Playcut v2 is coming MCP · Actor v2 · & much more
Prompt library

HappyHorse prompts for text, photos, casts and edits

HappyHorse prompts for text to video, image to video, nine-reference casts and video edits, written for HappyHorse 1.1 with spoken lines, no negative prompts.

Updated · Playcut prompt library

Give HappyHorse one plain paragraph that says who is in the shot, where they are, what moves and how the camera follows, with any spoken line in quotes after the word "saying". Leave out negative prompts, which HappyHorse rejects with an error, and name reference images as [Image 1], [Image 2] beside a description of each.

HappyHorse 1.1 is the version Playcut runs, and the 28 cards follow its four modes, from text-only scenes to edits of a clip you already have. Each tag gives the length and the credit cost; none comes with a sample clip.

HappyHorse prompts as one plain paragraph

Alibaba publishes API references for HappyHorse but no prompt guide of its own, so this method comes from those references and from how the model behaves on Playcut. The sample prompt in the HappyHorse text-to-video API reference is two plain sentences about a cardboard city at night: a subject, a setting and one slow movement.

Every card below uses the same five parts, in this order:

  1. Subject, with one detail that tells it apart: "a tailor in a grey waistcoat".
  2. Scene: the place, the light and the time of day.
  3. Motion: what moves, and how fast.
  4. Camera: one shot size and one move.
  5. Sound and line: effects, then any speech as saying: "…".

Speech is something Playcut checked on the model in July 2026, not something Alibaba documents: on Playcut, HappyHorse speaks dialogue you quote in the prompt. There is no audio upload and no voice choice.

Rules HappyHorse enforces

Rule HappyHorse 1.1 on Playcut
Negative prompts Rejected with an error on every mode. Write exclusions as positive description inside the prompt.
Prompt length Up to 5,000 characters, or 2,500 in Chinese. Anything longer is cut off with no warning.
Aspect ratios 16:9, 9:16, 1:1, 4:3, 3:4, 4:5, 5:4, 21:9, 9:21 on text and reference. Animate and Edit follow the source.
Resolution and length 720p (the default) or 1080p, 3 to 15 whole seconds. Edit output matches the source, up to 15 seconds.
References Up to 9 on Reference, up to 5 on Edit, one photo on Animate.
Sound Speech from quoted lines, generated by the model. No audio upload and no voice choice.
Credits 56 per second at 720p, 72 at 1080p. Edit bills source plus output seconds.

Alibaba's HappyHorse reference-to-video API reference sets the token rule: [Image N] must follow the order of the attached images, and each token should say what its image shows.

In credits, a 5-second 720p clip is 280, 8 seconds is 448 and 15 seconds is 840; at 1080p, 5 seconds is 360. The HappyHorse 1.1 limits and credits page has the full breakdown.

Text to video with a spoken line

Run on HappyHorse 1.1

Each card quotes its line after "saying:", with one or two speakers. On Playcut, HappyHorse voices the line itself; you cannot pick or upload a voice. Cards are 16:9 at 720p unless the tag says otherwise.

  1. Tailor at a fitting

    8 s · 720p · 448 cr
    A tailor in a grey waistcoat pins the shoulder of a navy jacket on a young man at a tall mirror, then steps back, saying: "Half an inch, no more." A narrow workshop lined with bolts of cloth, warm light. Slow push-in. Scissors clicking.
  2. Flower market at opening

    8 s · 720p · 448 cr
    At dawn in a covered flower market, a seller in a green apron lifts a bucket of peonies onto her stall and smiles at the first customer, saying: "These opened overnight." Cold blue light from a glass roof. Fixed medium shot. Water dripping, crates scraping.
  3. Three-piece band, two voices

    10 s · 720p · 560 cr
    In a small rehearsal room, the drummer counts in and the bassist raises a hand, saying: "Slower. Let it breathe." The guitarist laughs, saying: "You said that last time." Warm lamp light, cables on the floor. Slow pan from the drums to the guitarist. Amp hum.
  4. Crossing guard at the school gate

    6 s · 720p · 336 cr
    A school crossing guard in a fluorescent vest steps into the road with her sign raised and turns to the children on the kerb, saying: "Now we walk. Not run." Autumn leaves blow across the tarmac. Fixed wide shot from across the road.
  5. Bike mechanic's verdict

    6 s · 720p · 336 cr
    In a cramped bike workshop, a mechanic in a grease-streaked apron spins a wheel on the stand, listens, and turns to the customer, saying: "Spokes are fine. It is the bearing." Inner tubes hang behind him. Medium close-up, slow push-in.
  6. Climbing coach at the wall

    8 s · 720p · 448 cr
    At the foot of a bouldering wall, a coach in a chalk-dusted T-shirt checks a climber's harness knot and tugs it twice, saying: "Good. Now trust your feet." Coloured holds behind them. Medium two-shot, slow push-in. Chalk bags thudding.
  7. Hotel lift attendant

    6 s · 1080p · 432 cr
    In a brass-panelled hotel lift, an attendant in a burgundy uniform pulls the folding gate shut and turns to a guest, saying: "Top floor. The view is worth the wait." Warm light from a glass ceiling lamp. Fixed medium shot. The gate rattling.

Odd frames: 4:5, 5:4, 9:21 and 21:9

Run on HappyHorse 1.1

HappyHorse takes nine aspect ratios on text and reference cards, where Veo 3.1 takes two. These cards use the four unusual ones, 4:5, 5:4, 9:21 and 21:9, at 720p; set the ratio in the studio before you run.

Work out a 4:5 or 9:21 frame with the aspect ratio calculator

  1. Lookbook walk

    4:5 · 8 s · 448 cr
    A model in a loose sand-coloured linen suit walks toward the camera down a sunlit stone colonnade, turns once to show the back of the jacket, then walks on. Long shadows stripe the floor. Fixed full-length shot, centred. Heels on stone.
  2. Cinema poster in motion

    9:21 · 6 s · 336 cr
    A lone figure in a long coat stands on a cliff edge under a towering storm cloud lit from inside by lightning, the coat whipping as slanting rain begins. Slow crane up from the figure to the top of the cloud. Thunder rolling. Blue and amber palette.
  3. Canyon convoy

    21:9 · 10 s · 560 cr
    Three dusty off-road trucks drive in a line along a red canyon floor at golden hour, dust trailing between sandstone walls. Aerial tracking shot from the side, level with the rim. Engines rumbling, gravel crunching. Warm grade.
  4. Bookbinder at the bench

    5:4 · 8 s · 448 cr
    A bookbinder at a wooden bench folds a stack of paper, presses the fold with a bone folder and stitches the spine with waxed thread. An overhead lamp lights the tools. Fixed overhead shot. Paper creasing, thread pulling through.
  5. Boutique window at night

    4:5 · 6 s · 336 cr
    Inside a boutique window at night, knitwear on a rotating plinth turns slowly while snow falls outside and passers-by blur past. Warm spotlights on the wool, a cold blue street beyond. Fixed shot from the pavement. Muffled traffic.

Reference casts with [Image 1] to [Image 9]

Run on HappyHorse 1.1

Reference mode takes up to nine images, numbered in attach order, so the first is [Image 1]. Every token here has a description beside it, as Alibaba's reference docs ask. Faces are fine on HappyHorse, so these cards cast people.

  1. Band on a small stage

    8 s · 720p · 448 cr
    The singer in the silver jacket from [Image 1] leans into the microphone while the drummer with the shaved head from [Image 2] plays behind her on the club stage from [Image 3]. She grins, saying: "One more, then we are done." Handheld medium shot from the front row.
  2. Relay handoff in team kit

    6 s · 720p · 336 cr
    The runner with braided hair from [Image 1] passes the baton to the tall runner from [Image 2], both in the green and white team kit from [Image 3]. Slow-motion side tracking shot at the handoff, floodlights behind. Spikes striking the track.
  3. Two characters and a prop

    8 s · 720p · 448 cr
    Under a dripping shop awning at dusk, the woman in the navy peacoat from [Image 1] hands the red umbrella from [Image 3] to the man in the grey suit from [Image 2], saying: "You need this more than I do." Medium two-shot, slow push-in. Rain on canvas.
  4. Weekend bag and a vintage car

    8 s · 720p · 448 cr
    The man with a silver beard from [Image 1] swings the tan leather weekend bag from [Image 2] into the boot of the cream vintage car from [Image 3] and taps the roof, saying: "Coast road, no stops." Late afternoon sun. Fixed medium shot.
  5. Five images: a dinner toast

    10 s · 1080p · 720 cr
    At the long dinner table from [Image 5], the host in the burgundy dress from [Image 1] raises a glass, saying: "To the new kitchen." The man in round glasses from [Image 2], the woman with a grey bob from [Image 3] and the teenager in a hoodie from [Image 4] lift theirs. Slow orbit.
  6. Skater, costume and a frozen lake

    8 s · 720p · 448 cr
    The figure skater with a dark ponytail from [Image 1], wearing the sequinned blue costume from [Image 2], glides backward across the frozen lake from [Image 3] at dusk and slows into a spin. Low tracking shot across the ice. Blades hissing.

Animate one photo (image to video)

Run on HappyHorse 1.1

Animate uses one photo as the first frame and keeps its shape, because this mode takes no ratio setting. The prompt carries only motion, camera and any line. People in the photo are fine.

  1. A reader looks up

    6 s · 720p · 336 cr
    The woman in the photo looks up from her book, closes it on one finger and smiles toward the camera, saying: "You found my favourite spot." Her hair shifts in a light breeze. Slow push-in to a close-up. Café murmur.
  2. Bonsai in afternoon light

    5 s · 720p · 280 cr
    The bonsai in the photo stirs in a light draught, its tiny leaves trembling as afternoon light slides across the glazed pot. Fixed close-up. Faint room tone, a clock ticking.
  3. Road-trip photo comes to life

    8 s · 720p · 448 cr
    The couple in the photo, standing by their car at a roadside lookout, lean into each other and laugh as the wind lifts her scarf. He points down the road, saying: "Just one more hour." Slow pull back to reveal the valley below.
  4. Sign painter at a shopfront

    6 s · 720p · 336 cr
    The sign painter in the photo finishes the last gold letter on the shop window, steps back and nods, saying: "Now it looks like it has always been here." Traffic blurs behind the glass. Fixed medium shot. A brush tapping a jar.

Edit a clip you already have

Run on HappyHorse 1.1

Edit rewrites a source clip of 3 to 60 seconds and returns a clip of the same length, up to 15 seconds; a longer source is edited from its first 15. It takes up to five reference images and bills source plus output seconds: an 8-second source costs 896 credits at 720p.

  1. Restyle as cut-paper animation

    8 s source · 896 cr
    Turn the whole clip into cut-paper animation with visible paper textures, soft drop shadows and a warm autumn palette. Keep every person's movements, the camera move and the timing exactly as they are.
  2. Swap a bag from one reference

    10 s source · 1,120 cr
    Replace the bag on the man's shoulder with the tan leather satchel in the reference image, its strap and size natural as he walks. Leave his face, the background and the lighting as they are.

    Attach the satchel photo as the only reference image.

  3. Repaint a parked car

    6 s source · 672 cr
    Repaint the parked hatchback a glossy cherry red, with reflections moving naturally across the paint as the camera passes. Everything else in the clip stays exactly as filmed.
  4. Give a phone clip a film look

    12 s source · 1,344 cr
    Give the clip the look of 16mm film: soft grain, gentle halation around the highlights, slightly lifted blacks and warm skin tones. Keep the framing and every movement of the original.
  5. Move a conversation outdoors

    10 s source · 1,120 cr
    Move the two people and their conversation from the office to the seaside terrace in the reference image, matching the light direction to it. Keep their gestures, lip movements and the camera move unchanged.

    Attach one photo of the new location as the reference image. In the studio the model decides the new audio; to keep the original voices, run the edit over the API with the source audio kept.

  6. Clear a desk behind a demo

    5 s source · 560 cr
    Clear the desk behind the laptop so only a small potted plant and a lamp remain, against a plain, softly lit wall. Keep the hands, the laptop and every movement as filmed.

Where HappyHorse settings live in Playcut

With a card pasted into the video studio prompt box:

  1. Pick HappyHorse 1.1 in the video model picker. Veo 3.1 is the default.
  2. Choose the mode: Describe, Animate, Reference, or Edit 1.0, which carries "1.0" because it runs on HappyHorse's 1.0 edit model.
  3. Set the ratio on Describe and Reference cards. Resolution starts at 720p; 1080p costs 72 credits a second instead of 56.
  4. Attach references in token order: if [Image 2] is the drummer, that photo goes in second.

Playcut turns off the provider's watermark. In the studio, Edit leaves the new clip's audio to the model; over the API you can keep the source audio. Credits are held at submit and refunded if a job fails, and Edit refunds held seconds it did not use.

For a character who returns clip after clip, the persona builder helps you describe a character consistently before you cast it, and Playcut AI actors keep one on file; a tier of the actor video engine runs on HappyHorse 1.1. Attach photos only of people who have agreed to appear.

The prompt-fix guide covers what to change when a reference is ignored.

FAQ

How do I reference several images in a HappyHorse prompt?

Write [Image 1], [Image 2] and so on, with square brackets, numbered in the order you attached the images, and describe each one beside its token: "the drummer with the shaved head from [Image 2]". Reference mode takes up to nine images. If the two orders disagree, the token points at the wrong picture.

Can I use a negative prompt with HappyHorse?

No. HappyHorse has no negative-prompt field, and Playcut returns an error if one is sent, on every mode. Describe what you want instead: "a clear evening sky" rather than "no clouds", or "only a lamp on the desk" rather than a list of things to remove.

How long should a HappyHorse prompt be?

Under 5,000 characters, or 2,500 in Chinese; anything past that is cut off without a warning. Most cards here are one short paragraph, which leaves plenty of room. In a long prompt, keep the subject and the action near the start.

Can HappyHorse speak dialogue?

Yes, on Playcut: quote the line in the prompt, for example after "saying:", and the model speaks it. You cannot upload audio or choose a voice. Alibaba's API references do not document audio, so treat this as behaviour Playcut has checked rather than a vendor promise.

Can HappyHorse edit a video I already have?

Yes. Edit takes a source clip of 3 to 60 seconds and returns a clip of the same length, up to 15 seconds, with up to five reference images for swaps or style. It bills source plus output seconds, so an 8-second source costs 896 credits at 720p.