Free guide ·
16 AI animation prompts run in Google Flow, each with the clip it made, plus the six-part formula behind them and how to adapt it for Veo, Kling and Runway.
By John Shemron, graphic designer and video editor at PromptVault
A good AI animation prompt reads like a short shot list. It sets the look once, then says what happens second by second, where the camera goes and what you hear. The 16 prompts below all follow the same six-part formula: a style line, three timed beats, one camera sentence, a sound sentence, framing and text rules, then the format. Each one was run in Google Flow, and each comes with the unedited 10-second clip it made, including the two places where the prompt didn't fully hold.
If you're new to animating from a prompt, start with the prompt animation guide, which compares the three routes: text to video, image to video and prompt to code. This post goes deep on the prompts themselves.
The prompts and their reference frames are free on the AI Video With Code page, so you can download the exact frame each clip started from.
| Part | What it does | From The Ferryman |
|---|---|---|
| 1. Look | Sets the style before anything moves, so every later detail follows it | "Painterly 2D-3D hybrid animation like a prestige animated series: hand-painted brush textures…" |
| 2. Three timed beats | 0-3 s, 3-7 s and 7-10 s, with one change in each | "3-7 s: a hooded passenger holds out a coin that glows silver in the fog; the ferryman slowly looks up." |
| 3. Camera | One sentence, with at most two moves | "Side tracking shot, then a wide reveal." |
| 4. Sound | Its own sentence, with sounds that match the beats | "Sound: pole dipping in water, a creaking boat, a distant choir." |
| 5. Rules | Full frame, no captions, no watermark | "No subtitles, no on-screen text, no watermark, no logos." |
| 6. Format | Aspect ratio and length, at the very end | "16:9, 10 seconds." |
Why each part is there:
Copy any prompt as it is. The heading says which style group it belongs to, and the note under each prompt points out the one choice that makes it work.
Painterly 2D-3D hybrid animation like a prestige animated series: hand-painted brush textures, mature realistic proportions, muted shadows with warm highlights, soft volumetric haze, cinematic lighting and camera. 0-3 s: night on a wide river in thick blue fog; a ferryman in a long hooded coat poles a flat wooden ferry, one lantern swinging. 3-7 s: a hooded passenger holds out a coin that glows silver in the fog; the ferryman slowly looks up. 7-10 s: the fog parts to reveal a vast ghostly city of lights on the far shore. Side tracking shot, then a wide reveal. Sound: pole dipping in water, a creaking boat, a distant choir. Full-frame 16:9 that fills the whole frame edge to edge: no black bars at the top or bottom, no letterbox, no cinema crop. No subtitles, no on-screen text, no watermark, no logos. 16:9, 10 seconds.
What to notice: each beat has one job: the setup, the coin, the reveal. The city of lights doesn't appear until the fog parts in the last three seconds.
Painterly 2D-3D hybrid animation like a prestige animated series: hand-painted brush textures, mature realistic proportions, muted shadows with warm highlights, soft volumetric haze, cinematic lighting and camera. 0-3 s: dawn on an endless white salt flat cracked into hexagon tiles; a lone monk in a white hooded robe with a tall staff stands small in the frame. 3-7 s: the ground splits and a colossal crab-shaped machine of rusted copper plates rises behind him, salt pouring off its legs. 7-10 s: the monk calmly plants his staff; the machine freezes, then lowers its huge head to him. Low-angle wide shot, slow push-in. Sound: wind hiss, grinding metal, one deep resonant gong. Full-frame 16:9 with no letterbox bars. No subtitles, no on-screen text, no watermark, no logos. 16:9, 10 seconds.
What to notice: the threat and the answer are separate beats, so the machine rising and the monk's calm reply don't blend. The low angle makes the machine tower over him.
Painterly 2D-3D hybrid animation like a prestige animated series: hand-painted brush textures, mature realistic proportions, muted shadows with warm highlights, soft volumetric haze, cinematic lighting and camera. 0-3 s: in a candlelit stone cathedral, a tall woman with ash-blonde braids and huge dark feathered wings folded behind her stands with eyes closed. 3-7 s: her eyes snap open, the wings unfurl to their full span and every candle flame bends in the gust. 7-10 s: she launches straight up through a shaft of moonlight in the roof, black feathers drifting down toward the camera. Slow push-in, then tilt up. Sound: candle flicker, a deep wingbeat whoosh, rising choir. Full-frame 16:9 with no letterbox bars. No subtitles, no on-screen text, no watermark, no logos. 16:9, 10 seconds.
What to notice: stillness, then a snap, then flight. "Every candle flame bends in the gust" gives the room a visible reaction to the wings.
Painterly 2D-3D hybrid animation like a prestige animated series: hand-painted brush textures, mature realistic proportions, muted shadows with warm highlights, soft volumetric haze, cinematic lighting and camera. 0-3 s: night over snow-covered mountains; a blazing comet streaks across the sky and a small bronze shape breaks away, glowing hot. 3-7 s: it crashes into a snowy pine forest; a child in a fur hood walks up to the crater holding a lantern. 7-10 s: a bronze automaton lies in the snow; its round eyes flicker on and it slowly reaches out a hand to the child. Wide, then close-up. Sound: a rushing roar, a muffled impact, ticking gears, a soft music-box theme. Full-frame 16:9 with no letterbox bars. No subtitles, no on-screen text, no watermark, no logos. 16:9, 10 seconds.
What to notice: three places in ten seconds, held together by "Wide, then close-up." The close-up is saved for the moment that matters, when the automaton's eyes flicker on.
Painterly 2D-3D hybrid animation like a prestige animated series: hand-painted brush textures, mature realistic proportions, muted shadows with warm highlights, soft volumetric haze, cinematic lighting and camera. 0-3 s: a girl in a small wooden canoe glides through a misty mangrove lagoon at sunrise toward a green-bronze bell half sunk under a stone arch. 3-7 s: she strikes the bell with her oar; a ring of golden light ripples out across the water. 7-10 s: hundreds of glowing lanterns rise from beneath the surface all around her canoe. Slow push-in, then crane up. Sound: water lapping, one deep bell toll that echoes, soft swelling strings. Full-frame 16:9 with no letterbox bars. No subtitles, no on-screen text, no watermark, no logos. 16:9, 10 seconds.
What to notice: approach, strike, response. Each beat answers the one before it, so cause and effect stay in order. It's also one of the two clips that didn't follow every line, explained further down.
Painterly 2D-3D hybrid animation like a prestige animated series: hand-painted brush textures, mature realistic proportions, muted shadows with warm highlights, soft volumetric haze, cinematic lighting and camera. 0-3 s: night, heavy snow; a keeper in a long oilskin coat climbs the spiral stairs of a dark stone lighthouse, a small lantern in hand. 3-7 s: at the top he strikes a match and the great lens ignites with a roar of warm light. 7-10 s: the beam sweeps across the black sea and catches a sailing ship that turns away from the rocks just in time. Wide shot from the cliff. Sound: howling wind, a match strike, a deep foghorn, waves on rocks. Full-frame 16:9 with no letterbox bars. No subtitles, no on-screen text, no watermark, no logos. 16:9, 10 seconds.
What to notice: one fixed camera for the whole clip, chosen for the ending, where the tower, the beam and the ship have to share one frame.
Painterly 2D-3D hybrid animation like a prestige animated series: hand-painted brush textures, mature realistic proportions, muted shadows with warm highlights, soft volumetric haze, cinematic lighting and camera. 0-3 s: dusk over an ocean of orange dunes; a wooden sand-ship on runners with patched sails skims fast across the sand. 3-7 s: the dune beside it swells and cracks. 7-10 s: a colossal whale made of sand and stone breaches alongside the ship, sand cascading off its back, the crew frozen at the rail. Tracking shot alongside, then wide. Sound: wind in sails, runners hissing on sand, a deep whale call. Full-frame 16:9 with no letterbox bars. No subtitles, no on-screen text, no watermark, no logos. 16:9, 10 seconds.
What to notice: beat two is only a warning ("the dune beside it swells and cracks"). A short hint before the reveal builds tension the way a cut to a shaking glass does.
Semi-realistic animated adventure-film look, painterly detail, warm natural light. 0-3 s: a girl of about 14 with goggles on her forehead rides a giant sky-blue manta ray over a sea of white clouds at sunrise. 3-7 s: the ray dives straight down through the clouds. 7-10 s: it bursts out beside a floating island with waterfalls pouring into the sky and the girl throws her arms up, whooping. Tracking shot, then wide reveal. Sound: wind rush, a whoosh through cloud, waterfalls, a triumphant orchestral swell. Full-frame 16:9 with no letterbox bars. No subtitles, no on-screen text, no watermark, no logos. 16:9, 10 seconds.
What to notice: a shorter look line, and the dive gets a beat to itself, so the floating island can't appear before the ray bursts out of the clouds.
Painterly 2D-3D hybrid animation like a prestige animated series: hand-painted brush textures, mature realistic proportions, muted shadows with warm highlights, soft volumetric haze, cinematic lighting and camera. 0-3 s: in an open workshop above rice terraces at golden hour, a 12-year-old girl with a long braid ties the last bamboo rib of a giant red-and-gold paper kite. 3-7 s: a gust lifts the kite out of the workshop; she grabs the frame and is pulled off her feet. 7-10 s: she soars low over the glowing terraces, laughing, farmers below looking up. Tracking shot that swings out wide. Sound: wind chimes, paper snapping in the wind, joyful flute. Full-frame 16:9 with no letterbox bars. No subtitles, no on-screen text, no watermark, no logos. 16:9, 10 seconds.
What to notice: each beat takes her one step further from the ground: tying the kite, being pulled off her feet, flying over the terraces.
Painterly 2D-3D hybrid animation like a prestige animated series: hand-painted brush textures, mature realistic proportions, muted shadows with warm highlights, soft volumetric haze, cinematic lighting and camera. 0-3 s: a foggy harbour town at dusk; a young lamplighter with a long brass pole lights the first iron street lamp on the quay. 3-7 s: the next lamp lights by itself, then the next. 7-10 s: every lamp along the whole harbour ignites in a racing chain of light and the ships' bells ring as she turns back, amazed. Wide tracking shot. Sound: water lapping, a soft whoosh for each lamp, ship bells, gentle strings. Full-frame 16:9 with no letterbox bars. No subtitles, no on-screen text, no watermark, no logos. 16:9, 10 seconds.
What to notice: the light escalates in steps (one lamp by hand, the next by itself, then the whole harbour), and "a soft whoosh for each lamp" gives the sound a reason to repeat.
Painterly 2D-3D hybrid animation like a prestige animated series: hand-painted brush textures, mature realistic proportions, muted shadows with warm highlights, soft volumetric haze, cinematic lighting and camera. 0-3 s: night in a deep red-rock canyon; hundreds of paper lantern boats float down a calm river, warm light on the water. 3-7 s: a meteor shower streaks overhead and one meteor flares bright gold. 7-10 s: every lantern lifts off the water at once and rises into the sky like a second river of light. Low drone glide over the water, then tilt up. Sound: gentle water, crickets, a soft wooden flute that swells. Full-frame 16:9 with no letterbox bars. No subtitles, no on-screen text, no watermark, no logos. 16:9, 10 seconds.
What to notice: there's no main character, so the beats follow the lanterns. The gold meteor is its own beat, which gives the lanterns a reason to rise.
Bold 2D hand-drawn animated musical-adventure style: thick confident outlines, warm saturated colours, squash-and-stretch. 0-3 s: a busy night market under red paper lanterns; a short, round street chef in a big apron stirs his steaming wok. 3-7 s: a long playful dragon made of noodles rises out of the wok and loops around the stall. 7-10 s: the chef twirls his ladle like a conductor, the dragon bows to him and the crowd cheers. Comedic push-in, then wide. Sound: sizzling wok, gasps, a brass fanfare, cheering. Full-frame 16:9 with no letterbox bars. No subtitles, no on-screen text, no watermark, no logos. 16:9, 10 seconds.
What to notice: setup, surprise, punchline, like a joke. "Squash-and-stretch" in the look line asks for cartoon timing, not just a cartoon look.
Painterly 2D-3D hybrid animation like a prestige animated series: hand-painted brush textures, mature realistic proportions, muted shadows with warm highlights, soft volumetric haze, cinematic lighting and camera. 0-3 s: dawn mist over a vast reed marsh; a young queen in a cloak of grey heron feathers stands on a wooden walkway, silver dots painted on her cheekbones. 3-7 s: she slowly raises one arm. 7-10 s: hundreds of herons burst up out of the reeds around her and circle overhead as her cloak lifts in the wind. Slow push-in, then wide crane. Sound: reeds rustling, a rush of wingbeats, a single rising horn. Full-frame 16:9 with no letterbox bars. No subtitles, no on-screen text, no watermark, no logos. 16:9, 10 seconds.
What to notice: the middle beat is tiny: she raises one arm, nothing else. That pause makes it clear the herons rise because of her.
Semi-realistic animated adventure-film look, painterly detail, warm natural light. 0-3 s: a misty birch forest in autumn; a young woodcutter holding a lantern freezes as a soft pale glow moves through the mist between the trees; no other animals, only the stag appears. 3-7 s: an enormous white stag with glowing pale antlers steps out and lowers its head to the lantern. 7-10 s: the falling leaves around them turn to drifting golden light. Slow push-in, then close-up on the stag's eye. Sound: crunching leaves, a soft breath, a warm swelling choir. Full-frame 16:9 with no letterbox bars. No subtitles, no on-screen text, no watermark, no logos. 16:9, 10 seconds.
What to notice: "no other animals, only the stag appears" sits inside beat one, where the forest is set up. The limit sits right where the forest is described, the point where the model might add other animals.
Dramatic anime feature-film style: clean line art, bold silhouettes, strong rim light, high-contrast colour. 0-3 s: midnight in a clock-tower workshop; a young apprentice with ink on her cheek turns a giant brass hourglass over. 3-7 s: the sand glows blue and starts falling upward. 7-10 s: through the round clock-face window, the birds outside freeze mid-flight and the snow stops in the air; she smiles, close-up on her eyes. Sound: ticking clocks that suddenly stop, a shimmering chime, silence. Full-frame 16:9 with no letterbox bars. No subtitles, no on-screen text, no watermark, no logos. 16:9, 10 seconds.
What to notice: this is the only prompt with no camera sentence. The one camera direction, "close-up on her eyes", sits inside beat three. Add a line such as "Slow push-in." if you want the first two beats to move.
Stylized 2D noir-thriller animation: limited palette, flat shapes, hard shadows, film grain. 0-3 s: a sleeper train races across a moonlit plain. 3-7 s: in a velvet compartment, gloved hands open a briefcase full of glittering diamonds; the lights flicker and go out. 7-10 s: the lights come back on: the case is empty except for a single playing card, the queen of spades. Close-ups, hard cuts. Sound: rattling wheels, a click, darkness hum, a sharp brass sting. Full-frame 16:9 with no letterbox bars. No subtitles, no on-screen text, no watermark, no logos. 16:9, 10 seconds.
What to notice: "Close-ups, hard cuts." asks for cuts instead of one continuous move, which suits a heist.
Two clips didn't follow every line of their prompt. Both failures are useful.
[Style line: medium, era, finish, lighting]. 0-3 s: [where and when; who is there; what they're doing]. 3-7 s: [one change]. 7-10 s: [the payoff]. [Camera: one or two moves joined by "then"]. Sound: [three or four sounds, at least one tied to a beat]. Full-frame 16:9 with no letterbox bars. No subtitles, no on-screen text, no watermark, no logos. 16:9, 10 seconds.
Tips for filling it in:
The structure carries over to every tool. The details that change are clip length, how much the image does, and how to say what you don't want. Checked on 7 November 2026:
| Tool | What its own guide says | How to adapt these prompts |
|---|---|---|
| Google Veo (in Flow, the Gemini API and Vertex AI) | Google's prompt guide lists subject, action, scene, camera, lens, style, timing and audio, and says you don't need every element. It recommends separate sentences for audio, and for unwanted things it says: "Describe what you don't want to see." | Where there's a negative prompt field, move the rules there as plain words: "subtitles, on-screen text, watermark, logos, letterbox bars". |
| Runway Gen-4 | Runway's Gen-4 guide works from an input image plus a prompt, makes 5- or 10-second clips, and says to use positive phrasing only, describe the motion rather than the image, and refer to the subject simply ("the subject", "she"). | Drop the look line, since the image carries the style. Keep the beats and the camera, and rewrite the rules positively, for example "locked camera" instead of "no camera movement". |
| Kling 3.0 | Kling's API documentation says the prompt can hold positive and negative descriptions together, supports first and last frames, and has a multi-shot format that gives each shot its own length in seconds. | Keep the timed beats for one continuous shot. If you want cuts, like The Night Train, use Kling's multi-shot format instead. |
| Pika and others | Check the tool's own guide for length limits and negative-prompt support. | Start with the beats and camera, then add the rest one part at a time. |
If your tool makes shorter clips, scale the beats. For 8 seconds, use 0-2 s, 2-5 s and 5-8 s. For 5 seconds, use two beats and drop the middle one.
Video models are good at rich, one-off shots like the 16 above. They're weak at exact text, charts, and keeping a character identical across a whole series. For those jobs, prompt a coding agent instead: you describe the animation, the agent writes code, and your computer renders every frame, so nothing drifts and nothing gets garbled. Can Claude make animations? shows what that route can and can't do.
AI Video With Code teaches that route for narrated films, explainers and kinetic text: you describe the video, a coding agent (Claude Code, Cursor, Codex or Gemini) writes the scene, and your computer renders the MP4. It runs on Mac, Windows or Linux. For two-character dialogue shorts with the same cast in every episode, the Claude & Pixel Guide covers the same idea; it needs a Mac (see how to make animated dialogue videos). Both are independent courses from PromptVault, not made by Anthropic or Google.
A clear look, a few timed beats with one change each, one camera instruction and a sound line. Describe what changes over time, not only what the scene looks like. The 16 prompts above follow that pattern and come with the clips they made.
The prompts here are 96 to 143 words, enough for a look line, three beats, a camera move and sound. With an input image you can go shorter, because the image already sets the style. Runway's guide suggests starting simple and adding detail one element at a time.
The structure does. The look will differ, because every model draws differently and these were tested with Flow and a reference frame. In Runway, drop the look line and phrase the rules positively. In Kling, keep the beats, or switch to its multi-shot format if you want cuts.
Yes, the prompts are free to copy and change. What you can do with the clips you generate depends on the terms of the tool you made them in, so check those before you use a clip commercially. Text to animation for free lists the free plan terms of Flow, Runway, Pika, Kling and Firefly.
Usually one of three things: too much happens in one beat, the reference image disagrees with the words, or the prompt asks for readable text. Split busy beats, make the image match the prompt, and add text in the edit.
AI Video With Code: Films of any length, from code, for nothing per render.