Aliyun Happyhorse R2v

HappyHorse 1.0 Reference-to-Video

Validation

mkdir -p output/aliyun-happyhorse-r2v
python -m py_compile skills/ai/video/aliyun-happyhorse-r2v/scripts/r2v_happyhorse.py && echo "py_compile_ok" > output/aliyun-happyhorse-r2v/validate.txt

Pass criteria: command exits 0 and output/aliyun-happyhorse-r2v/validate.txt is generated.

Output And Evidence

• Save task IDs, polling responses, and final video URLs to output/aliyun-happyhorse-r2v/.
• Keep at least one end-to-end run log for troubleshooting.

Prerequisites

• Install dependencies (recommended in a venv):

python3 -m venv .venv
. .venv/bin/activate
python -m pip install requests

• Set DASHSCOPE_API_KEY in your environment, or add dashscope_api_key to ~/.alibabacloud/credentials.

Critical model names

• happyhorse-1.0-r2v — reference-to-video; 1-9 reference images fused into a single video, with character1..N references in the prompt

Capabilities

API endpoint (async only)

POST https://dashscope.aliyuncs.com/api/v1/services/aigc/video-generation/video-synthesis

Required headers:

• Authorization: Bearer $DASHSCOPE_API_KEY
• Content-Type: application/json
• X-DashScope-Async: enable

Singapore endpoint: replace dashscope.aliyuncs.com with dashscope-intl.aliyuncs.com.

Polling endpoint: GET https://dashscope.aliyuncs.com/api/v1/tasks/{task_id} — recommended interval 15s.

Normalized interface

Request

• model (string, required) — fixed happyhorse-1.0-r2v
• input.prompt (string, required) — up to 5000 non-CJK / 2500 CJK characters; reference subjects via character1, character2, ... matching media array order
• input.media (array, required) — 1 to 9 elements, each:
  • type: reference_image
  • url: public HTTP/HTTPS URL of a reference image
• parameters.resolution (string, optional) — 720P or 1080P (default: 1080P)
• parameters.ratio (string, optional) — 16:9 (default), 9:16, 1:1, 4:3, 3:4
• parameters.duration (integer, optional) — video length in seconds, range [3, 15] (default: 5)
• parameters.watermark (boolean, optional) — bottom-right "Happy Horse" watermark (default: true)
• parameters.seed (integer, optional) — range [0, 2147483647]

Media input limits

Reference image (type=reference_image):

• Formats: JPEG, JPG, PNG, WEBP
• Resolution: short side ≥ 400 pixels (720P or higher recommended)
• Max size: 10 MB per image
• Avoid blurry, over-compressed, or very small images

Character indexing rule

The first reference_image in the media array maps to character1, the second to character2, and so on up to character9. Reorder the array if you want a specific reference to bind to a specific characterN.

Response (task creation)

• output.task_id (string) — valid 24 hours
• output.task_status (string) — PENDING | RUNNING | SUCCEEDED | FAILED | CANCELED | UNKNOWN
• request_id (string)

Response (task result, on SUCCEEDED)

• output.video_url (string) — generated MP4 (H.264) URL, valid 24 hours
• output.orig_prompt (string)
• output.submit_time / output.scheduled_time / output.end_time (string)
• usage.duration (integer) — billable duration in seconds
• usage.output_video_duration (integer)
• usage.input_video_duration (integer) — fixed 0 for r2v
• usage.SR (integer) — output resolution tier
• usage.ratio (string)
• usage.video_count (integer) — fixed 1

Quick start (Python + HTTP)

import os
import time
import requests

API_KEY = os.getenv("DASHSCOPE_API_KEY")
BASE_URL = "https://dashscope.aliyuncs.com/api/v1"


def create_r2v_task(req: dict) -> str:
    """Create a reference-to-video task and return task_id."""
    refs = req["reference_images"]
    if not 1 <= len(refs) <= 9:
        raise ValueError("Need 1-9 reference images")
    payload = {
        "model": "happyhorse-1.0-r2v",
        "input": {
            "prompt": req["prompt"],
            "media": [{"type": "reference_image", "url": u} for u in refs],
        },
        "parameters": {
            "resolution": req.get("resolution", "1080P"),
            "ratio": req.get("ratio", "16:9"),
            "duration": req.get("duration", 5),
            "watermark": req.get("watermark", True),
        },
    }
    if req.get("seed") is not None:
        payload["parameters"]["seed"] = req["seed"]

    resp = requests.post(
        f"{BASE_URL}/services/aigc/video-generation/video-synthesis",
        headers={
            "Authorization": f"Bearer {API_KEY}",
            "Content-Type": "application/json",
            "X-DashScope-Async": "enable",
        },
        json=payload,
    )
    resp.raise_for_status()
    return resp.json()["output"]["task_id"]


def poll_task(task_id: str, interval: int = 15) -> dict:
    while True:
        resp = requests.get(
            f"{BASE_URL}/tasks/{task_id}",
            headers={"Authorization": f"Bearer {API_KEY}"},
        )
        resp.raise_for_status()
        data = resp.json()
        if data["output"]["task_status"] in ("SUCCEEDED", "FAILED", "CANCELED"):
            return data
        time.sleep(interval)

Usage examples

# Single subject — character1 references the only image
task_id = create_r2v_task({
    "reference_images": ["https://example.com/girl.jpg"],
    "prompt": "character1 walks slowly through a sunlit forest, cinematic shot.",
    "duration": 5,
})

# Multi-subject — character1=girl, character2=fan, character3=earring
task_id = create_r2v_task({
    "reference_images": [
        "https://example.com/girl.jpg",
        "https://example.com/folding-fan.jpg",
        "https://example.com/earring.jpg",
    ],
    "prompt": (
        "身着红色旗袍的女性 character1，轻抬玉手展开折扇 character2 时"
        "流苏耳坠 character3 随头部转动轻盈摆动。"
    ),
    "resolution": "720P",
    "ratio": "16:9",
    "duration": 5,
})

Error handling

Output location

• Default output: output/aliyun-happyhorse-r2v/videos/
• Override base dir with OUTPUT_DIR.

Anti-patterns

• Do not use any model ID other than happyhorse-1.0-r2v.
• Do not call this API synchronously — async header is required.
• Do not pass first_frame, last_frame, driving_audio, or video — only reference_image entries are accepted.
• Do not exceed 9 reference images, and do not omit the array entirely.
• Do not forget characterN tokens in the prompt — without them the model has no link from prompt to image.
• Video URLs expire after 24 hours; download and persist immediately.
• Do not use this skill for pure text-to-video (aliyun-happyhorse-t2v), single-image first-frame (aliyun-happyhorse-i2v), or video editing (aliyun-happyhorse-videoedit).

Workflow

1. Collect 1-9 high-quality reference images and decide character order.
2. Write a prompt that uses character1..N to bind subjects to references.
3. Create async task and poll /tasks/{task_id} every ~15s.
4. Download output.video_url before the 24-hour expiration.

References

• See references/api_reference.md for full HTTP API details.
• See references/sources.md for source links.