HeyGen

HeyGen Avatar IV

HeyGen Avatar IV is a photorealistic AI avatar generation model that creates talking videos from a single image and a script or audio input. The model synchronizes voice with facial motion, expressions, and gestures to produce lifelike avatar performances. It supports multilingual speech, realistic lip synchronization, and expressive body language, enabling scalable production of presenter-style videos without cameras, actors, or studio setups.

Model AIR ID: heygen:avatar@4.

Supported workflows: Text-to-video, Image-to-video, Audio-to-video.

Technical specifications:

• Positive prompt (positivePrompt): 1–5,000 characters. Optional scene description.
• Resolution (resolution): 720p or 1080p. Alternative to specifying width/height.
• Supported dimensions: 1280×720, 720×1280, 1920×1080, 1080×1920.
• Avatar (inputs.avatar): Built-in avatar ID. One of inputs.avatar or inputs.image is required.
• Custom avatar image (inputs.image): Image to animate as avatar (UUID, URL, Data URI, or Base64).
• Audio input (inputs.audio): Provide audio directly (UUID or URL). Mutually exclusive with speech.text.
• Background image (inputs.background): Custom background image (UUID, URL, Data URI, or Base64).
• Speech text (speech.text): Text to convert to speech. Required if inputs.audio is not provided.
• Voice (speech.voice): Voice identifier (default: auto). See voice library below.
• Speech speed (speech.speed): 0.5–1.5 (default: 1).
• Pitch (speech.pitch): -50 to 50 (default: 0).
• Language (speech.language): Language code for speech generation. See language list below.
• Expressiveness (settings.expressiveness): low (default), medium, high.
• Remove background (settings.removeBackground): Boolean (default: false).
• Background color (settings.backgroundColor): Hex color code (e.g., #ff0000).

Avatar library:

Voice library:

Supported languages:

{
  "taskType": "videoInference",
  "taskUUID": "b1c2d3e4-f5a6-7890-abcd-ef1234567890",
  "model": "heygen:avatar@4",
  "inputs": {
    "avatar": "woman_business_formal"
  },
  "speech": {
    "text": "Welcome to our quarterly earnings review. Today I will walk you through the key performance indicators and financial highlights for Q3 2026.",
    "voice": "sophia_narration_friendly_female_english"
  },
  "resolution": "1080p"
}

{
  "taskType": "videoInference",
  "taskUUID": "c2d3e4f5-a6b7-8901-bcde-f12345678901",
  "model": "heygen:avatar@4",
  "inputs": {
    "image": "c64351d5-4c59-42f7-95e1-eace013eddab"
  },
  "speech": {
    "text": "Hey everyone! Just wanted to share a quick update about our new feature launch. It is going to be amazing!",
    "voice": "chill_brian_male_english",
    "speed": 1.1
  },
  "width": 1280,
  "height": 720,
  "settings": {
    "expressiveness": "high"
  }
}

{
  "taskType": "videoInference",
  "taskUUID": "d3e4f5a6-b7c8-9012-cdef-123456789012",
  "model": "heygen:avatar@4",
  "inputs": {
    "avatar": "man_casual_young_adult",
    "audio": "a1b2c3d4-e5f6-7890-abcd-ef1234567890",
    "background": "https://example.com/office-background.jpg"
  },
  "resolution": "1080p",
  "settings": {
    "expressiveness": "medium"
  }
}

{
  "taskType": "videoInference",
  "taskUUID": "e4f5a6b7-c8d9-0123-def0-234567890123",
  "model": "heygen:avatar@4",
  "inputs": {
    "avatar": "woman_casual_sitting"
  },
  "speech": {
    "text": "Bienvenidos a nuestra plataforma de aprendizaje en línea. Hoy exploraremos las nuevas funciones disponibles para todos los estudiantes.",
    "voice": "luca_sabia_friendly_female_spanish",
    "language": "es-ES",
    "speed": 0.9
  },
  "resolution": "720p",
  "settings": {
    "removeBackground": true,
    "backgroundColor": "#1a1a2e"
  }
}

HeyGen Video Agent

HeyGen Video Agent is an AI video production model that generates complete, multi-scene videos from a single text prompt. It automates the full production pipeline — scriptwriting, avatar selection, shot planning, B-roll integration, motion graphics, captions, and editing — producing broadcast-ready videos with consistent branding. The agent supports customizable avatars and iterative editing, enabling scalable video content creation for marketing, training, and social media.

Model AIR ID: heygen:video-agent@0.

Supported workflows: Text-to-video.

Technical specifications:

• Positive prompt (positivePrompt): 1–10,000 characters. Describes the video topic, structure, and style.
• Supported dimensions: 1280×720 (16:9), 720×1280 (9:16).
• Duration (duration): 5–300 seconds.
• Avatar (inputs.avatar): Optional. Select a built-in avatar for presenter-style delivery (see avatar library below).
• Custom avatar image (inputs.image): Optional. Provide a custom image (UUID, URL, Data URI, or Base64) to animate as an avatar.

Avatar library: