菜单
Create HeyGen image-to-video/photo-avatar talking-head videos from a person's image plus custom audio. Use when the user wants a personal brand video, video nhân hiệu, talking avatar, lip-synced photo video, or a HeyGen video generated from one image and one audio file.
用 Codex 或 Claude 帮你安装 复制这段 Prompt，粘贴到 Codex、Claude 或其他助手里，让它检查 Skill 页面并帮你完成安装。
基于 SOC 职业分类
Download public Facebook Reel videos with yt-dlp into the repository raw asset pool or a specific video job. Use when a user provides a facebook.com/reel, facebook.com/watch, Facebook video, or fb.watch URL and asks to download, save, or ingest the video as a local asset.
Search LyricFind for song lyrics and lyric context, especially to correct Vietnamese music STT before writing subtitle captions; use the bundled Python client instead of generic web search when a song title or artist is known.
Default low-cost lipsync workflow using WaveSpeedAI InfiniteTalk. Use when the user asks for lipsync, lip sync, talking photo/avatar, digital human from one image plus audio, cheap avatar video generation, or asks to animate a person speaking from an image and audio unless they explicitly choose another provider.
Create AI videos with KIE.AI Google Veo 3.1. Use when the user wants text-to-video, image-to-video with first/last frames, reference/material-to-video, extending Veo 3.1 tasks, polling status, or upgrading/downloading 1080p/4K results through the KIE Veo3 API.
Convert a finished video job into a reusable project template and Codex skill. Use when the user asks to turn a Remotion job into a template, standardize a copied job, create a reusable video template skill, parameterize a job, or make future video requests token-light while preserving visual quality and render reliability.
Prepare effect/overlay videos so they can be composited over another video, especially stock overlays with black backgrounds, dust, light particles, petals, smoke, flares, or non-alpha MP4/MOV assets. Use when Codex needs to check whether overlay videos have real alpha, convert ordinary overlay footage into Remotion-ready assets, normalize fps/size, preserve color for screen blending, avoid black halos from bad alpha keying, or generate reusable overlay files and compositing snippets.
| name | heygen-photo-avatar-video |
| description | Create HeyGen image-to-video/photo-avatar talking-head videos from a person's image plus custom audio. Use when the user wants a personal brand video, video nhân hiệu, talking avatar, lip-synced photo video, or a HeyGen video generated from one image and one audio file. |
video nhân hiệu) from an image and an audio file.POST /v3/videos with type: "image"). It does not create a reusable avatar_id; for repeated videos with the same person, note that HeyGen's reusable Photo Avatar flow may be better..env from the repo root before running scripts. Require HEYGEN_API_KEY; never print the key.POST /v3/assets, then referenced as asset_id.audio_asset_id or audio_url; do not mix custom audio with script/voice_id.GET /v3/videos/{video_id} until completed or failed. Download video_url immediately when available.Create from uploaded image and uploaded audio:
{
"type": "image",
"image": {
"type": "asset_id",
"asset_id": "asset_image..."
},
"audio_asset_id": "asset_audio...",
"title": "Personal brand video",
"resolution": "1080p",
"aspect_ratio": "9:16"
}
Create from public URLs:
{
"type": "image",
"image": {
"type": "url",
"url": "https://example.com/person.jpg"
},
"audio_url": "https://example.com/narration.mp3",
"title": "Personal brand video"
}
Statuses: pending, processing, completed, failed. A completed response includes video_url.
Generate, poll, and download in one command:
python skills/heygen-photo-avatar-video/scripts/heygen_photo_avatar_video.py generate \
--env-file .env \
--image jobs/<job_id>/input/raw_assets/images/person.jpg \
--audio jobs/<job_id>/input/raw_assets/audio/narration.mp3 \
--title "Video nhân hiệu" \
--aspect-ratio 9:16 \
--resolution 1080p \
--output-dir jobs/<job_id>/input/raw_assets/videos/heygen
Use already-uploaded asset IDs:
python skills/heygen-photo-avatar-video/scripts/heygen_photo_avatar_video.py generate \
--env-file .env \
--image-asset-id asset_image_123 \
--audio-asset-id asset_audio_456 \
--title "Video nhân hiệu" \
--output-dir jobs/<job_id>/input/raw_assets/videos/heygen
Use public URLs:
python skills/heygen-photo-avatar-video/scripts/heygen_photo_avatar_video.py generate \
--env-file .env \
--image-url https://example.com/person.jpg \
--audio-url https://example.com/narration.mp3 \
--title "Video nhân hiệu" \
--output-dir jobs/<job_id>/input/raw_assets/videos/heygen
Check a video later:
python skills/heygen-photo-avatar-video/scripts/heygen_photo_avatar_video.py status \
--env-file .env \
--video-id <video_id> \
--output-dir jobs/<job_id>/input/raw_assets/videos/heygen \
--download
The script writes:
<output-dir>/heygen_video_<video_id>.json
<output-dir>/heygen_video_<video_id>.mp4
Use downloaded videos as raw assets. After download, continue with the usual pipeline:
.venv/bin/python -m tools.asset_index.exporter jobs/<job_id>/input/raw_assets/ --output jobs/<job_id>/source/asset_semantics.toml
--aspect-ratio 9:16 --resolution 1080p.--aspect-ratio 16:9 --resolution 1080p.--remove-background and optionally --background-color "#ffffff".Missing HEYGEN_API_KEY: add it to .env or export it in the shell.script and audio are mutually exclusive: this skill is for custom audio; use HeyGen script/voice only in a separate flow.failed status: inspect failure_message in the saved JSON report.401: invalid or inactive HeyGen API key.video_url field path and update the script if HeyGen changed the response shape.