# Animate Images with RunComfy Model Routing

Different animation tasks need different models and input formats. This skill selects a RunComfy route and supplies model-specific prompting guidance and CLI examples.

## Install

```bash
npx skillstore add gencraft-labs/image-to-video
```

## Metadata

- Status: approved
- Slug: gencraft-labs-image-to-video
- Skillstore revision: r1
- Version status: missing
- Tree hash: 9c50b2b32907fb02cb6ae3b02f8a7328ca6eaf4054b376100afa5591b5492007
- Author: gencraft-labs
- GitHub username: gencraft-labs
- License: MIT
- Repository: https://github.com/gencraft-labs/skills/tree/5d625f59c710d7a33316b50b302b7158cdee5971/image-to-video
- Ref: e406bf64b5db6e514879399e87b346fa7ff7c2fb
- 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, filesystem
- Quality score: 77
- Quality tier: bronze
- Public page: https://skillstore.pages.dev/skills/gencraft-labs-image-to-video
- Manifest: https://skillstore.pages.dev/api/skills/gencraft-labs-image-to-video/manifest

## Capabilities

- Route portrait, product, and general still-image animation requests to HappyHorse 1.0 I2V.
- Route custom voiceover requests to Wan 2.7 text-to-video for generated talking-head lip-sync.
- Route combined image, video, and audio reference requests to Seedance 2.0 Pro.
- Provide model-specific input fields, duration limits, resolution options, and reference-media requirements.
- Guide motion-focused prompting, preservation goals, negative prompts, and numbered reference cues.
- Provide RunComfy CLI invocation examples, output-directory configuration, and exit-code guidance.

## Use Cases

- Product Motion Previews: Prepare a short HappyHorse animation from a product image, with one camera movement and explicit packaging preservation goals.
- Localized Talking-Head Clips: Use Wan with different approved voiceover tracks to generate language variants of a described talking-head shot.
- Reference-Based Scene Concepts: Use Seedance to combine a subject image, a scene reference video, and an audio reference in one short clip.

## Prompt Templates

### Animate a Portrait

```
Prepare a five-second animation from [image URL], with subtle breathing and a gentle camera drift. Preserve facial features. Show the plan before submitting.
```

### Reveal a Product

```
Prepare a HappyHorse animation from [product image URL] at 1080P for eight seconds. Use a slow orbit and preserve packaging. Confirm before submission.
```

### Generate Voiceover Variants

```
Plan Wan talking-head clips using [English audio URL] and [Spanish audio URL]. Use vertical framing and a shared seed. Explain source-image limitations before submission.
```

### Combine Multimedia References

```
Plan an eight-second Seedance clip using [subject image URL], [scene video URL], and [voice audio URL]. Number reference cues and validate media before submission.
```

## Limitations

- Requires an installed RunComfy CLI, an authenticated account, accessible media URLs, and remote service availability; generation is not local.
- Wan's documented route has no source-image input and does not demonstrate animating an existing portrait for custom voiceover lip-sync.
- All documented routes cap output at 15 seconds; HappyHorse preserves input aspect ratio, while this Seedance template caps resolution at 720p.
- Selects one model per call; cross-model blending and stitching are outside scope, and output quality or identity consistency is not guaranteed.

## Best Practices

- Use media you are authorized to share, and confirm provider processing and current charges before remote submission.
- Validate model-specific media limits, duration, resolution, and aspect ratio before invoking the CLI.
- Describe one primary motion, state preservation goals, and reuse seeds when comparing supported variants.

## Anti Patterns

- Do not upload confidential media or place credentials in prompts or public media URLs.
- Do not assume the documented Wan route preserves a supplied portrait; it has no source-image field.
- Do not combine incompatible reference styles or request multiple model routes within a single generation call.

## Security Audit

- Audited at: 2026-10-07T15:00:23.694\+00:00
- Summary: All 67 static findings are false positives involving Markdown, placeholder URLs, ordinary CLI usage, or documented credential storage. No evidence found of malicious intent, unauthorized credential access, or prompt injection in SKILL.md. External CLI behavior and provider security claims cannot be verified from this documentation-only skill.

## Stats

- Views: 0
- Downloads: 1
- Favorites: 0
- Popularity score: 0