Create video with Happy Horse 1.1
Describe a shot or upload one reference image. The workspace reads the callable Happy Horse 1.1 modes from Star Task and requests a server quote before creating a task.
Happy Horse 1.1This model does not support reference video input
Results
Generated content will appear here
What Happy Horse 1.1 is
Happy Horse 1.1 (Chinese: Kuaile Xiao Ma) is a video generation model from Alibaba, offered through Alibaba Cloud Model Studio and the Happy family product page. It is the successor to Happy Horse 1.0 and is published as three model IDs, one per generation mode.
Alibaba's documentation lists the three routes: text-to-video for prompt-led clips, image-to-video that animates a first frame, and reference-to-video, which accepts 1 to 9 reference images and keeps their subjects consistent across the clip. According to Alibaba's help centre, clips run 3 to 15 seconds at 24 frames per second, come out as MP4, and the models generate audio as well as picture.
Alibaba describes 1.1 as a targeted upgrade over 1.0 rather than a new architecture: smoother motion in complex movement, stronger subject consistency in the multi-reference route, better handling of long prompts that describe several consecutive scenes, cleaner visual texture, and native audio. Those are the publisher's own descriptions of its own product, and this page repeats them as exactly that.
The three model IDs Alibaba publishes
Each generation mode is a separate model ID in Alibaba's catalogue. The rows below follow Alibaba's help centre and the product page, where every sample clip is bound to its model ID.
One boundary is worth stating plainly: the video-edit route exists only on Happy Horse 1.0. Alibaba's model list marks 1.1 as text-to-video, image-to-video and reference-to-video, so editing an existing video stays on the older model.
| Mode | Model ID | What it does |
|---|---|---|
| Text to video | happyhorse-1.1-t2v | Generates a clip purely from a prompt. On Alibaba's API this accepts storyboards that split one prompt into numbered shots. |
| Image to video | happyhorse-1.1-i2v | Animates a first-frame image while keeping its composition, subject and lighting. |
| Reference to video | happyhorse-1.1-r2v | Accepts 1 to 9 reference images; on Alibaba's API the prompt refers to them in order as character1, character2 and so on. This is the route Alibaba presents as 1.1's main commercial upgrade. |
What the official documentation specifies
These values come from Alibaba's help centre model list and the published API reference for the 1.1 routes.
The API documentation also shows the storyboard pattern: a single prompt can contain numbered shots such as Shot 1 [0~3s] wide angle, Shot 2 [3~6s] medium shot, and the model splits the timeline across them. Alibaba's material describes one prompt carrying 6 to 8 consecutive scenes this way.
| Specification | Value in the documentation |
|---|---|
| Clip length | 3 to 15 seconds, any integer; the API default is 5 |
| Resolutions | 480p, 720p and 1080p on the API; Alibaba bills per second with 480p at a reduced rate |
| Aspect ratios | 16:9, 9:16, 1:1, 4:3, 3:4, 4:5, 5:4, 9:21 and 21:9 on the API |
| Frame rate | 24 fps |
| Container | MP4 |
| Audio | The help centre lists audio among the models' capabilities |
| Prompt length | Up to 2,500 Chinese characters or 5,000 non-Chinese characters; excess is truncated |
| Determinism | A seed parameter (0 to 2147483647) is documented; the same seed and prompt can still differ because generation is probabilistic |
What this workspace exposes
The picker is the authority on what you can submit here, and it is narrower than Alibaba's catalogue. The rows below are read from this site's own model configuration and the public product catalog.
The most important gap sits between the catalogue and the picker: reference-to-video is the route Alibaba presents as 1.1's headline upgrade, and this workspace does not offer it. The site's catalog declares the route, but the existing reference control accepts a video file rather than a set of images, so the incompatible route stays hidden until a dedicated reference-image flow exists.
| Control | Options in this workspace |
|---|---|
| Modes | Text to video and image to video |
| Duration | 3 to 15 seconds, defaulting to 5 |
| Resolution | 720p or 1080p, defaulting to 720p |
| Aspect ratio | 16:9, 9:16, 1:1, 4:3 or 3:4, defaulting to 16:9 for text to video |
| Reference image | One image in image-to-video |
| Audio | No audio switch is rendered; output sound follows the route's service-side default |
| Cost | Billed per output second; the workbench estimate starts at 19 credits per second at 720p, and the quote printed before you submit is the binding figure |
What this workspace does not expose
These capabilities exist in Alibaba's documentation but have no control here. They are listed so you do not look for them.
- Reference-to-video with up to 9 images: not offered here; the picker accepts one image in image-to-video only.
- Video editing: a Happy Horse 1.0 route; Alibaba does not publish it for 1.1.
- 480p output: the API's economy tier; this workspace submits 720p or 1080p.
- Four extra aspect ratios: 4:5, 5:4, 9:21 and 21:9 are documented on the API but absent from this picker.
- Seed control: documented on the API, not rendered here.
- An audio switch: no toggle exists in this workspace's request shape.
How to generate here
Four steps, in the order the workspace asks for them.
- Pick the mode: text to video for a prompt-led clip, or image to video when a still should drive the shot.
- Write the prompt. Short scene descriptions work; numbered shots (Shot 1 [0~3s], Shot 2 [3~6s]) are the documented pattern for multi-scene clips.
- Set duration between 3 and 15 seconds, resolution and aspect ratio, then attach one reference image if you are animating a still.
- Read the quote, sign in when the workspace asks, then submit. If a task fails, read its status before retrying so you do not submit a duplicate.
What the upgrade claims are based on
The five upgrade dimensions come from Alibaba's own product material and developer-community articles. None of them is an independent measurement, and this page ran no comparison of its own.
- Motion and temporal consistency: Alibaba says 1.1 reworks motion modelling so dance, fight, driving and fluid scenes keep their continuity and physical weight.
- Subject consistency in reference-to-video: up to 9 reference images, with faces, logos and product details held stable across the clip and across multiple characters on screen.
- Long-prompt understanding: scene planning that splits one long prompt into 6 to 8 consecutive shots and assigns each a duration.
- Visual texture: skin, fabric and product surfaces rendered with less of the polished, synthetic look attributed to earlier models.
- Native audio: music and ambience matched to the scene natively rather than added in post. The third-party API comparison tables differ on this point, so treat it as a vendor claim.
Why other sites quote different numbers
Pages ranking for this model quote prices and capabilities in incompatible units. None of them is the number that applies to a task submitted here.
The prices below were read from the pages themselves in October 2026. They are third-party figures for third-party services, they change, and they are not quotes for tasks submitted on this page.
The only figure that binds a task here is the quote the workspace prints immediately before submission, which is resolved server-side from the live catalog.
- EvoLink, an API reseller, lists per-second prices of 0.056 US dollars at 480p, 0.112 at 720p and 0.145 at 1080p, with the same three 1.1 model IDs.
- Try Happy Horse AI, an unofficial generator site, publishes sample clips imported around the 1.1 model path and sells generation through its own subscription; its terms reserve output rights to the site.
- Alibaba Cloud's own platforms bill through Model Studio quotas and API plans rather than a public per-clip list price on the product page.
What this page does not promise
A model page should say where its own claims stop. These are the boundaries for this one.
- No free allowance is promised. Whatever credits exist are shown in your account, not here.
- Four of the five clips in the gallery are official Alibaba demonstrations published on the Happy family product page. None was generated on this site.
- The fifth clip is a third-party example published by Try Happy Horse AI with its prompt; the prompt is reproduced word for word from that site.
- The five upgrade dimensions are Alibaba's descriptions of its own product. This page ran no evaluation.
- Media licensing is stated on each card. The official clips carry no redistributable licence, so they are shown as third-party visual references without implying Alibaba's endorsement.
- Latency, queue behaviour and final billing depend on the service at the moment you submit, not on anything stated here.
Where the samples and numbers come from
Every figure and every clip on this page traces back to one of the sources below. Media licensing is stated with the gallery cards.
| Source | Used for | Reference |
|---|---|---|
| Alibaba Cloud help centre, video generation model list | The three 1.1 model IDs, audio, resolutions, 3-15 seconds, 24 fps, MP4, and the 1.0-only video-edit boundary | help.aliyun.com ↗ |
| Alibaba Cloud Happy family product page | The four official sample clips, each bound to its model ID and marked as officially AI-generated | www.aliyun.com ↗ |
| Alibaba Cloud developer community articles | The upgrade description, capability overview and model comparison graphics | developer.aliyun.com ↗ |
| EvoLink API reference for Happy Horse 1.1 | Request parameters, prompt limits, the storyboard pattern, aspect ratios and reseller pricing | evolink.ai ↗ |
| Try Happy Horse AI | The third-party sample clip and its published prompt | tryhappyhorseai.com ↗ |
| Star Task public product catalog for this site | Every control, route and price in the workspace sections | task.starhk.uk, site-18 |
Examples
Official text-to-video demonstration from Alibaba's Happy family product page, bound on the page to the happyhorse-1.1-t2v model ID and marked by Alibaba as officially AI-generated.
A second official text-to-video demonstration from the same product page, also bound to happyhorse-1.1-t2v and marked as officially AI-generated.
Official image-to-video demonstration, bound on the product page to the happyhorse-1.1-i2v model ID: a still first frame animated into a moving shot.
Official reference-to-video demonstration, bound on the product page to the happyhorse-1.1-r2v model ID. This is the multi-reference route Alibaba presents as 1.1's main upgrade, and it is not offered in this site's workspace.
A professional ballet dancer performs a powerful grand jeté across a dimly lit stage, arms extended, tutu flowing in slow motion. The camera tracks from a low side angle, warm spotlights create long shadows on the floor, cinematic stage lighting, subtle fabric movement, 10 seconds, 16:9.
A third-party text-to-video example published by Try Happy Horse AI around the 1.1 model path, shown with the prompt that site publishes for it, reproduced word for word.
Alibaba's own capability overview for the 1.1 upgrade, published in its developer community. A vendor description of its own product, not an independent measurement.
Alibaba's model comparison table for the Happy family, published with the same developer community article. It is the source for the mode boundaries repeated on this page.
Questions
What is Happy Horse 1.1?
It is Alibaba's video generation model, the successor to Happy Horse 1.0, offered through Alibaba Cloud Model Studio and the Happy family product page. It generates 3 to 15 second clips at 24 fps from a prompt, a first-frame image, or 1 to 9 reference images.
What is the difference from Happy Horse 1.0?
Alibaba describes 1.1 as stronger in five areas: motion smoothness, subject consistency across up to 9 reference images, long-prompt scene planning, visual texture, and native audio. The video-edit route remains on 1.0 only; 1.1 covers text-to-video, image-to-video and reference-to-video.
How long can a clip be?
3 to 15 seconds, and the API accepts any integer in that range with a default of 5. Billing is per output second, so a longer clip costs proportionally more.
Does it generate audio?
Alibaba's help centre lists audio among the 1.1 models' capabilities, and its product material describes native music and ambience matched to the scene. This workspace renders no audio switch, so output sound follows the route's service-side default.
Can I use multiple reference images here?
Not in this workspace. The model supports 1 to 9 reference images through its reference-to-video route on Alibaba's API, but this site's picker exposes text-to-video and single-image image-to-video only, because the current request shape has no multi-image reference control.
How much does a run cost here?
The workspace requests a server-side quote before submission, and that quote is the billing authority. Billing is per output second; the workbench estimate starts at 19 credits per second at 720p, which is 95 credits for the 5-second default.
Are the clips on this page generated on this site?
No. Four are official Alibaba demonstrations from the Happy family product page, each bound to one of the 1.1 model IDs. The fifth is a third-party example published by Try Happy Horse AI with its prompt. All were re-encoded at 1280 px for web delivery without any change to content.
Why do other sites quote different prices?
They are different businesses pricing their own services in different units: EvoLink lists per-second API prices, Try Happy Horse AI sells its own subscription, and Alibaba Cloud bills through Model Studio quotas. None of those figures applies to a task submitted here; the quote before you submit does.
Explore next
AI video generator
Generate video with four Seedance tiers: Seedance 2.5 and the 2.0, 2.0 Fast and 2.0 Mini tiers. Compare duration, resolution and inputs, then create.
Image and video tools
Every image and video job on this site in one place: generate an image, generate a video, read a prompt out of a photo, or open one model directly.