Wan 2.1 I2V 480p LoRA

Wan 2.1 I2V 480p LoRA

Playground

Generate unlimited 480P AI videos with WAN 2.1 Image-to-Video and custom LoRA support for personalized styles. Ready-to-use REST inference API, best performance, no coldstarts, affordable pricing.

Features

WAN 2.1 Image-to-Video 480p LoRA — wavespeed-ai/wan-2.1/i2v-480p-lora

wavespeed-ai/wan-2.1/i2v-480p-lora turns a single input image into a short, coherent video clip at 480p, with optional LoRA effects for stylized motion, transformations, or consistent visual “behavior” across runs.

What it’s good at

• Image-to-video animation from one still image (great for product shots, portraits, scenes)
• LoRA-driven motion/effects (for example: rotate, zoom-call framing, playful VFX-style transformations)
• Fast iteration for storyboard-style testing at a lightweight resolution

Inputs

• Prompt (required): describe the action, camera, and mood

• Image (required): the starting frame

• Negative prompt (optional): reduce artifacts or unwanted styles

• LoRAs (optional): add one or more LoRAs with:

  • Path: LoRA identifier
  • Scale: LoRA strength

• Size: 480p

• Duration: 5s or 10s

• Num inference steps: higher can improve detail/coherence (slower)

• Guidance scale: higher follows the prompt more strictly

• Flow shift: controls motion feel (lower is calmer, higher is more dynamic)

• Seed: set for repeatable results

How to use (Playground)

1. Upload your image.
2. Write a prompt that includes subject + action + camera movement + environment.
3. (Optional) Add a LoRA and start with a moderate scale, then adjust upward only if the effect is too subtle.
4. Keep guidance scale moderate for natural motion; increase it if the model ignores your requested action.
5. Run at 5s first, then move to 10s once the motion looks right.

Prompting tips (I2V)

• Start with a single clear action: walks, turns, smiles, waves, looks back, picks up an object.
• Add camera direction in plain film language: slow push-in, handheld, dolly left, tilt up, rack focus.
• If you use an “effect” LoRA (like deflate/rotate), describe the target as a non-graphic, clearly fictional or toy-like transformation to avoid unwanted safety filtering.

Example prompt style (safe wording): A beach photo becomes a playful VFX shot: the character turns into an inflatable toy version of itself and slowly deflates like a balloon, then settles gently on the sand. Smooth camera push-in, soft sunlight, natural motion, no distortion.

Use cases

• Social content and short ads: animate a key visual into a 5–10s loopable clip
• Product hero animation: subtle parallax, gentle rotation, light camera push for ecommerce
• Stylized LoRA effects: rotate, “video call” framing, playful toon-like transformations
• Prototype storytelling: turn concept art into quick motion beats for pitches

Pricing

Notes

• 480p is ideal for fast iteration, previews, and high-volume generation.
• LoRA effects can overpower the base motion if scaled too high; increase gradually.
• Safety checks may block certain real-person transformations; using fictional/toy framing typically works better.

# Submit the task
curl --location --request POST "https://api.wavespeed.ai/api/v3/wavespeed-ai/wan-2.1/i2v-480p-lora" \
--header "Content-Type: application/json" \
--header "Authorization: Bearer ${WAVESPEED_API_KEY}" \
--data-raw '{
    "loras": [
        {
            "path": "Remade-AI/Deflate",
            "scale": 1
        }
    ],
    "size": "832*480",
    "num_inference_steps": 30,
    "duration": 5,
    "guidance_scale": 5,
    "flow_shift": 3,
    "seed": -1
}'

# Get the result
curl --location --request GET "https://api.wavespeed.ai/api/v3/predictions/${requestId}/result" \
--header "Authorization: Bearer ${WAVESPEED_API_KEY}"