wan-2.2-spicy/image-to-video-lora

Key Features

• LoRA-powered personalization — apply custom-trained LoRA models to generate videos with unique, consistent visual styles.
• Triple LoRA architecture — standard LoRA, High Noise LoRA, and Low Noise LoRA slots for maximum control over style transfer.
• Up to 3 LoRAs per group — stack multiple LoRAs with individual strength weights for complex style blending (up to 9 total).
• Reproducible results — use the seed parameter to generate deterministic output for consistent iterations.
• Animate still images into smooth, high-quality video with natural motion at 30fps.
• Multiple resolution options — 480p for quick previews or 720p for production, with optional FHD upscaling.
• Flexible duration — generate 5-second or 8-second video clips.

Parameters

How to Use

1. Upload a high-quality source image that you want to animate.
2. Write a prompt describing the desired motion and visual style.
3. Add your LoRA filenames under Standard, High Noise, and/or Low Noise groups, with strength weights (default 1.0).
4. Optionally set a seed for reproducible results.
5. Set duration and resolution, then generate.

Code Examples

import os
import requests

response = requests.post(
    "https://aircube.ai/api/v3/wan-2.2-spicy/image-to-video-lora",
    headers={
        "Authorization": "Bearer " + os.environ["AIRCUBE_API_KEY"],
        "Content-Type": "application/json",
    },
    json={
    "prompt": "Camera slowly orbits around the subject, hair flowing gently in the breeze",
    "image": "https://example.com/portrait.jpg",
    "duration": 5,
    "loras": [
        "style.safetensors"
    ],
    "lora_strengths": [
        1
    ],
    "loras_high": [
        "style_lora_H.safetensors"
    ],
    "lora_strengths_high": [
        1
    ],
    "loras_low": [
        "style_lora_L.safetensors"
    ],
    "lora_strengths_low": [
        1
    ],
    "seed": -1
},
    timeout=300,
)
data = response.json()

if data["success"]:
    print("ID:", data["data"]["id"], "Status:", data["data"]["status"])
else:
    print("Error:", data["error"]["message"])

Pricing

Billing rules

• 480p / 5s: $0.20.
• 720p / 5s: $0.32.
• Longer durations scale proportionally.
• Failed generations are not charged.

Best Use Cases

• Branded content — apply trained brand LoRAs for consistent visual identity across generated videos.
• Character animation — use character-specific LoRAs for stylized, on-brand character motion.
• Art style transfer — blend artistic style LoRAs to animate images in specific visual aesthetics.
• Product showcases — combine product-trained LoRAs with motion prompts for styled marketing videos.

Pro Tips

• Standard LoRA applies across the full generation; High Noise LoRA affects the initial diffusion pass (coarse structure/style); Low Noise LoRA affects the refinement pass (fine details).
• Start with strength 1.0 and adjust — lower values (0.3-0.7) give subtle effects, higher values (1.2-2.0) for stronger style influence.
• You can stack different LoRAs: e.g. one for style + one for character in the same group.
• If the output looks over-stylized, reduce LoRA strength or use fewer groups.
• Use the seed parameter to lock in a good result, then experiment with LoRA combinations while keeping the same seed.
• Use short 5s durations to iterate quickly on LoRA combinations, then extend to 8s for final renders.

Notes

• LoRA files must be in safetensors format and accessible by the backend.
• Each LoRA group (Standard / High / Low) supports up to 3 LoRAs, for a maximum of 9 total.
• If lora_strengths is omitted, each LoRA defaults to strength 1.0.
• Seed -1 (default) generates a random seed each time.
• Input images should be under 10 MB (JPEG, PNG, WebP).