# Create AI Avatar Videos with Inference.sh

Teams need presenter videos without full video production. This skill guides avatar, lipsync, and dubbing workflows through the inference.sh CLI.

## Install

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

## Metadata

- Status: approved
- Slug: inference-sh-ai-avatar-video
- Skillstore revision: r1
- Version status: missing
- Tree hash: d8f0504bbaab6763b7d740e7830a58fe01ec6eeb0c454ca3618f24d4fc2a55cd
- Author: inference-sh
- GitHub username: inference-sh
- License: MIT
- Repository: https://github.com/inference-sh/skills/tree/main/skills/ai-avatar-video/
- Ref: a06681402992ceae98ba04d54cfd4ab004862696
- Supported tools: Claude, Codex, Claude Code
- Audit status: complete
- Agent install advisory: blocked
- Manual install advisory: allowed\_with\_warning
- 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/inference-sh-ai-avatar-video
- Manifest: https://skillstore.pages.dev/api/skills/inference-sh-ai-avatar-video/manifest

## Capabilities

- Shows how to install and log in to the inference.sh CLI.
- Lists OmniHuman, Fabric, PixVerse, and LatentSync choices for avatar video tasks.
- Provides CLI examples for image-and-audio avatar generation.
- Outlines a text-to-speech to avatar video workflow.
- Outlines a transcription, translation, and lipsync workflow for dubbing.
- Links to inference.sh documentation for app runs and streaming results.

## Use Cases

- Produce Presenter Demos: Create product demo videos from a presenter image and approved narration audio.
- Create Course Explainers: Turn lessons into talking head videos for online training and educational content.
- Localize Existing Videos: Plan transcription, translation, speech generation, and lipsync steps for multilingual releases.

## Prompt Templates

### Create a Simple Avatar

```
Use this skill to create an avatar video from my portrait URL and speech audio URL. Recommend the best model and command.
```

### Compare Model Options

```
Compare OmniHuman, Fabric, and PixVerse for my talking head video. Explain which one fits my image, audio, and realism goals.
```

### Plan a Dubbing Workflow

```
Help me dub an existing video into another language. Outline the inference.sh steps for transcription, translated speech, and lipsync.
```

### Build a Production Pipeline

```
Design a repeatable avatar video pipeline for multiple scripts. Include model selection, media preparation, review checkpoints, and final delivery steps.
```

## Limitations

- Requires an inference.sh account, CLI access, and network connectivity.
- Sends media URLs and generation requests to external inference.sh apps.
- Output quality depends on source image, audio quality, and model limits.
- Does not manage consent, likeness rights, or publication approvals.

## Best Practices

- Use only images, voices, and scripts that you have permission to use.
- Start with front-facing images and clean narration audio for stronger results.
- Review generated videos for identity, pronunciation, and brand accuracy before publishing.

## Anti Patterns

- Do not create videos that impersonate real people without clear consent.
- Do not publish generated avatars without checking rights, disclosures, and local rules.
- Do not use low quality or noisy source media for production results.

## Security Audit

- Audited at: 2026-07-06T15:38:57.557\+00:00
- Summary: Most external command findings are false positives caused by Markdown code fences and inline model IDs, not Ruby backtick execution. The remote installer piped to sh is a confirmed critical risk, and the related network URL is confirmed high risk. Other URLs are documentation links or placeholder media examples with no evidence of secret exfiltration.

## Stats

- Views: 171
- Downloads: 9
- Favorites: 0
- Popularity score: 0