GPT-6 Sol & Claude Opus 5.5 are liveGPT-6 Sol at half the price of 5.6 Sol

HappyHorse 1.1 API

AI Video

HappyHorse 1.1 is Alibaba's newest HappyHorse video family on OmniaKey. One card covers all three ways to use it, and each way has its own model id: happyhorse-1.1-t2v for text to video, happyhorse-1.1-i2v to animate a first frame, and happyhorse-1.1-r2v to drive the shot from up to nine reference images.

Alibaba

HappyHorse 1.1

Mode
Prompt0 / 12000
Duration
Resolution
Aspect ratio

Estimated $0.21$0.042/s × 5s

Your result will appear here once you run the model.

01

Where it fits

HappyHorse 1.1 is Alibaba's newest HappyHorse video family on OmniaKey. One card covers all three ways to use it, and each way has its own model id: happyhorse-1.1-t2v for text to video, happyhorse-1.1-i2v to animate a first frame, and happyhorse-1.1-r2v to drive the shot from up to nine reference images.

Pick HappyHorse 1.1 for the sharpest motion, lighting and subject consistency in the family, at $0.04 to $0.11 per second depending on resolution; step down to HappyHorse 1.0 only when prompts are already tuned to its look.

02

What it does

Text to video

happyhorse-1.1-t2v

Generate a short video clip from a text prompt.

Image to video

happyhorse-1.1-i2v

Animate a still image into a short video clip.

Reference to video

happyhorse-1.1-r2v

Guide a clip with reference media to keep a subject, style or motion.

03

In depth

HappyHorse 1.1 is Alibaba's current HappyHorse video family on OmniaKey, and it is listed as one card with three model ids: happyhorse-1.1-t2v for a prompt, happyhorse-1.1-i2v for a first frame, and happyhorse-1.1-r2v for reference-driven shots built from up to nine images.

The family covers 480P, 720P and 1080P, takes nine aspect ratios, and stops at 1080P. Alibaba's watermark setting is on by default and stamps a small two-word mark in the lower-right corner, so add watermark false to the parameters when the frame has to stay clean.

We ran three 720p prompts against Seedance 2.0 on 26 October 2026 and lost all three, then the same three against Kling 3.0 and won two: the noodle stall and the perfume orbit. Both sets, plus a watermark check and a 9:16 size check, appear below.

Fast turnarounds on three shots

HappyHorse finished every clip in our October test ahead of its competitor, including a 91.7-second image-to-video orbit. When a client needs three options before the end of the day, that head start is the reason to pick this family rather than a sharper one.

Signage and clean typography

In the noodle-stall test the painted sign read correctly and no invented lettering appeared in the background, which is the failure that spoils street scenes. Pair it with a prompt that explicitly forbids extra signs if the frame has to pass QC.

Character sheets with many references

The r2v id accepts up to nine reference images, so a costume, a face and a prop can be described in separate files instead of one crowded description. Send them as reference_image entries and keep the prompt about motion, light and camera.

04

Hands-on notes

  • In our three-prompt 720p test on 26 October 2026, HappyHorse 1.1 lost all three clips to Seedance 2.0: the skate jump, the noodle stall and the product shot.

    Read the test
  • Against Kling 3.0 on the same three 720p prompts, HappyHorse 1.1 won the noodle stall and the perfume orbit while Kling took the skate trick.

    Read the test
  • HappyHorse 1.1 stamped a small two-word mark in the lower-right corner by default, and setting watermark to false removed it on a separate 3-second 480P clip.

    Read the test
  • Our 14 September 2026 smoke run rendered a 5-second 1080P HappyHorse 1.1 text-to-video clip as a 1920x1080 MP4 of 5.16 seconds.

  • When we asked for 9:16 at 480P on 25 October 2026, HappyHorse 1.1 returned 480x848, while the same request on Wan 3.0 gave 480x854.

05

Limits to plan around

  • Three ids, not one: the operation decides which model name goes in the request, so a script that switches from text to image to reference has to swap the id as well as the media.
  • The watermark is on unless you turn it off, and it sits inside the picture rather than in metadata. Add watermark false to every job whose frames will be graded or keyed.
  • Stills drive the shot, but only as a first frame: there is no last frame and no video reference, and the family stops at 1080P, so a 4K master is out of reach.
  • The model can add what the prompt did not ask for: on the perfume orbit it introduced drifting smoke and deepened the colour of the liquid, which is harmless in a mood piece and wrong for a product shot.

06

Compare nearby models

ModelBest forOperationsListed price
HappyHorse 1.1This page
happyhorse-1.1
Three operation-specific ids, nine aspect ratios, 480P to 1080P, quick returns.Text to video · Image to video · Reference to video
$0.042–$0.108
per second
Seedance 2.0
dreamina-seedance-2-0-260128
Sharper on action and product fidelity in our tests, with a slower queue.Text to video · Image to video · Reference to video
$0.077–$0.407
per second
Kling 3.0
kling-3.0
Fixed 5, 10 or 15 second clips and a priced mode, with no reference operation.Text to video · Image to video
$0.80–$3.00
per call

08

Call it

Media models run asynchronously. Create a task, poll it until it finishes, then download the signed asset URL.

Create, poll, download
# 1. Create the job. 202 Accepted; the body carries the task id,
#    the pre-charged price, and a poll_url. Reuse one Idempotency-Key
#    across retries so a timeout never starts a second paid job.
curl --fail-with-body https://api.omniakey.com/v1/media/tasks \
  -H "Authorization: Bearer $OMNIAKEY_API_KEY" \
  -H "Content-Type: application/json" \
  -H "Idempotency-Key: happyhorse-1.1-001" \
  -d '{
    "model": "happyhorse-1.1-t2v",
    "type": "video",
    "operation": "text_to_video",
    "input": {"prompt": "a paper boat on a quiet lake at sunset"},
    "parameters": {"duration": 6, "resolution": "720p"}
  }'

# 2. The operation picks the model id on this route:
#      happyhorse-1.1-t2v     prompt only
#      happyhorse-1.1-i2v     media[] type "first_frame"
#      happyhorse-1.1-r2v     media[] type "reference_image"

# 3. Poll GET /v1/media/tasks/{id} until status is
#    succeeded | failed | canceled. Progress carries percent and stage.
curl --fail-with-body https://api.omniakey.com/v1/media/tasks/TASK_ID \
  -H "Authorization: Bearer $OMNIAKEY_API_KEY"

# 4. Download assets[0].url before expires_at, or poll again for a
#    fresh signed URL. Failed and canceled jobs are refunded.
curl -sS -o out.mp4 "$(jq -r '.assets[0].url' result.json)"

09

Pricing

Video jobs are billed per second of rendered output, not per token. Failed or canceled jobs are refunded automatically.

Per second of output

$0.042–$0.108per secondSave 40%

480P
$0.042per second
720P
$0.084per second
1080P
$0.108per second

More models from Alibaba

Related articles

Frequently asked questions

How do I call HappyHorse 1.1 on OmniaKey?

POST to /v1/media/tasks with the model id, type "video", and operation "text_to_video" — or "image_to_video" with a first frame in input.media[] with type "first_frame". Reuse one Idempotency-Key across retries so a timeout never starts a second paid job.

What does HappyHorse 1.1 cost?

HappyHorse 1.1 bills $0.042–$0.108 per second of rendered video. Creation pre-charges the requested duration up front from your prepaid balance, and failed or canceled jobs are refunded automatically.

Why does the API return a task id instead of the video?

Video renders take from about a minute to several minutes, so OmniaKey runs every video route through one asynchronous lifecycle. Poll GET /v1/media/tasks/{id} until the status is succeeded, then read the signed asset URL.

Which HappyHorse 1.1 id do I call?

One per operation: happyhorse-1.1-t2v for text to video, happyhorse-1.1-i2v when you send a first frame, and happyhorse-1.1-r2v when the shot is driven by reference images. All three share the resolution menu and the per-second billing.

How do I remove the HappyHorse watermark?

Put watermark false in the parameters. Left out, the setting is on and a small two-word mark is drawn in the lower-right corner; we confirmed the switch on a separate 3-second 480P clip, which came back with no mark. The mark is part of the picture, so it cannot be cropped away safely.

How many reference images does it take?

Up to nine on the r2v id, sent as reference_image entries in input.media. The i2v id is different: one first frame only, no last frame and no video reference. Pick the id by what you have, not by what you want to end up with.

Is HappyHorse 1.1 better than Seedance 2.0?

On our three 720p prompts it lost all three, mostly on action accuracy and on adding elements nobody asked for, though it returned every clip sooner. It beat Kling 3.0 on two of the same three prompts, so the answer depends on which model you are leaving.

Does it generate sound?

Our clips all carried a track, including the skate board clack and the dripping water in the perfume shot, and Alibaba's reference does not describe a sound switch for this family. Write the sound into the prompt and check loudness afterwards, since the levels varied a lot between clips.