← LeaderboardThe Skill Leaderboard

heygen

HeyGen avatar videos, photo animation, lipsync, video translation, TTS and voice cloning via inference.sh CLI. Apps: heygen/avatar-video (Avatar IV/V engines, up to 4K), heygen/create-avatar (digital twin, photo or prompt avatars), heygen/photo-video, heygen/lipsync, heygen/video-agent (prompt to finished video), heygen/video-translate (30+ languages with voice cloning and lip-sync), heygen/text-to-speech (Starfish), heygen/voice-clone (instant clone). Use for: AI presenters, digital twins, talking photos, dubbing and localizing videos, re-syncing lips to new audio, custom voices, prompt-to-vi

Published by inference-sh ↗inference-sh/skills

Try your first task

A source-grounded starter task has not been verified for this skill yet.

SKILL.md

name: heygendescription: "HeyGen avatar videos, photo animation, lipsync, video translation, TTS and voice cloning via inference.sh CLI. Apps: heygen/avatar-video (Avatar IV/V engines, up to 4K), heygen/create-avatar (digital twin, photo or prompt avatars), heygen/photo-video, heygen/lipsync, heygen/video-agent (prompt to finished video), heygen/video-translate (30+ languages with voice cloning and lip-sync), heygen/text-to-speech (Starfish), heygen/voice-clone (instant clone). Use for: AI presenters, digital twins, talking photos, dubbing and localizing videos, re-syncing lips to new audio, custom voices, prompt-to-video explainers. Triggers: heygen, heygen avatar, heygen api, digital twin, avatar iv, avatar v, photo avatar, talking photo, video translate, video translation, translate video, dub video, heygen lipsync, lip sync, heygen tts, starfish tts, voice clone, instant voice clone, video agent, ai spokesperson, ai presenter"allowed-tools: Bash(belt *)

Install the belt CLI skill: npx skills add belt-sh/cli

HeyGen

Create avatar videos, animate photos, translate and lipsync videos, and generate or clone voices with HeyGen via inference.sh CLI.

Quick Start

Requires inference.sh CLI (belt). Install instructions

belt login

# Animate a portrait photo into a talking video
belt app run heygen/photo-video --input '{
  "image": "https://portrait.jpg",
  "audio_url": "https://speech.mp3"
}'

Apps

AppApp IDWhat It DoesPricing
Avatar Videoheygen/avatar-videoTalking video from a HeyGen avatar ID + script, Avatar IV or V engine, 720p/1080p/4KPhoto avatar $0.05/s (720p/1080p), $0.067/s (4K); digital twin $0.067/s (720p/1080p), $0.083/s (4K)
Create Avatarheygen/create-avatarNew avatar from video (digital twin), photo, or text prompt; returns a look ID$1.00/call
Photo Videoheygen/photo-videoAnimate a portrait photo with script+voice or an audio file$0.05/s (720p/1080p), $0.067/s (4K)
Lip Syncheygen/lipsyncRe-sync an existing video's lips to new audio$0.033/s speed, $0.067/s precision
Video Agentheygen/video-agentFull video from a natural language prompt; agent picks avatar, script, production$0.033/s
Video Translateheygen/video-translateTranslate a video into 30+ languages with voice cloning and lip-sync$0.033/s speed or audio-only, $0.067/s precision
Text to Speechheygen/text-to-speechStarfish TTS, plain text or SSML, speed and language control$0.0007/s
Voice Cloneheygen/voice-cloneInstant clone from one short recording, optionally speak text in it$0.0007/s of audio; clone-only runs free

Choosing an App

You haveYou wantUse
A portrait photoA one-off talking videoheygen/photo-video
Footage, a photo, or a descriptionA reusable avatarheygen/create-avatar, then heygen/avatar-video
A HeyGen avatar IDA scripted videoheygen/avatar-video
Only an ideaA finished videoheygen/video-agent
A video in one languageThe same video in anotherheygen/video-translate
A video and new audioLips matching the new audioheygen/lipsync
TextSpeech audioheygen/text-to-speech
A voice recordingA reusable voice IDheygen/voice-clone

Examples

Avatar Video

belt app run heygen/avatar-video --input '{
  "avatar_id": "<avatar-look-id>",
  "script": "Welcome to our quarterly update. Here are three things that changed.",
  "engine": "avatar_iv",
  "resolution": "1080p",
  "aspect_ratio": "16:9"
}'
  • engine: avatar_iv (default) or avatar_v. Avatar V gives more natural motion and lip-sync but needs an eligible avatar look.
  • resolution: 720p, 1080p, 4k. aspect_ratio: 16:9, 9:16.
  • voice_id is optional; the avatar's default voice is used when omitted.
  • expressiveness (high, medium, low) and motion_prompt only apply to photo avatars.
  • remove_background and output_format (mp4, webm) are also available.

List available avatars and voices with the list_resources function:

belt app run heygen/avatar-video --function list_resources --input '{
  "resource_type": "both",
  "limit": 10
}'

resource_type is avatars, voices, or both.

Photo Video (Talking Photo)

With a script and voice:

belt app run heygen/photo-video --input '{
  "image": "https://portrait.jpg",
  "script": "Hi, I am excited to show you what we built.",
  "voice_id": "<voice-id>",
  "expressiveness": "medium",
  "motion_prompt": "nodding slowly",
  "aspect_ratio": "9:16"
}'

voice_id is required when script is set. Pass audio_url instead of script + voice_id to lip-sync to your own audio.

Create an Avatar

# From a photo
belt app run heygen/create-avatar --input '{
  "type": "photo",
  "name": "Product Presenter",
  "file": "https://portrait.jpg"
}'

# From a text prompt
belt app run heygen/create-avatar --input '{
  "type": "prompt",
  "name": "Friendly Teacher",
  "prompt": "a friendly middle-aged teacher in a bright classroom, warm smile"
}'

# Digital twin from video footage
belt app run heygen/create-avatar --input '{
  "type": "digital_twin",
  "name": "My Twin",
  "file": "https://footage-of-me.mp4"
}'

The output has avatar.id (the look ID to pass as avatar_id in heygen/avatar-video), avatar.supported_engines, and group.id. Pass avatar_group_id to attach a new look to an existing character. Digital twins can return group.consent_status and group.consent_url when consent is pending.

Video Agent (Prompt to Video)

belt app run heygen/video-agent --input '{
  "prompt": "A 45-second explainer about how solar panels convert sunlight into electricity, friendly presenter, simple diagrams",
  "orientation": "landscape"
}'

orientation is landscape or portrait (auto-detected if omitted). Optional avatar_id, voice_id, and style_id pin specific choices.

Lip Sync

belt app run heygen/lipsync --input '{
  "video": "https://original.mp4",
  "audio": "https://new-voiceover.mp3",
  "mode": "precision"
}'

mode is speed (default) or precision. Optional: enable_caption, enable_dynamic_duration (default true), disable_music_track, start_time/end_time for a partial sync. Inputs must be publicly accessible URLs.

Text to Speech

belt app run heygen/text-to-speech --input '{
  "text": "Thanks for watching. See you next week.",
  "voice_id": "<voice-id>",
  "speed": 1.1,
  "locale": "en-US"
}'
  • text: 1-5000 characters. input_type: text or ssml.
  • speed: 0.5-2.0. language (ISO code, e.g. es) or locale (BCP-47, e.g. pt-BR).
  • The voice must support the Starfish engine.

Voice Clone

belt app run heygen/voice-clone --input '{
  "audio": "https://my-voice-60s.mp3",
  "text": "This sentence is spoken in my cloned voice.",
  "voice_name": "My Voice",
  "keep_voice": true
}'
  • 30-60 seconds of clean, single-speaker speech works best. remove_background_noise defaults to true.
  • keep_voice defaults to false: the clone is deleted after generating audio because the HeyGen account has limited clone slots. It is forced on when no text is given.
  • Output: audio, voice_id, kept, preview_audio_url. A kept voice_id works with heygen/text-to-speech and avatar videos.

Manage cloned voices:

belt app run heygen/voice-clone --function list_voices --input '{"voice_type": "private", "limit": 20}'
belt app run heygen/voice-clone --function delete_voice --input '{"voice_id": "<voice-id>"}'

voice_type: private (this workspace's clones) or public (HeyGen catalog).

Workflows

Create an Avatar, Then Make Videos With It

# 1. Create a photo avatar (one-time, $1.00)
belt app run heygen/create-avatar --input '{
  "type": "photo",
  "name": "Brand Host",
  "file": "https://host-portrait.jpg"
}' -o avatar.json

# 2. Use avatar.id from step 1 as avatar_id
belt app run heygen/avatar-video --input '{
  "avatar_id": "<avatar.id-from-step-1>",
  "script": "Here is what is new this week.",
  "aspect_ratio": "9:16",
  "resolution": "1080p"
}'

Reuse the same avatar_id for every future video. Check avatar.supported_engines before setting "engine": "avatar_v".

Translate a Video

belt app run heygen/video-translate --input '{
  "video": "https://original-english.mp4",
  "output_language": "Spanish (Spain)",
  "mode": "precision",
  "speaker_num": 2,
  "enable_caption": true
}'
  • output_language takes names like Spanish (Spain), Chinese (Mandarin, Simplified), French.
  • input_language is auto-detected if omitted.
  • translate_audio_only: true translates the audio track and keeps the original frames.
  • speaker_num improves speaker separation in multi-speaker videos.
  • start_time/end_time translate only part of the video.

Your Own Voice on an Avatar

# 1. Clone and keep the voice
belt app run heygen/voice-clone --input '{
  "audio": "https://my-voice-60s.mp3",
  "voice_name": "Founder Voice"
}' -o voice.json

# 2. Use voice_id from step 1 in an avatar video
belt app run heygen/avatar-video --input '{
  "avatar_id": "<avatar-look-id>",
  "voice_id": "<voice_id-from-step-1>",
  "script": "A quick message from our founder."
}'

With no text, the clone is kept automatically and only the voice_id is returned.

Replace the Voiceover on Existing Footage

# 1. Generate new speech
belt app run heygen/text-to-speech --input '{
  "text": "Our new pricing starts today.",
  "voice_id": "<voice-id>"
}' -o speech.json

# 2. Re-sync the presenter's lips to the new audio
belt app run heygen/lipsync --input '{
  "video": "https://presenter.mp4",
  "audio": "<audio-url-from-step-1>",
  "mode": "precision"
}'

Tips

  • Check pricing with belt app pricing heygen/<app>; use --estimate before long or 4K renders.
  • precision mode in lipsync and video-translate costs about twice speed mode.

Related Skills

# Full platform skill (all apps)
npx skills add inference-sh/skills@infsh-cli

# Other avatar and lipsync models (P-Video-Avatar, OmniHuman, Fabric)
npx skills add inference-sh/skills@ai-avatar-video

# Talking head production workflows
npx skills add inference-sh/skills@talking-head-production

# ElevenLabs dubbing (alternative to video-translate)
npx skills add inference-sh/skills@elevenlabs-dubbing

# Voice cloning with other providers
npx skills add inference-sh/skills@ai-voice-cloning

# Video generation
npx skills add inference-sh/skills@ai-video-generation

Browse all HeyGen apps: belt app search "heygen"

Documentation