30 |
| Happy Souls Alive | 4:58 | 365 | 73 |
| One More Day | 5:38 | 183 | 32 |
| Lifeguard | 4:35 | 183 | 40 |
| Seasons | 2:49 | 166 | 59 |
| Drop of Water | 3:21 | 150 | 45 |
| In Our Time | 3:47 | 131 | 35 |
| The Miracle | 10:04 | 118 | mostly uploads |
| Catch Out | 4:06 | 95 | 23 |
| Duro | 9:14 | 94 | 10 |
| Boxing at the Opera | 0:17 | 62 | 219 |
| La fille dans le parc | 0:16 | 53 | 199 |
| To My Grandma | 1:38 | 45 | 28 |
| Anger is Stolen from the Market | 5:30 | 45 | 8 |
| Battery Boy | 3:11 | 23 | mostly uploads |
| a slow kind of still | 0:29 | 23 | 48 |
| Model | Kind | Generations | Films |
|---|---|---|---|
| Seedance 2.0 | Video | 575 | 10 |
| Seedance 2.5 | Video | 392 | 10 |
| Seedance Pro | Video | 186 | 2 |
| Kling 3.0 Pro | Video | 132 | 3 |
| Kling 3.0 4K | Video | 89 | 1 |
| Kling O3 Pro | Video | 28 | 4 |
| MiniMax H3 | Video | 27 | 6 |
| Seedance 2.0 Mini | Video | 25 | 3 |
| Nano Banana Pro | Image | 575 | 12 |
| GPT Image 2 | Image | 302 | 10 |
| Nano Banana 2 | Image | 260 | 9 |
| Midjourney | Image | 49 | 6 |
| Eleven v3 | Voice | 44 | 6 |
A short, character-driven film whose shots, and often its voices and sound, are generated with AI image and video models. The craft is the same as any drama: a character, a turn in their feelings, and the sound around it. Made in Flick, they run from 16 seconds to 18 minutes.
On Flick TV, free and without an account, along with AI short films in every genre. Each film also has its own page on Flick, where its workflow can be opened.
Yes. In Flick, Seedance 2.0 lists native audio with lip sync, and Seedance 2.5 and Kling 3.0 4K generate native audio. Write the line, the speaker and the delivery into the prompt, and the model generates the voice with the picture.
Anywhere from a few dozen to several hundred. Across 15 AI short films made in Flick, the count ran from 23 to 617 images, clips and voice lines, with a median of 131, not counting uploaded footage.
Make a character sheet first, a turnaround or an expression sheet of one face, then reference it in every shot prompt. In Flick, attach the sheet and call it by name with an @ reference, so every shot draws the same face.
No single model. Across 17 AI drama films made in Flick, Seedance 2.0 and 2.5 made 967 of the 1,547 video clips, and Kling was picked for dialogue and multi-shot handheld scenes. Most filmmakers mix models on one canvas: an image model for characters and a video model for shots.
You can start for free. Flick gives you 100 credits when you sign up, and 200 more for joining the Flick Discord, and the free plan includes MiniMax H3 Fast, MiniMax H3 Max and Pruna Draft video and every image model. Seedance and Kling, the video models drama filmmakers use most, come with the Standard plan.
Yes. On a film's page in a desktop browser, choose Workflow to preview the canvas, then Get Workflow to clone it into your own project. Cloning asks you to sign up.
From under half a minute to over 18 minutes. On Flick, short experiments run 16 to 29 seconds, and Residency films run 2:49 to 18:04.
Characters, close-ups, dialogue and sound on one canvas, with the models these filmmakers used.
Try for free, no credit card required.
Ask AI about this page:ChatGPTPerplexityClaude