# Ai Video Generation Deepinfra

> A Claude skill from marketintellabs/agent-skills.

- **Type:** Skill
- **Install:** `agentstack add skill-marketintellabs-agent-skills-ai-video-generation-deepinfra`
- **Verified:** Yes — security-reviewed for prompt injection and unsafe behavior
- **Seller:** [marketintellabs](https://agentstack.voostack.com/s/marketintellabs)
- **Installs:** 0
- **Category:** [Content & Media](https://agentstack.voostack.com/c/content-and-media)
- **Latest version:** 0.1.0
- **License:** MIT
- **Upstream author:** [marketintellabs](https://github.com/marketintellabs)
- **Source:** https://github.com/marketintellabs/agent-skills/tree/main/ai-video-generation-deepinfra

## Install

```sh
agentstack add skill-marketintellabs-agent-skills-ai-video-generation-deepinfra
```

Requires the [AgentStack CLI](https://agentstack.voostack.com/docs/cli). Works with Claude Code, Cursor, and any MCP-compatible agent.

## About

# AI Video Generation (DeepInfra)

Generate short-form video clips using DeepInfra's text-to-video models. Ideal for intros, teasers, animated data visualizations, and branded bumpers. Uses your existing DeepInfra API key — no additional subscriptions required.

## Requirements

- A DeepInfra API key (set as `DEEPINFRA_API_KEY` environment variable)
- Any HTTP client

## API Reference

**Endpoint:** `https://api.deepinfra.com/v1/inference/{model}`
**Auth:** `Authorization: Bearer $DEEPINFRA_API_KEY`

Check DeepInfra's current text-to-video model catalog for available models and pricing: https://deepinfra.com/models/text-to-video

### Request

```json
{
  "prompt": "Detailed video description",
  "num_frames": 49,
  "width": 1024,
  "height": 576,
  "seed": -1
}
```

### Parameters

| Parameter | Description |
|-----------|-------------|
| `prompt` | Detailed description of the video scene, motion, and style |
| `num_frames` | Number of frames to generate (more frames = longer clip, higher cost) |
| `width` | Output width in pixels |
| `height` | Output height in pixels |
| `seed` | Random seed for reproducibility (-1 for random) |

## Use Cases

| Type | Duration | Purpose |
|------|----------|---------|
| Intro animation | 3–5 sec | Opening sequence for video content |
| Content teaser | 5–10 sec | Social media previews for articles or reports |
| Data visualization | 3–5 sec | Animated chart/data sequences |
| Channel bumper | 2–3 sec | Branded intro/outro clips |

## Prompt Engineering

Structure prompts as: `{scene_description}, {motion_direction}, {style}, {lighting}, {duration_hint}`

```
Example for an intro:
"Smooth camera pan across a dark environment with glowing holographic
charts floating in space, cinematic lighting, professional broadcast
quality, slow deliberate motion"

Example for a data visualization:
"Abstract particles flowing upward in a spiral pattern against a dark
background, representing rising data, smooth slow motion, minimalist
modern design"
```

Tips:
- Describe motion explicitly (pan, zoom, flow direction)
- Keep prompts focused on a single clear scene
- Shorter clips (3–5 sec / fewer frames) are cheaper and more reliable
- Avoid requesting realistic human faces (uncanny valley risk)
- Video is visual only — audio/narration must be added separately

## Cost Control

- Video generation costs significantly more than image generation per request
- Prefer shorter clips (fewer frames) over longer ones
- Reuse evergreen clips (intros, bumpers) — generate once, use many times
- Set budget guardrails appropriate to your use case
- Test prompts with lower frame counts before generating full-quality clips

## Workflow

1. Define the video concept and purpose
2. Craft a detailed prompt with motion, style, and lighting
3. Generate a test clip with lower frame count to validate the prompt
4. Generate the final clip at desired resolution and frame count
5. Download the output and upload to your storage backend
6. Combine with audio/narration in post-production if needed

## Example (Python)

```python
import requests, os

model = "your-chosen-text-to-video-model"
resp = requests.post(
    f"https://api.deepinfra.com/v1/inference/{model}",
    headers={"Authorization": f"Bearer {os.environ['DEEPINFRA_API_KEY']}"},
    json={
        "prompt": "Abstract blue and gold particles flowing upward against a dark background, smooth slow motion, minimalist modern design",
        "num_frames": 49,
        "width": 1024,
        "height": 576,
        "seed": -1,
    },
)

video_url = resp.json().get("video_url") or resp.json().get("output")
print(f"Video: {video_url}")
```

## Example (curl)

```bash
curl -X POST "https://api.deepinfra.com/v1/inference/YOUR_MODEL" \
  -H "Authorization: Bearer $DEEPINFRA_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "prompt": "Smooth cinematic camera pan across abstract data visualization",
    "num_frames": 49,
    "width": 1024,
    "height": 576,
    "seed": -1
  }'
```

## Source & license

This open-source skill is cataloged on AgentStack and links to its original source — we do not rehost the code.

- **Author:** [marketintellabs](https://github.com/marketintellabs)
- **Source:** [marketintellabs/agent-skills](https://github.com/marketintellabs/agent-skills)
- **License:** MIT

Install and usage instructions live in the source repository linked above.

## Pricing

- **Free** — Free

## Security capabilities

Automated source analysis of v0.1.0 — what this tool can access:

- **Network access:** yes
- **Filesystem access:** no
- **Shell / process execution:** no
- **Environment & secrets:** yes
- **Dynamic code execution:** no

*"Yes" means the capability is present in the source — more access means more to trust, not that it is unsafe.*


## Versions

- **0.1.0** — security scan: passed — Imported from the upstream source.

## Links

- Listing page: https://agentstack.voostack.com/l/skill-marketintellabs-agent-skills-ai-video-generation-deepinfra
- Seller: https://agentstack.voostack.com/s/marketintellabs
- Browse the marketplace: https://agentstack.voostack.com/browse

---
Listed on AgentStack — the marketplace for AI agent skills and MCP servers. Every listing is security-reviewed. Creators keep 70%.