Skyreels V3 Talking Avatar

Skyreels V3 Talking Avatar

Playground

SkyReels V3 Talking Avatar is a 19B-parameter multimodal model that generates talking avatars from portrait and audio with precise lip sync, supporting up to 20 seconds at 720p resolution. Ready-to-use REST inference API, best performance, no coldstarts, affordable pricing.

Features

SkyReels V3 Talking Avatar generates realistic talking head videos by combining a portrait image with audio. Upload a face image and audio clip — the model creates a lip-synced video with natural facial movements and expressions that match the speech.

Why Choose This?

• Realistic lip-sync Generates accurate lip movements synchronized to the audio input.

• Natural expressions Creates realistic facial movements and expressions during speech.

• Flexible duration Support for audio clips up to 20 seconds.

• Resolution options Choose between standard and 720p output quality.

• Simple workflow Just upload an image and audio — no complex setup required.

• Prompt Enhancer Optional prompt to guide the generation style.

Parameters

How to Use

1. Upload your image — provide a clear portrait image with visible face.
2. Upload your audio — add the audio clip you want the avatar to speak (up to 20 seconds).
3. Add prompt (optional) — describe any specific style or expression preferences.
4. Select resolution — choose 720p for higher quality or lower for faster processing.
5. Run — submit and download your talking avatar video.

Pricing

Billing Rules

• Minimum charge: 5 seconds
• Maximum duration: 20 seconds
• 720p rate: 2× standard rate
• Standard rate: $0.15 per 5 seconds

Best Use Cases

• Digital Presenters — Create talking avatars for video presentations and tutorials.
• Social Media Content — Generate engaging talking head videos for TikTok, Reels, and Shorts.
• Marketing & Ads — Produce spokesperson videos without filming.
• E-learning — Create instructor avatars for educational content.
• Localization — Generate lip-synced videos in different languages from the same avatar.

Pro Tips

• Use clear, front-facing portrait images with good lighting for best results.
• Ensure the face is clearly visible and not obscured by hair or accessories.
• Use high-quality audio with clear speech for accurate lip-sync.
• Keep audio clips within 15 seconds for optimal results.
• Set a specific seed for reproducible results across multiple generations.

Notes

• Both image and audio are required fields.
• Maximum audio duration: 20 seconds.
• Minimum charge: 5 seconds (shorter clips billed as 5 seconds).
• 720p resolution costs 2× the standard rate.
• Ensure uploaded file URLs are publicly accessible.

Related Models

• DreamActor V2 — Motion transfer from driving video to character image.
• Wan 2.6 Image-to-Video — General image-to-video generation.
• AI Twerk — Fun dancing video generation from images.

Duration limit

The maximum supported audio duration is 200 seconds. Longer audio is automatically trimmed to 200 seconds before processing.

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'
{
  "image": "https://interactive-examples.mdn.mozilla.net/media/cc0-images/painted-hand-298-332.jpg",
  "audio": "https://interactive-examples.mdn.mozilla.net/media/cc0-audio/t-rex-roar.mp3",
  "resolution": "720p",
  "seed": -1
}
JSON
)

# 1. Submit the prediction.
SUBMIT_RESPONSE=$(curl --silent --show-error --fail-with-body \
  -X POST "https://api.wavespeed.ai/api/v3/wavespeed-ai/skyreels-v3/talking-avatar" \
  -H "Authorization: Bearer ${WAVESPEED_API_KEY}" \
  -H "Content-Type: application/json" \
  -d "${REQUEST_BODY}")

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 "${PREDICTION_ID}" ]; then
  printf 'Submission response did not contain a prediction id
' >&2
  exit 1
fi
RESULT_URL=$(printf '%s' "${TASK}" | jq -r '.urls.get // empty')
if [ -z "${RESULT_URL}" ]; then RESULT_URL="https://api.wavespeed.ai/api/v3/predictions/${PREDICTION_ID}/result"; fi

# 2. Poll until the prediction finishes.
while true; do
  RESPONSE=$(curl --silent --show-error --fail-with-body \
    "${RESULT_URL}" \
    -H "Authorization: Bearer ${WAVESPEED_API_KEY}")
  RESULT=$(printf '%s' "${RESPONSE}" | jq 'if type == "object" and has("data") then .data else . end')
  STATUS=$(printf '%s' "${RESULT}" | jq -r '.status // empty')

  case "${STATUS}" in
    completed) printf '%s\n' "${RESULT}" | jq '.outputs'; break ;;
    failed|cancelled|timeout) printf '%s\n' "${RESULT}" | jq . >&2; exit 1 ;;
    created|processing) sleep 2 ;;
    *) printf 'Unexpected status: %s
' "${STATUS}" >&2; exit 1 ;;
  esac
done

Parameters

Task Submission Parameters

Request Parameters

Response Parameters

Result Request Parameters

Result Response Parameters