Multitalk API Documentation

Multitalk API Documentation

Playground

MultiTalk converts one image and audio into audio-driven talking/singing videos (Image-to-Video), supporting up to 10 minutes. Ready-to-use REST inference API, best performance, no coldstarts, affordable pricing.

Features

Generate realistic talking videos from a single photo with MultiTalk — MeiGen-AI’s revolutionary audio-driven conversational video framework. Unlike traditional talking head methods that only animate facial movements, MultiTalk creates lifelike videos with perfect lip synchronization, natural expressions, and dynamic body language.

Why It Looks Great

• Perfect lip sync: Advanced audio analysis ensures precise mouth movements matching every syllable.
• Full-body animation: Goes beyond faces — animates natural body movements and gestures.
• Camera dynamics: Built-in Uni3C controlnet enables subtle camera movements for professional results.
• Instruction following: Control scene, pose, and behavior through text prompts while maintaining sync.
• Multi-person support: Animate conversations with multiple speakers in the same scene.
• Extended duration: Generate videos up to 10 minutes long.

How It Works

MultiTalk combines three powerful technologies for optimal results:

Through sophisticated attention mechanisms, MultiTalk perfectly aligns lip movements with audio while maintaining natural facial expressions and body language.

Parameters

How to Use

1. Upload your image — a clear portrait photo works best.
2. Upload your audio — speech, singing, or any vocal audio.
3. Run — click the button to generate.
4. Download — preview and save your talking video.

Pricing

Per 5-second billing based on audio duration.

Best Use Cases

• Virtual Presenters — Create AI spokespeople for videos and training content.
• Content Localization — Dub content into different languages with matching lip movements.
• Music Videos — Generate singing performances from static photos.
• E-learning — Produce instructor-led courses without filming.
• Social Media — Create engaging talking-head content at scale.
• Multi-person Conversations — Animate group discussions and dialogues.

Pro Tips for Best Results

• Use clear, front-facing portrait photos with good lighting.
• Ensure faces are clearly visible without obstructions.
• High-quality audio with minimal background noise produces better sync.
• Neutral or slightly open mouth expressions in source images work best.
• For conversations, provide distinct audio tracks for each speaker.
• Test with shorter clips before generating longer videos.

Related Workflows

• Wan2.1 T2V/I2V — Text-to-video and image-to-video generation
• Uni3C Camera Control — Camera motion transfer for dynamic videos

Notes

• Maximum supported video length is 10 minutes.
• If using URLs, ensure they are publicly accessible.
• Processing time scales with audio duration.
• Best results come from portrait-style images with clear facial features.

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",
  "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/multitalk" \
  -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