Plimo

HappyHorse AI: HappyHorse 1.1 Videos with Sound from Your Prompt and Up to 9 Photos

Pet Dance

What is HappyHorse AI?

HappyHorse is a video model from Alibaba's ATH team that first got attention by leading a public blind-vote leaderboard before anyone knew who had made it. It is good at believable, expressive motion, builds sound into every clip, and can keep the people and things from several reference photos consistent in one scene. On Plimo you get version 1.1, already selected in the generator above.

  • HappyHorse 1.1

    Turns up to 9 reference photos into one video.

    • 3โ€“15 s
    • 720P / 1080P
    • Always with sound
    • Up to 9 reference photos

    From 165 credits

Where the model came from

The series is young, with two releases so far, both in 2026.

  1. April 2026

    HappyHorse 1.0

    It appeared without a name on the Artificial Analysis video arena in early April and led several of its blind-vote rankings for text-to-video and image-to-video. A few days later Alibaba confirmed its ATH unit had built it. It generated picture and sound together, from text or a single start image.

  2. June 2026

    HappyHorse 1.1

    Added a reference mode that keeps the subjects from up to 9 photos consistent in one clip, along with smoother, more expressive motion and steadier detail from frame to frame. Clips run 3 to 15 seconds with sound built in.

How to make a HappyHorse video on Plimo

  1. Write the prompt first

    Happy Horse AI, as many people write it, always needs written directions, even when you add a photo, so describe the action, the setting and what we should hear.

  2. Choose a start frame or a reference set

    In the mode menu, Standard takes one picture for the clip to open on, or none at all; Reference takes up to 9 photos of people, pets or objects to bring together. One run uses one mode.

  3. Set length and resolution

    Clips run 3 to 15 seconds at either of two resolutions. With a start frame, the video keeps your picture's proportions.

  4. Generate and pick a favorite

    The Generate button shows the cost, and the finished clip plays in My creations. Remake reloads your photos, prompt and settings when you want a variation.

Why use HappyHorse on Plimo

  • Side by side with other video models

    It sits in the same menu as Kling, Seedance and Hailuo AI, so you can test one prompt on several and keep the clip that moves the way you pictured.

  • Nine photos in one scene

    Reference mode helps when the people or things in your video come from different pictures, like a friend, a pet and a favorite jacket.

  • Sound without extra steps

    Every clip arrives with audio, so there's no separate sound tool to open before you share it.

  • A clear price before you start

    The credit figure on the Generate button changes with length and resolution, and reference photos don't add to it. A run that fails gives the credits back.

  • Ready to post or sell

    Videos made with an active membership come without a watermark and can be used commercially, from shop listings to client work.

  • Your photos stay out of training

    Plimo doesn't use the portraits, pets or products you upload, or the clips made from them, to train any model.

Compared with other video models

Every model below has its own page. Here is the short version of what sets each one apart from this one.

  • Kling

    Kling 4.0 offers start and end frames and stretches to 30 seconds; HappyHorse 1.1 takes a start frame but no end frame and tops out at 15 seconds.

  • Veo

    Veo 3.1 Fast always makes 8 seconds and takes up to 3 reference photos, where you can set the length here and use up to 9.

  • Seedance

    Seedance pairs reference photos with reference clips and has a sound switch; its 2.0 Mini version runs on free credits.

  • Hailuo AI

    Hailuo AI's MiniMax H3 Max shares the 9-photo reference limit and always-on sound, adds end frames and reference clips, and focuses on lifelike faces.

  • Grok Imagine

    Grok Imagine 1.5 takes up to 7 reference photos but has no start frame slot, runs 6 to 15 seconds, and new accounts can make a short clip with it on free credits.

  • Wan

    Wan 3.0, a sibling from Alibaba, runs as long as 30 seconds, accepts an end frame and reference clips, and has an audio switch.

  • Vidu

    Vidu Q4 Preview takes up to 15 reference photos and can animate a start frame with an empty prompt box.

  • FLUX

    FLUX 3 Video has no reference photos but supports an end frame and clips of up to 20 seconds.

  • SkyReels

    SkyReels V4 keeps sound on too, and swaps the 9-photo reference mode for 3 photos or a start-and-end-frame pair.

Ideas to try

  • Friends who live far apart

    Add a portrait of you and one of a friend who moved abroad, and describe the two of you clinking glasses at a rooftop bar while the city hums below.

  • A pet and its favorite toy

    Use one photo of your cat and one of its battered toy mouse, and ask for a dramatic slow-motion pounce across the living room rug.

  • An old family photo that turns to smile

    Set a scanned family photo as the start frame and describe everyone turning toward the camera, with birdsong and distant traffic in the audio.

  • An outfit check

    Combine a full-length photo of yourself with photos of a jacket and a pair of boots, and describe walking toward the camera on a busy street wearing them.

  • A small product ad

    Put photos of a candle, its box and a sprig of lavender in reference mode and ask for a cozy close-up as a match is struck. Made during a membership, it can go in a shop listing.

  • A drawing that flies away

    Use a child's crayon dragon as the start frame and describe it flapping off the page while the paper rustles and a small voice cheers.

  • A holiday greeting

    Describe a snowy cabin, a dog in a red scarf and a warm voice saying happy holidays from all of us, then send it instead of a card.

  • A team photo that cheers

    Set a group shot from your club or office as the start frame and describe everyone throwing their arms up as a crowd roars and a whistle blows. It makes a fun end-of-season post.

Common questions

No. Plimo is an independent site that uses the model through an API. It is not run by Alibaba or the team behind the model and has no affiliation with them.

This model needs a membership. The 100 free credits new accounts get once can go toward Seedance 2.0 Mini or Grok Imagine 1.5 instead, two video models that run on free credits. Plans are listed on pricing.

The version card shows where prices start. Each extra second and the higher resolution add to the cost, while reference photos don't, and the Generate button always shows the exact number first.

From 3 to 15 seconds, at one of two resolutions shown on the version card. You can pick wide, vertical, square, 4:3 or 3:4; with a start frame, the clip follows your picture instead.

Yes. Audio comes with every video and there is no switch to turn it off. Put the sounds or spoken words you want into the prompt.

Yes. Some models can animate a photo from an empty prompt box, but version 1.1 asks for a description every time, start frame or not. A prompt on its own, with no photos, is enough.

Version 1.1 takes a start frame but no end frame. For a clip that ends on a picture you pick, try Kling 4.0 or Wan 3.0, or browse more options on image to video.

No. Both come from Alibaba, but they are separate model families. Wan 3.0 goes up to 30 seconds and accepts reference clips, while this one centers on a start frame or a set of up to 9 photos.

Clear, well-lit pictures where each person, pet or object is easy to see. Use one subject per photo when you can, then name them in the prompt, for example the woman in the first photo wearing the jacket from the second. Files can be JPG, PNG or WebP.

Open it in My creations and press Remake. Your photos, prompt and settings go back into the generator, so you can reword the action or change the length before you run it again. Each new run uses credits like the first one did.

Yes, as a member: videos made with an active membership have no watermark and are cleared for business use. My creations keeps them for between 30 and 360 days, by plan.

More ways to make videos

All models