Video GenerationActive
alibaba / happyhorse 1.0 reference
happyhorse 1.0 reference
Edit a short MP4/MOV clip or generate a new 720p/1080p video from up to 9 reference images and a detailed prompt, with optional watermark and seed control.
Image to VideoVideo to Video
Model ID
happyhorse 1.0 reference
Provider
alibaba
Updated
1777315461
wiro playground—alibaba/happyhorse 1.0 reference
Updated 1777315461
## Overview
HappyHorse 1.0 Reference is a reference-driven video model by Alibaba. It supports two workflows: edit an existing clip, or generate a new clip from reference images.
You provide a short video plus an instruction, or a set of reference images plus a scene description. The model outputs a new short video clip that follows your instruction while trying to keep subjects consistent across frames.
This is useful when you need repeatable characters, styling, or camera feel across iterations.
## What you can build
- Product video variations from one source clip, like “change the background to a studio.”
- Character-consistent clips from a small image set, for ads and storyboards.
- Style transfers on video, like “make it look like 35mm film.”
- Prompt-driven motion from stills, like turning concept art into a short shot.
- Quick social edits that keep the original audio.
## Inputs
- A workflow choice: either edit an input video, or generate from reference images.
- For video editing: 1 input video file in MP4 or MOV (H.264). The input can be 3–60 seconds, but output is capped at 15 seconds.
- For video editing: the input video must be within 100 MB. Its longer side must be 2160 px or less, and the shorter side must be 320 px or more.
- For video editing: the input video aspect ratio must be between 1:2.5 and 2.5:1.
- Reference images (optional for video editing, required for reference-to-video): JPEG/JPG/PNG/WEBP files.
- Reference image count: 0–5 for video editing, or 1–9 for reference-to-video.
- Reference image size limit: up to 10 MB each.
- A written instruction or scene description. It can be up to 5,000 non‑Chinese characters or 2,500 Chinese characters.
- If you upload multiple reference images, you can bind them by order in the prompt. Use labels like character1, character2, and so on.
- For reference-to-video: choose the target clip duration from 3 to 15 seconds.
- Choose an output resolution tier: 720P or 1080P.
- For reference-to-video: choose an aspect ratio, or let the system auto-detect from the first reference image.
- For video editing: choose audio handling. You can preserve the original audio, or let the model decide.
- Optional watermark toggle to add a “HappyHorse” mark in the lower-right corner.
- Optional seed value for repeatable results (0 to 2,147,483,647).
## Outputs
The model returns a generated video clip.
- A downloadable video file (typically MP4).
- Resolution matches your chosen tier (720P or 1080P).
- Duration is capped at 15 seconds.
- In video editing mode, the output can keep the original audio if you choose that option.
- If watermark is enabled, the output includes a “HappyHorse” watermark in the lower-right.
## Limitations
- Output length is capped at 15 seconds. Longer input videos are trimmed.
- Video editing requires a single input clip in MP4 or MOV (H.264).
- Input video constraints apply. Oversized frames, odd ratios, or large files may fail.
- Reference-to-video needs at least 1 reference image and allows up to 9.
- Video editing allows up to 5 reference images. More images won’t be used.
- Low-quality inputs can reduce results. Heavy compression, motion blur, shaky camera, and jump cuts can hurt edits.
- Reference images that are inconsistent in lighting or face angle can reduce identity consistency.
## Safety & compliance
- Only upload videos and images you have rights to use.
- Don’t generate misleading or deceptive content, like impersonations intended to trick viewers.
- Use watermarking when you need clear disclosure that a clip is AI-generated.
- Follow local laws and platform rules for generated media.
Example prompts
Great starting points for happyhorse 1.0 reference.
The young woman is sitting on a wooden bench in a sunlit autumn park, holding a takeaway coffee cup. The golden retriever puppy trots up to her, tail wagging, and places its paws on her lap. She laughs softly, leans down, and gently scratches behind the puppy's ears. Warm golden hour light filters through orange leaves overhead, a gentle breeze drifts a few leaves through the frame. Cinematic 35mm shallow depth of field, naturalistic colors.Video Generation
The elegant woman stands in a softly lit luxurious marble dressing room with floor-length sheer curtains gently swaying in a warm breeze. She gracefully picks up the golden perfume bottle from a marble vanity, lifts it close to her neck, and lightly sprays once. Fine mist drifts through the warm sunlight beams. She closes her eyes briefly, then opens them with a soft confident smile toward the camera. Cinematic luxury commercial aesthetic, slow elegant motion, golden warm grade, 35mm anamorphic.Video Generation
Transform the scene into a warm summer festival evening. Replace the bright midday sky with a deep golden-hour sky fading into dusk. Crisscross strings of warm glowing string lights and small paper lanterns between the houses above the street. Light up the windows with warm yellow interior light. Add bursts of colorful fireworks rising over the sea in the background. Add a few softly motion-blurred locals dancing on the sides of the street. Keep the woman's identity, yellow sundress, straw hat, walking motion, and exact camera framing identical. Cinematic Italian summer evening look.Video Generation
Transform the scene into a glamorous Hollywood movie premiere at night. Replace the daytime sky with a deep evening sky with klieg searchlight beams sweeping across it. Roll out a long red carpet down the center of the sidewalk where he is walking, lined with velvet ropes on both sides. Add a dense crowd of paparazzi photographers with bright camera flashes constantly popping on the left side, and an excited cheering crowd of fans behind the velvet ropes on the right side, slightly motion-blurred. Light up the theater marquee in the background with bright glowing white bulbs and a movie title. Add subtle haze and warm spotlight beams crossing through the air. Keep the man's identity, navy suit, walking motion, and exact camera framing identical. Cinematic premiere-night look.Video Generation
API quick start
Run happyhorse 1.0 reference with a single API call.
POST https://api.wiro.ai/v1/Run/alibaba/happyhorse 1.0 reference
{
"prompt": "The young woman is sitting on a wooden be…",
"mode": "videoedit",
"resolution": "720P",
"inputVideo": "https://your-cdn.com/input.mp4"
}