# Create AI Avatar and Talking Head Videos

Avatar video production often requires separate speech, animation, and lip-sync tools. This skill organizes inference.sh models into reusable presenter, dubbing, and UGC workflows.

## Install

```bash
npx skillstore add 101-skills/ai-avatar-video
```

## Metadata

- Status: approved
- Slug: 101-skills-ai-avatar-video
- Skillstore revision: r1
- Version status: missing
- Tree hash: 70f78fe157b0ea8bc79439517f6d4422d95ed9acc70bbb35b084cb87013a36c1
- Author: 101-skills
- GitHub username: 101-skills
- License: MIT
- Repository: https://github.com/101-skills/skills/tree/main/tools/video/ai-avatar-video/
- Ref: d71c7417a35d5c2624161bd2fe8de8a41a362128
- Supported tools: Claude, Codex, Claude Code
- Audit status: complete
- Agent install advisory: confirmation\_required
- Manual install advisory: allowed
- Artifact signature: available
- Audit attestation: unavailable
- Human verification: not\_verified
- Risk factors: external\_commands, network
- Quality score: 38
- Quality tier: warning
- Public page: https://skillstore.pages.dev/skills/101-skills-ai-avatar-video
- Manifest: https://skillstore.pages.dev/api/skills/101-skills-ai-avatar-video/manifest

## Capabilities

- Generate talking-head videos from a portrait and written script with P-Video-Avatar.
- Animate portraits from existing audio with P-Video-Avatar, OmniHuman, Fabric, or PixVerse.
- Control voice, language, delivery style, scene direction, and output resolution.
- Combine portrait generation, text-to-speech, transcription, translation, and lip synchronization.
- Run batch presenter variants with different voices through shell workflows.

## Use Cases

- Produce product explainers: Turn an approved presenter portrait and concise script into a narrated product walkthrough.
- Localize training videos: Transcribe a training video, translate its narration, generate speech, and synchronize the new audio.
- Create character dialogue: Animate an authorized character image with expressive speech for game scenes or interactive prototypes.

## Prompt Templates

### Create a basic presenter

```
Create a 720p talking-head video from [portrait URL]. Use this script: [script]. Choose a clear, neutral English voice.
```

### Direct voice and scene

```
Create a 1080p presenter video from [portrait URL] and [script]. Use an enthusiastic voice and describe a professional studio setting.
```

### Build a dubbing workflow

```
Dub [video URL] into [language]. Transcribe, translate, generate approved speech, and synchronize it while preserving the original visual timing.
```

### Generate approved presenter variants

```
Create three disclosed synthetic presenter variants for [campaign]. Use authorized portraits, distinct voices, consistent messaging, and vertical 1080p output.
```

## Limitations

- The skill requires the belt CLI, an authenticated inference.sh account, and access to remote models.
- Model availability, pricing, speed, supported voices, and output quality can change.
- Media inputs must be available through URLs accepted by the selected model.
- The skill does not verify likeness rights, voice consent, advertising compliance, or synthetic-media disclosure.

## Best Practices

- Use front-facing, well-lit portraits and clean audio for stable facial animation.
- Confirm likeness, voice, script, and media rights before sending inputs to remote models.
- Review generated speech, lip synchronization, claims, and disclosure labels before publication.

## Anti Patterns

- Do not animate a real person without documented permission.
- Do not embed unescaped user text directly inside shell-quoted input arguments.
- Do not publish synthetic testimonials as genuine customer endorsements.

## Security Audit

- Audited at: 2026-07-12T09:45:58.132\+00:00
- Summary: Most detections are Markdown backticks, code fences, placeholder media URLs, or documentation links rather than executable backticks or suspicious destinations. Real risks remain in third-party package installation, remote media processing, shell command construction, remote image loading, and synthetic testimonial workflows without consent safeguards.

## Stats

- Views: 0
- Downloads: 6
- Favorites: 0
- Popularity score: 0