# Create AI Avatar Videos

Producing talking head videos often requires editing tools, voice generation, and model selection. This skill gives Claude, Codex, and Claude Code CLI workflows for avatar video generation.

## Install

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

## Metadata

- Status: approved
- Slug: inference-skills-ai-avatar-video
- Skillstore revision: r1
- Version status: missing
- Tree hash: dc92e9e6c7af1f7d9f6eb816b5eccd4ba14f4b23e2905e1929852f4de9b04307
- Author: inference-skills
- GitHub username: inference-skills
- License: MIT
- Repository: https://github.com/inference-skills/skills/tree/main/tools/video/ai-avatar-video/
- Ref: a25199bc7d6b82598536822d1738eb5d5f54025b
- Supported tools: Claude, Codex, Claude Code
- Audit status: complete
- Agent install advisory: allowed
- Manual install advisory: allowed
- Artifact signature: available
- Audit attestation: unavailable
- Human verification: not\_verified
- Risk factors: external\_commands, network
- Quality score: 69
- Public page: https://skillstore.pages.dev/skills/inference-skills-ai-avatar-video
- Manifest: https://skillstore.pages.dev/api/skills/inference-skills-ai-avatar-video/manifest

## Capabilities

- Runs P-Video-Avatar examples through the inference.sh belt CLI.
- Shows model options for P-Video-Avatar, OmniHuman, Fabric, and PixVerse Lipsync.
- Supports text-to-avatar workflows with built-in text to speech.
- Supports avatar videos from portrait and audio URL inputs.
- Outlines portrait generation followed by avatar video creation.
- Outlines dubbing workflows using transcription, translation, speech generation, and lip sync.

## Use Cases

- Product Presenter Videos: Create short avatar-led product demos from a portrait and approved script.
- Training Explainers: Generate consistent talking head lessons for onboarding, support, or course content.
- Localized Video Dubbing: Plan a workflow to transcribe, translate, voice, and lip-sync existing videos.

## Prompt Templates

### Create First Avatar Video

```
Use this skill to create a 20-second presenter video from my portrait URL and short script. Choose the recommended model and explain expected cost factors.
```

### Use Custom Voice Audio

```
Create a talking head video from my portrait URL and approved audio URL. Select the best model for lip sync quality.
```

### Build Dubbing Workflow

```
Plan a dubbing workflow for my existing video. Include transcription, translation review, speech generation, and final lip sync steps.
```

### Compare Production Options

```
Compare the available avatar models for a campaign with ten short videos. Recommend settings for speed, cost, quality, and review checkpoints.
```

## Limitations

- Requires an inference.sh account, belt CLI access, and network connectivity.
- Model availability, pricing, speed, and output quality can change outside the skill.
- Users must provide media URLs and confirm they have rights to each asset.
- The skill does not verify consent, likeness rights, copyright, or disclosure requirements.

## Best Practices

- Use only portraits, voices, and scripts that are approved for synthetic video generation.
- Start with low-cost settings, then raise resolution after the script and timing are final.
- Keep scripts concise and review generated videos for lip sync, tone, and disclosure needs.

## Anti Patterns

- Do not create avatars of real people without permission and a clear review process.
- Do not include confidential, regulated, or customer-sensitive information in scripts or media URLs.
- Do not treat generated videos as final without checking likeness accuracy and rights.

## Security Audit

- Audited at: 2026-07-05T18:50:43.606\+00:00
- Summary: The static backtick findings are false positives from Markdown code fences, inline model names, and example CLI commands, not executable Ruby or shell evaluation. The hardcoded URLs are documentation links, image assets, or placeholder media inputs for the intended inference workflow. One semantic marketplace risk remains: the synthetic media workflows need explicit consent, rights, and disclosure guidance.

## Stats

- Views: 81
- Downloads: 10
- Favorites: 0
- Popularity score: 0