What does a HappyHorse 1.1 text-to-video request need?

HappyHorse 1.1 text-to-video requires one field: prompt, up to 2,500 characters. aspect_ratio, duration_seconds and resolution are optional and fall back to 16:9, five seconds and 720p. The schema is closed, so a request carrying any field the endpoint does not define is rejected before it reaches the model.

How long can a HappyHorse 1.1 text-to-video prompt be?

HappyHorse 1.1 text-to-video accepts up to 2,500 characters in prompt, with a minimum of one. That budget covers subject, wardrobe, action, camera movement, light and pacing written as a single paragraph, which is the shape this endpoint reads best. No second field exists for style notes or exclusions.

How do I write a prompt for HappyHorse 1.1 text-to-video?

Write the HappyHorse 1.1 text-to-video prompt as a shot, in the order things happen: who or what is in frame, what they do, how the camera behaves, what the light is doing, how the clip is paced. Naming the camera move matters most, since a brief that only describes a place tends to come back as a near-static frame.

Does HappyHorse 1.1 text-to-video have a seed parameter?

HappyHorse 1.1 text-to-video does not expose seed. Save the prompt, aspect_ratio, duration_seconds and resolution with each export. The image-to-video and reference-to-video tasks support an optional seed for controlled comparisons.

Can one HappyHorse 1.1 brief cover several aspect ratios?

Yes. Hold prompt, duration_seconds and resolution fixed on HappyHorse 1.1 text-to-video and submit once per value of aspect_ratio. Nine values are available, so a feed crop, a print crop and a cinema crop all trace back to one description. Each submission runs as its own task and returns its own take.

When should I use HappyHorse 1.1 text-to-video instead of the image or reference endpoints?

Use HappyHorse 1.1 text-to-video when nothing has been shot yet. image-to-video needs an approved opening frame and carries subject, wardrobe and shape from it. reference-to-video needs one to nine pictures and holds specific elements steady across the shot. Text alone is the quickest route to a moving version of an idea.

Is the HappyHorse 1.1 text-to-video API synchronous?

No. HappyHorse 1.1 text-to-video runs as an asynchronous task: the call returns a task id, and the finished video arrives through polling or a callback once the render completes. Build the client around that pattern, and keep the task id stored with the request so each result can be matched back to the settings that produced it.

Is there a negative prompt on HappyHorse 1.1 text-to-video?

No. HappyHorse 1.1 text-to-video has no negative-prompt field, and the schema rejects fields it does not define, so anything you want kept out of frame has to be handled inside prompt. Describing what should be in the shot works better here than listing what should stay out of it.

What file does HappyHorse 1.1 text-to-video return?

HappyHorse 1.1 text-to-video returns an MP4 video, delivered as a link on the finished task. Alibaba's documentation describes HappyHorse output as MP4 with H.264 encoding, so the file drops into an editing timeline without a conversion pass. Nothing else comes back with it, such as stills or project files.

What did HappyHorse 1.1 change compared with 1.0?

Alibaba describes HappyHorse 1.1 as optimizing motion modeling and temporal consistency for smoother movement in complex action sequences, and as reading several reference images more consistently. Every endpoint on this page runs the 1.1 version, which this catalog dates to 23 June 2026.

HappyHorse 1.1 API

alibaba/happyhorse-1.1/text-to-video

HappyHorse 1.1 text-to-video is Alibaba's video model turning a written brief into a 3 to 15 second clip with generated sound. Choose 720p or 1080p and one of nine aspect ratios from 21:9 to 9:21; the default is 720p at 16:9.

Filtering differs by endpoint: Image to video — Uncensored, Reference to video — Uncensored, Text to video — Filtered.

From $0.14/s

Run model View API

Input

Text to video

Prompt601/2500

Aspect ratio

Resolutionaffects price

Durationaffects price

Safety checkerOff

SpicyAPI does not run a safety checker on this model. Your prompt and media go to the model as written, and the model itself decides what comes back.

Checking your account before enabling submission and pricing.

Output

Ready for a new run

Example result

Extend

Video upscaling

Video editing

{
  "model": "alibaba/happyhorse-1.1/text-to-video",
  "input": {
    "aspect_ratio": "16:9",
    "resolution": "720p",
    "duration_seconds": 5,
    "prompt": "Handheld 35mm, warm amber street light cutting through venetian blinds into a dim roadside motel room at night. An adult woman sits on the edge of the bed in a black lace slip and tall cowboy boots, a denim jacket dropped on the floor. She hooks one boot heel against the other toe, works it off and lets it thud onto the carpet, then peels the second one off, flexes her bare feet on the rug and falls slowly back across the quilt with her arms spread, laughing and looking sideways into the lens. Bars of light crawling over her, dust in the beam, shallow focus, grain, crickets and a distant truck."
  }
}

Introduction

HappyHorse 1.1 Text to video is the prompt-only endpoint of Alibaba's HappyHorse 1.1 family: a prompt of up to 2,500 characters returns a clip of 3 to 15 whole seconds at 720p or 1080p, with a generated soundtrack. Nine aspect ratios are available, from 21:9 to 9:21, including the 4:5 and 5:4 feed and print crops. This endpoint screens the prompt, and a brief that crosses its content line is rejected outright rather than softened.

Choose a HappyHorse 1.1 mode by starting material

What you have at the start	Mode
a written scene or script	HappyHorse 1.1 Text to video
an approved opening image	HappyHorse 1.1 Image to video
one to nine reference images	HappyHorse 1.1 Reference to video

Key features & highlights

Nine aspect ratios in one field

aspect_ratio accepts 16:9 (the default), 9:16, 1:1, 4:3, 3:4, 4:5, 5:4, 9:21 and 21:9. A set of deliverables in several shapes becomes a loop over one parameter rather than a round of re-framing in post. The rain-hood-convenience example was written for 9:16 at 1080p, the vertical end of that list, and spends its words on one continuous action away from camera.

A soundtrack with no switch

Every clip comes back with generated audio, and there is no field to turn it off. Describe the sound you want: the desert-tailgate brief asks for cicadas and a distant highway, and in rain-hood-convenience the rain hammering the awning arrives in the same pass as the picture. When the clip goes under your own mix, mute or replace the track in the edit.

Draft at 720p, deliver at 1080p

Duration runs from 3 to 15 whole seconds with 5 as the default, and resolution defaults to 720p. Both drive the price. Draft at 720p, then re-run the take you want at 1080p; each pass draws its own take, so keep the wording fixed rather than expecting the same frames back at a higher resolution.

Request parameters

Request parameters for HappyHorse 1.1 · Text to Video
Parameter	Required	Description
`prompt`	Yes	Write it as a shot: subject, action, camera, light, pacing. This endpoint screens the prompt: a brief that crosses its content line is rejected outright rather than softened. at least 1 characters · at most 2500 characters
`aspect_ratio`	No	Nine values, default 16:9. 4:5 and 5:4 are the feed and print crops. `16:9` · `9:16` · `1:1` · `4:3` · `3:4` · `4:5` · `5:4` · `9:21` · `21:9` · Default: `16:9`
`resolution`	No	720p or 1080p. Draft at 720p, deliver at 1080p. `720p` · `1080p` · Default: `720p`
`duration_seconds`	No	Whole seconds, 3 to 15. Drives the price. `≥ 3` · `≤ 15` · Default: `5`

Pricing

Pricing
Specification	Price
`720p`	$0.14 / second
`1080p`	$0.18 / second

Use cases

Vertical clips with a phone-camera look

The karaoke-room example asks for a device look by name, with rolling shutter, phone-camera noise, slight shake and mixed neon light, in 9:16 at 720p. Two people share one vertical frame, trade a chorus on a single microphone and collapse into each other laughing, and the room sound comes back with the picture.

Mood pieces carried by light

Moving light can hold a whole clip together. In motel-blinds-boots, amber street light through Venetian blinds crawls over the subject while she works off two boots and falls back across a quilt, a chain of actions held inside the default five seconds at 720p. Nothing lights the room except what the prompt describes.

Settle framing and pacing before building frames

When a shot needs an exact subject, or the brief goes past what this endpoint accepts, the path runs through an opening frame on image-to-video or a reference set on reference-to-video. Text-to-video is where framing, light and pacing can be settled in a few quick drafts before those inputs are prepared.

Explore key use-case tutorials

Prompt guide & pro tips

Write it as a shot

Give the subject, action, camera, light and pacing in the order the viewer sees them. The desert-tailgate prompt does this in one paragraph: who is on the tailgate and what she wears, what she does (boots swinging, tipping a bottle, squinting into the low sun), then "warm orange backlight, lens flare, handheld 35mm", then the sound.

Keep the brief within what this endpoint accepts

The prompt is screened before generation, and a description past its content line comes back as a rejection, not as a milder clip. The published examples work through wardrobe, light and gesture rather than spelling everything out. For work beyond that line, start from an image on image-to-video or reference-to-video. With a 2,500-character limit, cut adjectives before you cut actions.

Set the frame before comparing takes

The defaults are 16:9 and five seconds. If the deliverable is 4:5 or 9:21, set it before the first draft, because a composition written for one frame does not translate to another. Keep the move resolvable within the duration you choose, and remember that a written shot leaves wardrobe, body and framing to the model, so re-runs drift.

API quick start

# Generate the idempotency key once; reuse it when retrying after a lost response.
export SPICY_IDEMPOTENCY_KEY="${SPICY_IDEMPOTENCY_KEY:-$(uuidgen)}"
curl -X POST https://api.spicyapi.ai/api/v1/jobs/createTask \
  -H "Authorization: Bearer $SPICY_API_KEY" \
  -H "Content-Type: application/json" \
  -H "Idempotency-Key: $SPICY_IDEMPOTENCY_KEY" \
  -d '{
  "model": "alibaba/happyhorse-1.1/text-to-video",
  "input": {
    "prompt": "A slow aerial shot drifting over a misty pine forest at sunrise",
    "resolution": "720p",
    "aspect_ratio": "16:9",
    "duration_seconds": 5
  }
}'

# Copy data.taskId from the response into TASK_ID, then poll the status.
TASK_ID="TASK_ID_FROM_RESPONSE"
curl "https://api.spicyapi.ai/api/v1/jobs/recordInfo?taskId=$TASK_ID" \
  -H "Authorization: Bearer $SPICY_API_KEY"

See the API reference for authentication, the complete schema, polling, and error handling.

Related models and next tasks

Missing the model you need?

Tell us what to add and where you would use it. Requests go to the people who decide what gets onboarded next.

One written brief, nine deliverable ratios

HappyHorse 1.1 text-to-video takes a written prompt and returns one 3 to 15 second clip at 720p or 1080p, with nothing to upload. prompt is the only required field; aspect_ratio, duration_seconds and resolution all carry defaults. aspect_ratio alone covers nine frames, and every clip comes back with generated sound that the request cannot switch off.

Open the playground

One written brief, nine deliverable ratios

Input: prompt only, required, up to 2,500 characters
Duration: 3-15 whole seconds, default 5
Resolution: 720p or 1080p, default 720p
Aspect ratios: 9 values: 16:9, 9:16, 1:1, 4:3, 3:4, 4:5, 5:4, 9:21, 21:9
Audio: Sound generated alongside the video, no switch to disable it
Execution: Asynchronous task, billed per second of output

Ratio loop: Nine ratios from one fixed briefHold prompt, duration_seconds and resolution steady and submit the same brief once per aspect_ratio value. A feed crop, a print crop and a 21:9 wide all come back from one description, so a placement list turns into a loop over a single field.
No upload: The one endpoint here that takes text aloneimage-to-video needs an opening frame and reference-to-video needs at least one reference image. text-to-video has neither field, which makes it the endpoint to reach for while a scene still exists only as a sentence.
Closed request: Four fields and no hidden switchesThe schema accepts prompt, aspect_ratio, duration_seconds and resolution, and rejects anything else. There is no seed, no negative prompt, no closing-frame field and no audio toggle, so everything you control sits in those four values.

What the HappyHorse 1.1 text-to-video API does well

Nine ratios behind one field

aspect_ratio carries 16:9, 9:16, 1:1, 4:3, 3:4, 4:5, 5:4, 9:21 and 21:9, and falls back to 16:9 when the field is left out. The four uncommon values are the ones worth knowing: 4:5 and 5:4 are the feed and print crops that usually mean reframing after the fact, and 21:9 and 9:21 are the extremes at either end.

Two fields carry the price

duration_seconds takes whole seconds from 3 to 15 and resolution takes 720p or 1080p. Both are marked as price-affecting in the schema, so the estimate shown before you submit moves as you change either one, and the wording of the prompt leaves it alone.

A 2,500-character prompt
The prompt field holds up to 2,500 characters, which is room for subject, wardrobe, action, camera behavior, light and pacing in the order they happen. A single character is technically enough to pass validation, but a one-line brief leaves the model to invent everything the sentence left out.
No source material needed
Neither an image field nor a reference array appears on this endpoint, so a shot can exist before the photography does. That suits pitch work and pre-production, where the point is to put a moving version of an idea in front of people while the cast, set and location are still being argued about.
Every submission draws a fresh take
There is no seed field here, so two runs of one brief come back as two different takes of it. Store the exact prompt, duration_seconds, resolution and aspect_ratio next to each export, because that record is the only reliable way back to the version you approved.
Sound written into the same brief
No audio field exists on this endpoint, yet every clip returns with sound generated alongside the picture. Name the sounds you want in the same paragraph as the action, such as rain on a window or traffic far below, and swap the track out in post when the clip has to sit under its own music bed.

Who HappyHorse 1.1 text-to-video is for

Campaign deliverables
Loop the ratio, hold everything else
Write one brief, then submit it once per placement changing only aspect_ratio. Feed, story, print and cinema crops all come from the same wording, duration_seconds and resolution, so the review conversation stays about the shot itself and never turns into nine separate briefs.
Storyboards and pitches
Motion before anything has been booked
A written scene becomes a moving reference while the photographer, cast and location are still proposals. Fifteen seconds carries a full beat of a story, and 720p is enough to read on a screen in a pitch meeting, which keeps the exploration cheap.
Social content teams
Nine placements, one submission pattern
A content calendar that lists vertical, square and widescreen versions of the same idea maps onto one prompt and nine values of aspect_ratio. The brief gets written once and approved once, and each placement runs as its own task with its own estimate.
Teams with no source frame
Start where the other two endpoints cannot
image-to-video and reference-to-video both need a picture before they accept a request. When there is nothing to animate yet, text-to-video is the only endpoint in this family that can run, and its output often becomes the reference material for the work that follows.

How to call the HappyHorse 1.1 text-to-video API

Write the shot as one paragraph
Cover subject, action, camera, light and pacing in the order they happen, inside the 2,500-character prompt field. A brief written as a time sequence gives the model an order to follow; an unordered pile of details gives it nothing to sequence.
Set aspect_ratio for the placement
Nine values are available and 16:9 applies when the field is left out. Once a brief is producing more than one crop, name the value on every submission so the placement is recorded in the request and not just in someone's head.
Set duration_seconds between 3 and 15
Whole seconds only. Three seconds holds one gesture, fifteen holds a short sequence, and the field moves the price in step with the length you ask for, so pick the length the edit actually needs.
Pick resolution before you submit
720p is the default and reads fine for review; 1080p is the delivery pass. Every run draws its own take, so treat the 1080p submission as a new take of the brief and judge it on its own terms.
Poll the task or take the callback
The endpoint is asynchronous: the call returns a task id and the finished MP4 arrives when the render completes. Build the client around polling or a callback, and store the request beside the result.

HappyHorse 1.1 text-to-video examples

A product teaser in a single ratio pass

One paragraph names the product, the surface, the lighting and the camera behavior, which is enough for a clean rotate-and-reveal with no reference photography anywhere in the pipeline.

A minimalist skincare bottle in matte black glass rotates slowly on a marble podium under a single overhead spotlight, background shifting from teal to violet, faint steam rising from a cracked lid, studio product lighting, shallow depth of field, 35mm.

A street scene with camera movement stated

Naming what the camera does, and not only where the scene is set, keeps the shot from resolving into a static establishing frame with a little ambient motion in it.

A neon-lit night market street, string lights swaying, steam rising from a food stall, a cyclist crossing the middle distance, handheld tracking shot, warm sodium light against cool blue shadows, light rain reflections on the pavement.

A brand hero shot pushing from wide to close

Stating the push from a wide establishing frame to a close detail gives the model a camera arc to execute, which is the part a scene description on its own leaves open.

A vintage convertible parked on a coastal cliff road at golden hour, camera pushing in slowly from a wide establishing shot to a close detail of the chrome grille, ocean waves breaking far below, warm backlight and lens flare, cinematic color grade.

HappyHorse 1.1 text-to-video questions

Write the whole shot in one paragraph, in the order it happens: who is in frame and what they wear, what they do, how the camera moves, what the light does, how it is paced. Then set duration_seconds to the length that beat actually needs, since three seconds and fifteen seconds want different amounts of action described. Once a brief holds up at 720p in one ratio, loop aspect_ratio over the other eight values to build the placement set.

Workbench