Wan 2.1 V2V 480p LoRA Ultra Fast API Documentation

Wan 2.1 V2V 480p LoRA Ultra Fast API Documentation

Playground

Wan 2.1 V2V 480p is an ultra-fast video-to-video model that generates unlimited AI videos and supports custom LoRAs for personalization. Ready-to-use REST inference API, best performance, no coldstarts, affordable pricing.

Features

Wan 2.1 V2V 480p LoRA Ultra Fast is a speed-optimized video-to-video model for prompt-guided edits while preserving the original motion and timing of an input video. Upload a source video, describe what should change, and tune strength to control how closely the output follows the original footage. It supports up to 3 LoRAs to enforce a consistent style, character look, or branded aesthetic—now with lower latency for rapid iteration.

Key capabilities

• Ultra-fast video-to-video transformation anchored to an input video (480p output)
• Prompt-guided edits while keeping motion continuity and pacing
• Strength control to balance preservation vs. transformation
• LoRA support (up to 3) for stable style/identity steering
• Fine control over motion behavior via flow_shift

Use cases

• Rapid V2V restyling for social clips and creative iteration
• Apply a consistent “house style” across multiple clips using LoRAs
• Lighting/mood changes (cinematic grade, neon, golden hour) without re-animating motion
• Brand-safe refresh: keep composition and timing, update textures/colors/details
• Quick A/B testing by changing prompts, LoRAs, or seed

Pricing

Inputs

• video (required): source video to transform
• prompt (required): what to change and how the result should look
• negative_prompt (optional): what to avoid (artifacts, jitter, unwanted elements)
• loras (optional): up to 3 LoRA items for style/identity steering

Parameters

• num_inference_steps: sampling steps
• duration: output duration (seconds)
• strength: how strongly to transform the input video (lower = preserve more; higher = change more)
• guidance_scale: prompt adherence strength
• flow_shift: motion/flow behavior tuning
• seed: random seed (-1 for random; fixed for reproducible results)

LoRA (up to 3 items):

• loras: list of LoRA entries (max 3)

• path: owner/model-name or a direct.safetensors URL

• scale: LoRA strength

Prompting guide (V2V)

Write prompts that explicitly separate preservation from transformation:

Template: Keep the same camera motion and timing from the input video. Change [style/lighting/environment]. Keep faces natural and consistent. Avoid flicker and warping.

Example prompts

• Keep the original motion and composition. Apply a candid, cinematic look with warm sunlight, soft depth of field, and natural skin texture.
• Preserve timing and camera movement. Restyle into a clean anime look with consistent shading and no flicker.
• Keep the same scene and people. Change the color grade to sunset golden hour, add subtle lens flare, maintain realistic shadows.

Authentication

For authentication details, please refer to the Authentication Guide.

API Endpoints

Submit Task & Query Result

set -euo pipefail

export WAVESPEED_API_KEY="your-api-key"

REQUEST_BODY=$(cat <<'JSON'
{
  "prompt": "A cinematic ocean wave at sunrise, highly detailed",
  "video": "https://interactive-examples.mdn.mozilla.net/media/cc0-videos/flower.mp4",
  "num_inference_steps": 30,
  "duration": 5,
  "strength": 0.9,
  "guidance_scale": 5,
  "flow_shift": 3
}
JSON
)

# 1. Submit the prediction once. Never retry this request automatically:
# a retry after an unclear failure can create a second, billed task.
CHECK='The task may already exist. Check History or the predictions list before submitting again.'
if ! SUBMIT_RESPONSE=$(curl --silent --show-error --fail-with-body \
  --connect-timeout 10 --max-time 60 \
  -X POST "https://api.wavespeed.ai/api/v3/wavespeed-ai/wan-2.1/v2v-480p-lora-ultra-fast" \
  -H "Authorization: Bearer ${WAVESPEED_API_KEY}" \
  -H "Content-Type: application/json" \
  -d "${REQUEST_BODY}"); then
  printf 'Submit failed. %s\n' "${CHECK}" >&2
  exit 1
fi

TASK=$(printf '%s' "${SUBMIT_RESPONSE}" | jq 'if type == "object" and has("data") then .data else . end')
PREDICTION_ID=$(printf '%s' "${TASK}" | jq -r '.id // empty')
if [ -z "${PREDICTION_ID}" ]; then
  printf 'Submission response did not contain a prediction id. %s\n' "${CHECK}" >&2
  exit 1
fi
RESULT_URL="https://api.wavespeed.ai/api/v3/predictions/${PREDICTION_ID}/result"

# 2. Poll until the prediction reaches a final status.
BODY_FILE=$(mktemp)
trap 'rm -f "${BODY_FILE}"' EXIT
while true; do
  # Transient errors (429, 5xx) are retried, honoring Retry-After.
  HTTP_CODE=$(curl --silent --show-error --fail-with-body \
    --connect-timeout 10 --max-time 60 --retry 5 --retry-delay 2 \
    --output "${BODY_FILE}" --write-out '%{http_code}' \
    "${RESULT_URL}" \
    -H "Authorization: Bearer ${WAVESPEED_API_KEY}" || true)
  case "${HTTP_CODE}" in
    2??) ;;
    000|429|5??)  # dropped connection or temporary error: keep polling
      printf 'Task %s: HTTP %s, polling again in 10s\n' "${PREDICTION_ID}" "${HTTP_CODE}" >&2
      sleep 10; continue ;;
    *) printf 'Task %s: HTTP %s\n' "${PREDICTION_ID}" "${HTTP_CODE}" >&2
      cat "${BODY_FILE}" >&2; exit 1 ;;
  esac
  RESULT=$(jq 'if type == "object" and has("data") then .data else . end' "${BODY_FILE}")
  STATUS=$(printf '%s' "${RESULT}" | jq -r '.status // empty')

  case "${STATUS}" in
    completed) printf '%s\n' "${RESULT}" | jq '.outputs'; break ;;
    failed|cancelled|timeout|deleted)
      printf 'Task %s ended with status %s\n' "${PREDICTION_ID}" "${STATUS}" >&2
      printf '%s\n' "${RESULT}" | jq . >&2; exit 1 ;;
    *) sleep 2 ;;
  esac
done

Parameters

Task Submission Parameters

Request Parameters

Every model also accepts the global request options enable_sync_mode and enable_base64_output; see their pages for details.

Response Parameters

Result Request Parameters

Result Response Parameters