本文へ移動
cccskills
無料GitHub で公開

openclaw-plugin-heygen

Official HeyGen plugin for OpenClaw — adds HeyGen as a first-class provider for the built-in video_generate tool. Identity-first avatar videos: pick an avatar_id, a voice, a script, and get an on-brand presenter video back. Use when: (1) generating a video where a specific avatar/person speaks specific lines, (2) "make a video of me / our brand presenter", (3) outreach, announcements, pitch decks, product walkthroughs, or any presenter-led video, (4) needing landscape or portrait output with deterministic aspect ratios, (5) integrating HeyGen into an OpenClaw agent through the standard provider interface (no bespoke REST glue required). Covers: HeyGen v3 Video Agent (api.heygen.com), avatar-presenter video, identity-preserving video, talking-head AI video, scripted avatar narration, brand spokesperson video. NOT for: cinematic b-roll (use Veo / Runway / Kling), video translation, TTS-only, streaming avatars, or photo-to-talking-head v2 pipelines.

heygen-com/openclaw-plugin-heygen
インストール方法を見る

含まれるファイル（19）

  • SKILL.md4.4 KB
  • .github/workflows/publish.yml1.1 KB
  • .github/workflows/release-please.yml1.0 KB
  • .gitignore122 B
  • .release-please-manifest.json19 B
  • docs/heygen.md7.1 KB
  • index.ts402 B
  • INSTALL_FOR_AGENTS.md8.6 KB
  • LICENSE1.0 KB
  • openclaw.plugin.json1.1 KB
  • package-lock.json172.7 KB
  • package.json1.4 KB
  • plugin-registration.contract.test.ts249 B
  • README.md7.3 KB
  • release-please-config.json816 B
  • tsconfig.json464 B
  • video-generation-provider.test.ts24.6 KB
  • video-generation-provider.ts20.7 KB
  • vitest.config.ts1.2 KB

SKILL.md（原文）

インストールする前に、エージェントに与えられる指示の中身を確認できます。

HeyGen Video Agent — OpenClaw Plugin

Adds HeyGen as a first-class provider for OpenClaw's built-in video_generate tool. Same interface every other video provider in OpenClaw uses (Google Veo, Runway, Kling, Wan, MiniMax, Dashscope) — now with identity-preserving avatar/presenter video.

Install

openclaw plugin install heygen

Set HEYGEN_API_KEY in your environment (get one at https://app.heygen.com/settings/api).

Quick start

const result = await video_generate({
  model: "heygen/video_agent_v3",
  prompt: "Hi team, here's the product update for this week.",
  providerOptions: {
    avatar_id: "<your-avatar-id>",
    voice_id: "<your-voice-id>",
    orientation: "landscape", // or "portrait"
  },
});
// result.url → downloadable .mp4

Avatar / voice ids: list from the HeyGen dashboard or the heygen-skills companion skill (which composes opinionated avatar/voice picking on top of this plugin).

When HeyGen vs other video providers

  • Avatar-presenter / talking-head video → HeyGen.
  • Cinematic / scene generation → Veo, Runway, Kling.
  • Translation / dubbing → use HeyGen's translate APIs (separate skill).
  • Real-time interactive avatar → HeyGen Streaming Avatar (separate surface, not this plugin).

Provider options

FieldTypeNotes
avatar_idstringHeyGen avatar id; falls back to models.providers.heygen.defaultAvatarId.
voice_idstringHeyGen voice id; falls back to models.providers.heygen.defaultVoiceId.
style_idstringOptional style template.
orientation"landscape" | "portrait"Maps from aspect ratio. 1:1 rejected.
mode"generate" | "chat"Default generate. chat keeps full poll budget for interactive flows.
callback_urlstringAsync webhook completion.
callback_idstringCaller-supplied correlation id for the webhook.
incognito_modebooleanSkip retention on HeyGen side where supported.

Errors surfaced cleanly

  • 401 → HeyGen API key missing or invalid
  • 402 / quota → HeyGen credit limit reached
  • Generate-mode + 8 consecutive thinking polls → fast-fail with a hint pointing at avatar_id / voice_id.
  • failed status surfaces HeyGen's failure_message rather than a generic error.

Polling

Three-tier backoff: 5s × 6 → 15s × 12 → 30s until MAX_POLL_ATTEMPTS. Long renders (20–45 min) stay under typical HeyGen per-key rate limits.

Source + tests

License

MIT. See LICENSE.

レビュー

まだレビューはありません。使ってみた感想をお寄せください。

ログインしてレビューを書く
このスキルの問題を報告する