Four ways to create in one model
With no media, HappyHorse starts from your text; with one image, it animates it; with 2 to 9 images, it combines them; with a video, it edits it. On Deepnia the mode simply follows what you add.
Video model · Alibaba
HappyHorse — Make a new video or rework your own, with one model.
HappyHorse is Alibaba’s video model, unveiled in April 2026, and it covers four ways of working: from text, from one image, from several reference images, or from a video you want to change. On Deepnia it makes 3–15 second clips in 720p or 1080p, with sound, and reworks footage you upload, for example to restyle it or swap an object.
- 3–15 s
- 720p or 1080p
- Up to 9 images
- Video editing
- Sound included
- 9:16 · 1:1 · 16:9 · 4:3 · 3:4
The video tool opens with HappyHorse already selected. The exact cost shows before every creation; you sign up when you generate.
- Your source video
- Up to 5 images (optional)
- Your editing instruction
- The edited video
At a glance
- Maker
- Alibaba
- Modes
- Text, image, multi-reference, video editing
- Length
- 3 to 15 s (5 s by default)
- Resolution
- 720p or 1080p (720p by default)
- Aspect ratios
- 9:16, 1:1, 16:9, 4:3, 3:4
- References
- Up to 9 images
- Editing
- Source video of 3–60 s
- Sound
- Made by the model, or your video’s own when editing
What HappyHorse does best
Up to 9 reference images
Bring a person, a product and a setting from different images into one video, for an ad or a product page. Alibaba aims HappyHorse at advertising, ecommerce, short-form video and social media marketing.
Edit a video you already have
Add a video and up to 5 reference images, then describe the change. Alibaba lists style transfer, replacing an element and inserting a subject: swap a shop sign or a product, for example, without a reshoot.
Sound, created or kept
With Auto, the default, HappyHorse creates the video’s sound. When you edit your own video, pick Original under “Source audio” to keep its voices, music and ambience.
Several shots in one video
Alibaba says HappyHorse can cut between several shots while keeping characters stable across the cuts, which suits short dramas. Describe them in your prompt, numbered (Shot 1, Shot 2…).
Vertical to widescreen, in 720p or 1080p
Choose 9:16, 1:1, 16:9, 4:3 or 3:4 for text or several images; from a single image or a video, HappyHorse keeps the source’s aspect ratio. Test in 720p, then switch to 1080p for the final cut.
How to use HappyHorse on Deepnia
Open the video tool
The “Try HappyHorse” button opens Deepnia’s video tool with the model already selected.
Add your media, or none
Add 1 image to animate or 2 to 9 images to combine, or add a video to edit along with up to 5 images. With no media, the video starts from your text.
Describe the scene or the change
The prompt is required, up to 5,000 characters: scene, action and camera for a new video, the change you want for an edit.
Set it up and generate
Choose the length (3–15 s), resolution and aspect ratio; when editing, the length follows your video and the “Source audio” setting offers Auto or Original. Check the cost shown, then create.
HappyHorse on Deepnia: what you can set
- Inputs
- Text only, 1 image, 2–9 reference images, or 1 video to edit with up to 5 images
- Mode
- Set by your media: text, image, multi-reference or video editing
- Length
- 3–15 s (5 s by default); when editing, the source video’s length
- Source video (editing)
- 3–60 s, up to 100 MB; Alibaba says the edited video lasts 15 s at most
- Resolution
- 720p or 1080p (720p by default)
- Aspect ratios
- 9:16, 1:1, 16:9, 4:3, 3:4 for text and multi-reference; from an image or a video, the source’s aspect ratio
- Images
- Up to 10 MB each
- Sound
- Auto by default: the model creates the sound; when editing, “Source audio” also offers Original, which keeps your video’s sound
- Prompt
- Required, up to 5,000 characters
Prompt ideas to try
Copy a prompt, open the tool and adapt it to your project.
Teaser for a fitness app launch
- Text only
- 9:16
- 10 s
- 1080p
A coach finishes a set of squats in a bright gym, wipes her forehead and smiles at the camera; close-up on her phone showing a timer, then a wide shot of the gym at sunrise. Smooth camera moves, vivid colors, energetic pace. No on-screen text.
Animate a dish photo for a restaurant
- 1 image: photo of the dish
- 6 s
- 720p
Animate this photo: steam rises gently from the dish, the sauce glistens under warm light, a hand sets a small bowl of fresh herbs beside the plate, the camera slowly pushes in.
Lookbook for a Nairobi fashion label
- 3 images: model, dress, bag
- 9:16
- 10 s
- 1080p
The woman from the first image wears the dress from the second image and carries the bag from the third image. She walks across a Nairobi rooftop at sunset: medium shot, then a close-up on the bag, smooth camera move, high-end editorial mood.
Restyle a shop video shot on a phone
- Video to edit: 12 s
- 1 image: the new shop sign
- Source audio: Original
- 720p
Turn the whole video into colorful 3D animation, keeping every movement of the people and the camera exactly the same, and replace the shop sign with the one from the reference image.
Swap the product in an ad
- Video to edit: 8 s
- 1 image: the new bottle
- Source audio: Auto
- 1080p
Replace the bottle the person is holding with the bottle from the reference image, at the same size and with the same grip; keep the lighting, the reflections and everything else in the video unchanged.
Three-shot wedding film
- Text only
- 16:9
- 12 s
- 1080p
Shot 1: wide shot of a reception hall decorated with white flowers, guests raising their glasses. Shot 2: close-up of the couple’s hands exchanging rings. Shot 3: the couple dances in the middle of the room as the camera slowly circles them, soft warm light.
Tips for better results
- With several images, refer to each one by its position (“the woman from the first image”, “the bag from the third image”) and say what it brings: a person, a product or a setting.
- When editing, say what changes and what must stay the same: movements, framing, faces, lighting.
- Keep the source video to 15 seconds or less: according to Alibaba, it is the first 15 seconds that get edited.
- To animate a photo, describe the motion and the camera rather than what is already visible: the video keeps the image’s framing and aspect ratio.
- To chain several shots, number them in the prompt (Shot 1, Shot 2…) and give each one its framing.
- Test the idea in 720p, then run it again in 1080p for the final cut.
Good to know
- To tell a longer story, create several 15-second clips and join them in your editor.
- Pick 9:16 for Reels, TikTok and Shorts, and 16:9 for YouTube and the web; from an image or a video, HappyHorse keeps the source’s aspect ratio.
- To keep a character or product identical from one video to the next, reuse the same reference images and refer to them the same way in the prompt.
HappyHorse or another model?
Deepnia’s other video models, to pick the right one for your project.
| Model | Length | Resolution | Aspect ratios | Best for |
|---|---|---|---|---|
| HappyHorseThis model | 3–15 s | 720p–1080p | 9:16 · 1:1 · 16:9 · 4:3 · 3:4 | Make a new video or rework your own, with one model. |
| Kling 3.0 | 3–15 s | Set by the mode | 9:16 · 1:1 · 16:9 | A shot-by-shot editor, persistent characters and optional sound. |
| Seedance 2.5 | 4–30 s | 480p–720p | 16:9 · 9:16 · 1:1 · 3:4 · 4:3 · 21:9 | Videos up to 30 s with up to 16 references, sound always on. |
| Seedance 2.0 | 4–15 s | 480p–1080p | 16:9 · 9:16 · 1:1 · 3:4 · 4:3 · 21:9 | Videos up to 1080p, guided by image and video references. |
| Motion Control | Set by the source video | 720p–1080p | From the source | Copying the moves from a real video onto a character image. |
| Gemini Omni | 4, 6, 8 or 10 s | 720p–4K | 9:16 · 16:9 | Clips up to 4K with 7 reference images. |
Frequently asked questions
Who made HappyHorse?
Alibaba. Its official announcement names the Alibaba Token Hub (ATH) business unit as the developer, and HappyHorse was unveiled as a limited beta in April 2026.
How long is a HappyHorse video?
From 3 to 15 seconds (5 by default) from text or images. When editing, the length follows the source video, accepted from 3 to 60 seconds; Alibaba says the edited video lasts 15 seconds at most.
Can HappyHorse edit an existing video?
Yes. Add a video (3–60 s, up to 100 MB) and up to 5 reference images, then describe the change: a new style, a replaced object or sign. The “Source audio” setting keeps the original sound (Original) or lets the model create the sound (Auto).
Does HappyHorse generate sound?
Yes. With the Auto setting, used by default, HappyHorse creates the video’s sound. When you edit your own video, “Source audio” also offers Original, which keeps your video’s sound.
How many images can I use?
One image to animate it, 2 to 9 images to combine people, products and settings, or up to 5 images with a video to edit. Each image can be up to 10 MB.
Which resolutions and aspect ratios are available?
720p or 1080p (720p by default). With text or several images: 9:16, 1:1, 16:9, 4:3 or 3:4. From a single image or a video, the result keeps the source’s aspect ratio.
Can I write the prompt in any language?
Yes: Alibaba says the prompt can be written in any language. On Deepnia the prompt is required and can be up to 5,000 characters long.
HappyHorse or Kling 3.0?
HappyHorse to edit an existing video or combine up to 9 images, in 720p or 1080p. Kling 3.0 to split a scene into up to 5 shots and keep 3 characters or products identical from shot to shot.
Ready to try HappyHorse?
The video tool opens with HappyHorse already selected. The exact cost shows before every creation; you sign up when you generate.