Image-to-Video API — Input Handling and Workflow

How Image-to-Video Works

Every image-to-video model on Muapi takes a source image_url and a text prompt describing the motion, camera move, or scene continuation, and returns a video that animates from that starting frame. The source image sets the visual content and, on most models, the output aspect ratio — submitting a portrait image to a model without an explicit aspect_ratio parameter typically produces a portrait video, not a cropped landscape one.

Muapi exposes 150+ image-to-video models — Kling, Veo, Wan, Seedance, Runway, PixVerse, Vidu, MiniMax, and more — through the same unified REST pattern: one API key, one request/poll flow, pay-per-generation pricing with no subscription.

First/Last-Frame Control (Start-End-to-Video)

Many image-to-video models accept a second image — usually named last_image (Seedance, Kling, Wan, Veo, Vidu) or end_image_url (MiniMax Hailuo) — that fixes the final frame instead of leaving the model to freely animate the motion. This is the standard technique for a controlled transition between two known states: product A turning into product B, a before/after reveal, or a scene that must land on a specific composition. Vidu Q2 Pro Start-End additionally exposes a movement_amplitude parameter to control how much motion happens between the two endpoints. Not every model supports this — Runway's and PixVerse's image-to-video endpoints, for example, take only a single starting image — so check the model's parameter list before assuming first/last-frame support.

Representative Models

Duration, Resolution, and Prompt Structure

• Duration is typically a fixed choice (commonly 5s or 8s) rather than a free-form value — check each model's enum before submitting an arbitrary number.
• Resolution usually trades directly against price and generation time — draft at 480p/720p, re-run the final take at 1080p once the motion and composition are confirmed.
• Some models split resolution and quality into separate parameters (Wan 2.2), letting you keep resolution fixed while trading off a faster/cheaper generation mode.
• Prompt structure matters more for motion than for content — the image already fixes the subject and composition, so the prompt should describe camera movement, action, and pacing rather than re-describing what's already visible in the source image.
• A fixed-camera option (Seedance's camera_fixed) is useful when you want subject motion only, with no camera drift — check for this before assuming every model free-animates the camera by default.

Output Persistence — Download Before the URL Expires

Generated video URLs are hosted by the underlying provider and are not permanent — how long they stay reachable depends on which provider is behind the model. As a rule of thumb: WaveSpeed-backed and Runway-backed outputs stay reachable for about 7 days, while Replicate-backed and fal.ai-backed outputs typically expire after about 1 hour. Muapi does not expose which provider serves a given request, so the safe pattern is to always download and persist the output to your own storage immediately after the request completes, rather than storing only the returned URL for later use.

How to Call an Image-to-Video Model

1. Host the source image. Upload via POST /api/v1/upload_file, or use any publicly reachable URL.
2. Submit the request. POST /api/v1/{model-slug} with image_url, a motion-describing prompt, and any duration/resolution/first-last-frame fields the model supports.
3. Poll for completion. Check GET /api/v1/predictions/{request_id}/result until status is completed.
4. Download immediately. Save the output video to your own storage rather than relying on the returned URL staying valid long-term.

curl -X POST https://api.muapi.ai/api/v1/wan2.2-image-to-video \
  -H "Content-Type: application/json" \
  -H "x-api-key: YOUR_API_KEY" \
  -d '{
    "image_url": "https://example.com/product-shot.jpg",
    "prompt": "slow push-in, gentle studio light sweep",
    "resolution": "720p",
    "duration": 5
  }'

import requests

response = requests.post(
    "https://api.muapi.ai/api/v1/wan2.2-image-to-video",
    headers={"x-api-key": "YOUR_API_KEY"},
    json={
        "image_url": "https://example.com/product-shot.jpg",
        "prompt": "slow push-in, gentle studio light sweep",
        "resolution": "720p",
        "duration": 5,
    },
)
request_id = response.json()["request_id"]

result = requests.get(
    f"https://api.muapi.ai/api/v1/predictions/{request_id}/result",
    headers={"x-api-key": "YOUR_API_KEY"},
)
print(result.json())  # download result["output_url"] immediately

Frequently Asked Questions