Runware Image & Video generation — 技能

技能详情（站内镜像，无评论）

Generate images and videos via Runware API. Access to FLUX, Stable Diffusion, Kling AI, and other top models. Supports text-to-image, image-to-image, upscaling, text-to-video, and image-to-video. Use when generating images, creating videos from prompts or images, upscaling ima…

媒体与内容

许可证：MIT-0

MIT-0 ·免费使用、修改和重新分发。无需归因。

版本：v1.0.0

统计：⭐ 1 · 1.9k · 6 current installs · 6 all-time installs

⭐ 1

安装量（当前） 6

🛡 VirusTotal ：良性 · OpenClaw ：可疑

Package：26medias/runware

安全扫描（ClawHub）

VirusTotal ：良性
OpenClaw ：可疑

OpenClaw 评估

The skill appears to implement the claimed image/video functionality, but its metadata omits the required API credential and it will upload local files (base64) to a remote service — the combination of metadata mismatch, unknown source/homepage, and automatic upload of user files is a risk the user should understand before installing.

目的

Name, description, SKILL.md, and included scripts all align: they call https://api.runware.ai/v1 to perform text->image, img2img, upscaling, text->video and img->video. The required functionality (network calls, uploading images, downloading results) is coherent with the stated purpose.

说明范围

Runtime instructions and the scripts are narrowly scoped to interacting with the Runware API and reading/writing image/video files. There are no instructions to read unrelated system config, shell history, or to contact third-party endpoints outside runware.ai for control.

安装机制

No install spec — code files are shipped with the skill and there are no external downloads or package installs. This minimizes install-time risk, though the included scripts will make outbound network calls at runtime.

证书

SKILL.md and both scripts require a Runware API key (RUNWARE_API_KEY or --api-key) but the registry metadata lists no required environment variables or primary credential — this is an inconsistency. Additionally, the scripts will read local image files and embed them as base64 in API requests (i.e., they upload the raw image content), which is expected for img2img/upscale but has privacy implications that the metadata does not call out.

持久

The skill does not request permanent presence (always:false) and does not modify other skills or global agent config. It runs as-needed and only performs network I/O and local file read/write under user-invoked commands.

安装（复制给龙虾 AI）

将下方整段复制到龙虾中文库对话中，由龙虾按 SKILL.md 完成安装。

请把本段交给龙虾中文库（龙虾 AI）执行：为本机安装 OpenClaw 技能「Runware Image & Video generation」。简介：Generate images and videos via Runware API. Access to FLUX, Stable Diffusion, K…。
请 fetch 以下地址读取 SKILL.md 并按文档完成安装：https://raw.githubusercontent.com/openclaw/skills/refs/heads/main/skills/26medias/runware/SKILL.md
（来源：yingzhi8.cn 技能库）

SKILL.md

打开原始 SKILL.md（GitHub raw）

---
name: runware
description: Generate images and videos via Runware API. Access to FLUX, Stable Diffusion, Kling AI, and other top models. Supports text-to-image, image-to-image, upscaling, text-to-video, and image-to-video. Use when generating images, creating videos from prompts or images, upscaling images, or doing AI image transformation.
---

# Runware

Image and video generation via Runware's unified API. Access FLUX, Stable Diffusion XL, Kling AI, and more.

## Setup

Set `RUNWARE_API_KEY` environment variable, or pass `--api-key` to scripts.

Get API key: https://runware.ai

## Image Generation

### Text-to-Image

```bash
python3 scripts/image.py gen "a cyberpunk city at sunset, neon lights, rain" --count 2 -o ./images
```

Options:
- `--model`: Model ID (default: `runware:101@1` / FLUX.1 Dev)
- `--width/--height`: Dimensions (default: 1024x1024)
- `--steps`: Inference steps (default: 25)
- `--cfg`: CFG scale (default: 7.5)
- `--count/-n`: Number of images
- `--negative`: Negative prompt
- `--seed`: Reproducible seed
- `--lora`: LoRA model ID
- `--format`: png/jpg/webp

### Image-to-Image

Transform an existing image:

```bash
python3 scripts/image.py img2img ./photo.jpg "watercolor painting style" --strength 0.7
```

- `--strength`: How much to transform (0=keep original, 1=ignore original)

### Upscale

```bash
python3 scripts/image.py upscale ./small.png --factor 4 -o ./large.png
```

### List Models

```bash
python3 scripts/image.py models
```

## Video Generation

### Text-to-Video

```bash
python3 scripts/video.py gen "a cat playing with yarn, cute, high quality" --duration 5 -o ./cat.mp4
```

Options:
- `--model`: Model ID (default: `klingai:5@3` / Kling AI 1.6 Pro)
- `--duration`: Length in seconds
- `--width/--height`: Resolution (default: 1920x1080)
- `--negative`: Negative prompt
- `--format`: mp4/webm/mov
- `--max-wait`: Polling timeout (default: 600s)

### Image-to-Video

Animate an image or interpolate between frames:

```bash
# Single image (becomes first frame)
python3 scripts/video.py img2vid ./start.png --prompt "zoom out slowly" -o ./animated.mp4

# Two images (first and last frame)
python3 scripts/video.py img2vid ./start.png ./end.png --duration 5
```

### List Video Models

```bash
python3 scripts/video.py models
```

## Popular Models

### Image
| Model | ID |
|-------|-----|
| FLUX.1 Dev | `runware:101@1` |
| FLUX.1 Schnell (fast) | `runware:100@1` |
| FLUX.1 Kontext | `runware:106@1` |
| Stable Diffusion XL | `civitai:101055@128080` |
| RealVisXL | `civitai:139562@297320` |

### Video
| Model | ID |
|-------|-----|
| Kling AI 1.6 Pro | `klingai:5@3` |
| Kling AI 1.5 Pro | `klingai:3@2` |
| Runway Gen-3 | `runwayml:1@1` |

Browse all: https://runware.ai/models

## Notes

- Video generation is async; scripts poll until complete
- Costs vary by model — check https://runware.ai/pricing
- FLUX models are excellent for quality; Schnell is faster
- For best video results, use descriptive prompts with motion words