Wan v2.2 A14B Image-to-Video A14B with LoRAs

video

fal-ai/wan/v2.2-a14b/image-to-video/lora

Animate a still image with Wan 2.2 and your motion LoRAs.

Image-to-video on Wan 2.2 A14B with LoRA support: give it a start image, a prompt, and the LoRAs you trained with the Wan 2.2 trainers. It exposes deep video controls, from frame interpolation to a second-stage guidance scale, but the defaults are sensible.

Open in fal playground ↗Official API docs ↗

Parameters

Schema facts from the fal API; the notes are ours.

Required

image_urlstringrequired

Input image for image-to-image, editing or video conditioning.

Raw schema description

URL of the input image. If the input image does not match the chosen aspect ratio, it is resized and center cropped.

promptstringrequired

The text description of what to generate.

Tip: If your LoRA used a trigger word, include it. Describe the scene around the subject normally.

Raw schema description

The text prompt to guide video generation.

Optional

num_framesintegerdefault: 8117 – 161

Number of video frames to generate.

Raw schema description

Number of frames to generate. Must be between 17 to 161 (inclusive).

frames_per_secondintegerdefault: 164 – 60

Frames per second of the generated video. Must be between 4 to 60. When using interpolation and `adjust_fps_for_interpolation` is set to true (default true,) the final FPS will be multiplied by the number of interpolated frames plus one. For example, if the generated frames per second is 16 and the number of interpolated frames is 1, the final frames per second will be 32. If `adjust_fps_for_interpolation` is set to false, this value will be used as-is.

negative_promptstring

What the model should avoid generating.

Raw schema description

Negative prompt for video generation.

seedinteger

Random seed. Same seed plus same inputs gives a nearly identical image.

Tip: Fix the seed when comparing LoRA scales or parameters, so the only thing changing is the thing you are testing.

Raw schema description

Random seed for reproducibility. If None, a random seed is chosen.

resolutionenumdefault: 720p480p | 580p | 720p

Output or training resolution.

Tip: Higher costs more and trains slower. Match it to how you will actually generate.

Raw schema description

Resolution of the generated video (480p, 580p, or 720p).

aspect_ratioenumdefault: autoauto | 16:9 | 9:16 | 1:1

Aspect ratio of training samples or generated output.

Raw schema description

Aspect ratio of the generated video. If 'auto', the aspect ratio will be determined automatically based on the input image.

num_inference_stepsintegerdefault: 272 – 40

Number of denoising steps per image. More steps, more refinement, more latency.

Tip: Defaults are tuned per model. Cutting steps in half is the quickest way to trade quality for speed.

Raw schema description

Number of inference steps for sampling. Higher values give better quality but take longer.

enable_safety_checkerbooleandefault: false

Runs a safety filter on outputs.

Raw schema description

If set to true, input data will be checked for safety before processing.

enable_output_safety_checkerbooleandefault: false

If set to true, output video will be checked for safety after generation.

enable_prompt_expansionbooleandefault: false

Lets the endpoint rewrite your prompt with more detail before generating.

Raw schema description

Whether to enable prompt expansion. This will use a large language model to expand the prompt with additional details while maintaining the original meaning.

accelerationenumdefault: regularnone | regular

Speed and cost preset for the generation backend.

Tip: 'regular' balances quality and latency. Use higher acceleration for drafts and iteration.

Raw schema description

Acceleration level to use. The more acceleration, the faster the generation, but with lower quality. The recommended value is 'regular'.

guidance_scalenumberdefault: 3.51 – 10

How strictly generation follows the prompt (classifier-free guidance).

In the atelier: How tightly you hold the painter to the brief. Too tight and the work gets stiff and oversaturated; too loose and he wanders.

Tip: Stay near the endpoint default. Adjust in steps of 0.5.

Raw schema description

Classifier-free guidance scale. Higher values give better adherence to the prompt but may decrease quality.

guidance_scale_2numberdefault: 41 – 10

Guidance for the second stage of Wan 2.2's two-stage sampler, separate from the main guidance_scale.

Raw schema description

Guidance scale for the second stage of the model. This is used to control the adherence to the prompt in the second stage of the model.

shiftnumberdefault: 51 – 10

Shift value for the video. Must be between 1.0 and 10.0.

interpolator_modelenumdefault: filmnone | film | rife

The model to use for frame interpolation. If None, no interpolation is applied.

num_interpolated_framesintegerdefault: 10 – 4

Extra frames synthesized between each generated pair, smoothing motion and raising the effective frame rate.

Tip: The default of 1 doubles 16 fps output to 32 fps. Set 0 to see the raw model motion.

Raw schema description

Number of frames to interpolate between each pair of generated frames. Must be between 0 and 4.

adjust_fps_for_interpolationbooleandefault: true

If true, the number of frames per second will be multiplied by the number of interpolated frames plus one. For example, if the generated frames per second is 16 and the number of interpolated frames is 1, the final frames per second will be 32. If false, the passed frames per second will be used as-is.

video_qualityenumdefault: highlow | medium | high | maximum

The quality of the output video. Higher quality means better visual quality but larger file size.

video_write_modeenumdefault: balancedfast | balanced | small

The write mode of the output video. Faster write mode means faster results but larger file size, balanced write mode is a good compromise between speed and quality, and small write mode is the slowest but produces the smallest file size.

loraslistdefault: []

List of LoRA weights to load, each with a path (URL or HF repo) and a scale.

In the atelier: Which bracelets the painter wears for this painting, and how hard he leans on each.

Tip: Order does not matter; scales do. Start every LoRA at 1.0 and adjust one at a time.

Watch out: Stacking three strong LoRAs usually degrades all of them. Lower each scale when combining.

Raw schema description

LoRA weights to be used in the inference.

reverse_videobooleandefault: false

If true, the video will be reversed.

end_image_urlstring

Optional final frame. The video is generated to land on it.

Raw schema description

URL of the end image.

Call it

import { fal } from "@fal-ai/client";

const result = await fal.subscribe("fal-ai/wan/v2.2-a14b/image-to-video/lora", {
  input: {
    "image_url": "https://your-cdn.com/input.jpg",
    "prompt": "a photo of TOK on a sunny windowsill",
    "loras": [
      {
        "path": "https://your-cdn.com/lora.safetensors",
        "scale": 1
      }
    ]
  },
  logs: true,
});
console.log(result.data);

Train the LoRA with

Wan-2.2 LoRA Trainer

fal-ai/wan-22-trainer/i2v-a14b

Wan-2.2 LoRA Trainer

fal-ai/wan-22-trainer/t2v-a14b

Wan 2.2 14B Image Trainer

fal-ai/wan-22-image-trainer