Compétences ai-avatar-video
🎥

ai-avatar-video

Sûr 🌐 Accès réseau

AI 태킹 헤드 비디오 생성

텍스트, 오디오 또는 이미지를 사용하여 AI 아바타로 전문적인 태킹 헤드 비디오를 만드세요. 촬영 없이 제품 데모, 튜토리얼 및 가상 프레젠테이터에 완벽합니다.

Prend en charge: Claude Codex Code(CC)
📊 70 Adéquat
1

Télécharger le ZIP du skill

2

Importer dans Claude

Allez dans Paramètres → Capacités → Skills → Importer un skill

3

Activez et commencez à utiliser

Tester

Utilisation de "ai-avatar-video". Generate a talking head video from a portrait photo with a voice-over explaining a product feature.

Résultat attendu:

Uses infsh app run bytedance/omnihuman-1-5 --input '{"image_url": "https://portrait.jpg", "audio_url": "https://speech.mp3"}' to create the video. Result: The AI avatar from portrait.jpg speaks the audio track with synchronized lip movements.

Utilisation de "ai-avatar-video". Create a dubbed version of an educational video in a new language.

Résultat attendu:

1. Transcribe original video with infsh/fast-whisper-large-v3. 2. Generate speech in new language with kokoro-tts. 3. Sync to original video with latentsync-1-6. Result: Original video with AI avatars speaking the new language.

Audit de sécurité

Sûr
v1 • 2/4/2026

Documentation file containing example CLI commands. Static scanner misidentified markdown code blocks as code execution patterns. No actual security vulnerabilities found.

1
Fichiers analysés
154
Lignes analysées
1
résultats
1
Total des audits
Audité par: claude

Score de qualité

38
Architecture
100
Maintenabilité
87
Contenu
21
Communauté
100
Sécurité
91
Conformité aux spécifications

Ce que vous pouvez construire

제품 데모 비디오

촬영 없이 AI 프레젠테이터로 전문 제품 데모를 만드세요. 기능을 소개하는 SaaS 회사에 이상적입니다.

교육 콘텐츠

일관된 설명 비디오와 튜토리얼을 생성하세요. 온라인 강좌 및 교육 플랫폼에 완벽합니다.

다국어 콘텐츠

AI 아바타가 새로운 언어로 말하도록 기존 비디오를 여러 언어로 더빙하세요. 글로벌 청중을 확장하세요.

Essayez ces prompts

기본 아바타 비디오 생성
Generate a talking head video from this image: [IMAGE_URL] using the specified audio: [AUDIO_URL]. Use the OmniHuman 1.5 model for best quality.
립싱크 비디오 생성
Create a lipsync video where this static image speaks the provided audio. Use the PixVerse Lipsync model for realistic lip movements.
다국어 더빙 생성
Create a dubbed version of this video in English: [VIDEO_URL] using these translated subtitles as audio: [TRANSLATED_AUDIO_URL]. Use LatentSync for video-to-audio sync.
TTS + 아바타 워크플로우 생성
First generate speech from this text: [TEXT] using kokoro-tts. Then create a talking head video from this image: [IMAGE_URL] with the generated speech as audio.

Bonnes pratiques

  • Use high-quality front-facing portrait photos with good lighting for best avatar results
  • Ensure audio is clear with minimal background noise for accurate lipsync
  • Use Claude Code's multi-step workflow for complex video generation tasks
  • Verify URL accessibility before using images or audio in commands
  • Test with small segments before generating full-length videos

Éviter

  • Don't use low-quality images with poor lighting - avatar quality will be poor
  • Avoid using copyrighted images for commercial avatar videos without permission
  • Don't mix languages in the same video without proper dubbing workflow
  • Don't forget to save output files immediately after generation
  • Avoid using very long audio tracks without testing lip sync quality

Foire aux questions

What is inference.sh and why do I need it?
Inference.sh is a cloud platform that provides AI model APIs. The inference.sh CLI allows you to run these AI models locally without managing infrastructure. You need it to access OmniHuman, Fabric, and PixVerse models for avatar video generation.
Which model should I use for my avatar videos?
Use OmniHuman 1.5 for multi-character videos and best quality. Use Fabric 1.0 for static images that need to speak. Use PixVerse Lipsync for highly realistic lip sync on existing images.
Can I create videos in languages other than English?
Yes, you can create videos in any language supported by the kokoro-tts text-to-speech model. Use the dubbing workflow to create multilingual versions of existing videos.
How do I authenticate with inference.sh?
Run infsh login in your terminal. You'll need to authenticate with your inference.sh account credentials. This provides access to all available AI models.
What image formats work best for avatar videos?
Use high-resolution portrait photos (400x400 or larger) with front-facing orientation. JPG and PNG formats work well. Good lighting and neutral background produce better results.
Can I use this skill with Claude Code?
Yes, this skill is compatible with Claude, Codex, and Claude Code. You can use the command examples in SKILL.md with Claude Code to build automated video generation workflows.

Détails du développeur

Structure de fichiers

📄 SKILL.md