图像安装

Name: 图像安装
Author: xramjetx

AI 智能社区

作者仓库星标 0
作者更新于实时读取
作者仓库 free-image-and-video-generation-skill

流狐档案事实层信息：领域、Agent、信任分、运行环境、权限与网络

领域: AI 智能
兼容 Agent: Claude Code
Cursor
Cline
Codex
Windsurf
Gemini CLI

+20
信任分: 83 / 100 · 社区维护
作者 / 版本 / 许可: @xramjetx · 未声明 license
Token 消耗评级: 低消耗
接入复杂程度: 需简单配置
是否需要外部 API Key: 需要 · Vendor-specific
兼容的系统: 未声明（默认跨平台）
底层运行要求: Python >=3.10
文件与系统权限: 只读
允许写入 / 修改
Shell 执行
读取环境变量
网络行为: 允许外网请求
安装命令数: 26 条

档案由构建时根据 SKILL.md 与安装命令自动衍生，可能与作者实际意图存在差异。

需要注意： 未限定 allowed-tools，默认拥有全部工具权限。；检出高风险片段：pipe_curl_to_shell

输出预览 free-image-and-video-generation.preview

---
name: free-image-and-video-generation
description: Free local AI image and video processing toolkit with cloud AI generation. Local tools: upscale…
category: AI 智能
runtime: Python
---

# free-image-and-video-generation 输出预览

## PART A: 任务判断
- 适用问题：提示词、Agent 工作流、模型评估或自动化推理。
- 输入要求：目标材料、限制条件、期望输出和验收方式。
- 证据边界：围绕“Prerequisites / Available Tools / Usage”读取原文规则，不把推断写成作者承诺。

## PART B: 执行结果
- **01** 任务判断：确认你的需求是否属于提示词、Agent 工作流、模型评估或自动化推理，并标出输入、限制和预期结果。
- **02** 执行计划：优先按“Prerequisites / Available Tools / Usage”拆成步骤，说明每一步会读取什么、修改什么、产出什么。
- **03** 交付结果：给出可复制的命令、文件改动、检查清单或内容草稿，并说明如何继续迭代。
- **04** 风险边界：结合 读取文件、写入/修改文件、执行终端命令、读取环境变量、会按任务需要访问外部网络、需要准备 Vendor-specific API Key 给出执行前确认项。

## Running Rules
- 读取文件、写入/修改文件、执行终端命令、读取环境变量；会按任务需要访问外部网络；需要准备 Vendor-specific API Key。
- 先小样例验证，再放大到真实任务。
- 交付时同时给结果、检查口径和下一步迭代建议。

SKILL.md SKILL.md

---
name: free-image-and-video-generation
description: Free local AI image and video processing toolkit with cloud AI generation. Local tools: upscale…
category: AI 智能
source: xramjetx/free-image-and-video-generation-skill
---

# free-image-and-video-generation

## 什么时候使用
- 把 AI / Agent方向的常用动作沉淀成 Agent 可调用的技能 适合处理AI Agent、提示词、模型评估与自动化推理，核心价值是把输入、判断、执行、验证和交付边界固定下来，避免 Agent 泛泛回答。 把任务拆成可执行、可检查…
- 面向提示词、Agent 工作流、模型评估或自动化推理，优先处理能明确输入、步骤和验收标准的工作。

## 需要提供什么
- 目标材料、目录范围、期望结果和不可改动内容。
- 是否允许联网、执行命令、读写文件或调用外部服务。

## 执行规则
- 围绕「Prerequisites / Available Tools / Usage」组织步骤，不把推断写成作者事实。
- 读取文件、写入/修改文件、执行终端命令、读取环境变量；会按任务需要访问外部网络；需要准备 Vendor-specific API Key。
- 先跑小样例，确认结果可检查后再扩大任务范围。

## 输出要求
- 给出最终产物、关键证据、验证方式和下一步动作。
- 信息不足时标记 unknown，不编造命令、平台或依赖。

证据边界与执行链路

作者原文负责流程事实；仓库文件负责来源和命令；流狐只补充适用场景、限制和质量判断。

skill "free-image-and-video-generation" {
  输入层     -> 用户目标 + 目标文件 + 禁止范围 + 验收标准
  上下文层   -> Prerequisites / Available Tools / Usage
  规则层     -> SKILL.md 触发条件 / 执行顺序 / 输出格式
  运行层     -> Python | 读取文件、写入/修改文件、执行终端命令、读取环境变量 | 会按任务需要访问外部网络
  安全层     -> 需要准备 Vendor-specific API Key + 小任务验证 + diff / 日志复核
  输出层     -> 可复制结果 + 检查清单 + 下一步迭代
}

解读按原文结构重写，命令、链接、术语均保留；右侧可核对作者原始 SKILL.md

先判断是否适合

核心用途：把 AI / Agent方向的常用动作沉淀成 Agent 可调用的技能适合处理AI Agent、提示词、模型评估与自动化推理，核心价值是把输入、判断、执行、验证和交付边界固定下来，避免 Agent 泛泛回答。把任务拆成可执行、可检查、可继续迭代的步骤…
适用任务：更适合 AI Agent、提示词、模型评估与自动化推理，尤其是输入、步骤、验收口径可以反复复用的场景。
不宜硬套：如果任务只有一次性问答、来源材料很少，或需要作者没有说明的命令和外部服务，应先降级为普通对话。

作者设计意图

触发依据：使用时机、上下文收集、输出格式和失败处理共同决定它是否适合当前任务。
来源边界：作者原文负责事实，仓库文件负责命令和文件证据；Fluxly 只补充用途、限制和使用判断。

作者的方法与取舍

执行主线：任务围绕「Prerequisites / Available Tools / Usage」组织，把触发条件、必要上下文、执行规则和交付标准连成一条可复核的链路。
设计重点：重点不在重复教程，而在解释作者为什么规定这些步骤、哪些条件不可省略，以及结果偏离时应回到哪一层检查。
使用判断：先确认任务与原文边界一致，再决定是否启用；来源没有说明的命令、平台和依赖保持未知，不由 Fluxly 补写。

边界和复核

依赖条件：需要提前准备 Vendor-specific API Key，未配置前不要直接跑完整任务。
权限提醒：会涉及 read / write / shell-exec / env-read 权限，执行前要说明可读写范围、命令边界和回退方式。
质量标准：结果应包含明确产物、必要证据和下一步动作；如果只是泛泛解释，需要补充输入和验收条件后重跑。

Free Image & Video Processing Toolkit

7 free local AI tools + cloud AI generation (300+ models via Atlas Cloud API).

Local tools run 100% on your machine — no API keys, no cloud costs. Cloud generation tools provide access to state-of-the-art AI models for image and video creation.

Prerequisites

Python 3.10+ installed
uv installed (curl -LsSf https://astral.sh/uv/install.sh | sh)
FFmpeg installed (brew install ffmpeg / apt install ffmpeg / winget install ffmpeg)

Available Tools

Tool	Script	What It Does
Image Upscale	`scripts/upscale.py`	2x/4x super resolution using Real-ESRGAN
Face Enhance	`scripts/face-enhance.py`	Restore and enhance faces using GFPGAN + CodeFormer
Background Remove	`scripts/bg-remove.py`	Remove image backgrounds, output transparent PNG
Object Erase	`scripts/erase.py`	Erase unwanted objects using LaMa inpainting
Face Swap	`scripts/face-swap.py`	Swap faces between images using InsightFace
Smart Segment	`scripts/segment.py`	Segment anything in images using FastSAM
Media Process	`scripts/media-process.py`	Convert, compress, resize, extract with FFmpeg
AI Generate	`scripts/ai-generate.py`	Generate images/videos with 300+ cloud AI models

Usage

All scripts use uv run for zero-setup execution — dependencies are automatically installed on first run.

Image Upscale (Real-ESRGAN)

Upscale low-resolution images by 2x or 4x with AI super resolution.

# 4x upscale (default)
uv run scripts/upscale.py input.jpg

# 2x upscale
uv run scripts/upscale.py input.jpg --scale 2

# Upscale with face enhancement
uv run scripts/upscale.py input.jpg --face-enhance

# Batch upscale a folder
uv run scripts/upscale.py ./photos/ --scale 4

# Custom output path
uv run scripts/upscale.py input.jpg -o upscaled.png

Face Enhance (GFPGAN + CodeFormer)

Restore old photos, enhance blurry faces, fix low-quality portraits.

# Enhance faces in an image (GFPGAN, default)
uv run scripts/face-enhance.py photo.jpg

# Use CodeFormer (better fidelity control)
uv run scripts/face-enhance.py photo.jpg --method codeformer

# Adjust fidelity (0=quality, 1=fidelity, default 0.5)
uv run scripts/face-enhance.py photo.jpg --method codeformer --fidelity 0.7

# Also upscale background (2x)
uv run scripts/face-enhance.py photo.jpg --bg-upscale 2

# Batch process
uv run scripts/face-enhance.py ./old-photos/

Background Remove (rembg)

Remove backgrounds from images, output transparent PNG. Supports multiple AI models.

# Remove background (default u2net model)
uv run scripts/bg-remove.py product.jpg

# Use specific model
uv run scripts/bg-remove.py photo.jpg --model isnet-general-use

# Batch process folder
uv run scripts/bg-remove.py ./products/ -o ./transparent/

# Keep only the foreground (alpha matting for fine edges)
uv run scripts/bg-remove.py portrait.jpg --alpha-matting

# Available models: u2net, u2netp, u2net_human_seg, u2net_cloth_seg,
#                   silueta, isnet-general-use, isnet-anime, sam

Object Erase (LaMa Inpainting)

Remove unwanted objects from images using a mask.

# Erase objects (white area in mask = erase)
uv run scripts/erase.py image.png --mask mask.png

# Auto-generate mask from coordinates (x,y,width,height)
uv run scripts/erase.py image.png --region 100,200,150,150

# Batch erase with matching masks (image1.png + image1_mask.png)
uv run scripts/erase.py ./images/ --mask-dir ./masks/

Face Swap (InsightFace)

Swap faces between two images.

# Swap face from source to target
uv run scripts/face-swap.py --source face.jpg --target photo.jpg

# Swap specific face index (when multiple faces detected)
uv run scripts/face-swap.py --source face.jpg --target group.jpg --face-index 0

# Custom output
uv run scripts/face-swap.py --source face.jpg --target photo.jpg -o result.png

Smart Segment (FastSAM)

Segment any object in an image using text prompt, point, or bounding box.

# Segment everything
uv run scripts/segment.py image.jpg

# Segment by text prompt
uv run scripts/segment.py image.jpg --text "the dog"

# Segment by point (x, y)
uv run scripts/segment.py image.jpg --point 400,300

# Segment by bounding box (x1,y1,x2,y2)
uv run scripts/segment.py image.jpg --box 100,100,400,400

# Output mask only
uv run scripts/segment.py image.jpg --text "car" --mask-only

Media Process (FFmpeg)

Convert, compress, resize, extract frames, merge audio/video — powered by FFmpeg.

# Convert format
uv run scripts/media-process.py convert input.mp4 output.webm

# Compress video (target size in MB)
uv run scripts/media-process.py compress input.mp4 --target-size 10

# Resize video
uv run scripts/media-process.py resize input.mp4 --width 1280 --height 720

# Extract frames as images
uv run scripts/media-process.py frames input.mp4 --fps 1 --output ./frames/

# Extract audio
uv run scripts/media-process.py audio input.mp4 -o audio.mp3

# Create GIF from video
uv run scripts/media-process.py gif input.mp4 --start 5 --duration 3 --fps 15

# Trim video
uv run scripts/media-process.py trim input.mp4 --start 00:01:00 --end 00:02:30

# Merge multiple videos
uv run scripts/media-process.py merge video1.mp4 video2.mp4 video3.mp4 -o combined.mp4

# Add watermark
uv run scripts/media-process.py watermark input.mp4 --image logo.png --position bottom-right

# Get media info
uv run scripts/media-process.py info input.mp4

Generate images and videos using 300+ state-of-the-art AI models. Requires an Atlas Cloud API key.

IMPORTANT for AI agents: Before calling this script, you MUST first use Atlas Cloud MCP tools to find the correct model ID and its required parameters:

Call atlas_list_models to browse available models, or atlas_search_docs to search for a specific model
Call atlas_get_model_info with the model ID to get the exact parameter schema (different models use different parameters — some use size, others use aspect_ratio + resolution, etc.)
Then call the script with --model <full_model_id> and the correct parameters

# Generate image (pass full model ID from Atlas Cloud)
uv run scripts/ai-generate.py image "A cat astronaut on the moon" --model black-forest-labs/flux-schnell --size 1024*1024

# Models using aspect_ratio + resolution (e.g. Nano Banana 2, Imagen4)
uv run scripts/ai-generate.py image "Anime girl with blue hair" --model google/nano-banana-2/text-to-image --aspect-ratio 1:1 --resolution 1k

# Models using size presets (e.g. Seedream)
uv run scripts/ai-generate.py image "Product photo on marble" --model bytedance/seedream-v5.0-lite --size 2048*2048

# Edit existing image
uv run scripts/ai-generate.py image "Make the sky sunset orange" --model bytedance/seedream-v5.0-lite/edit --image photo.jpg

# Generate video
uv run scripts/ai-generate.py video "Timelapse of cherry blossoms" --model alibaba/wan-2.6/text-to-video --size 1280*720

# Image-to-video
uv run scripts/ai-generate.py video "The person starts walking" --model alibaba/wan-2.6/image-to-video --image portrait.jpg

# Pass extra model-specific parameters as JSON
uv run scripts/ai-generate.py image "A logo" --model google/imagen4-ultra --extra '{"num_images": 4}'

# NSFW mode
uv run scripts/ai-generate.py image "Artistic figure study" --model black-forest-labs/flux-dev-lora --nsfw

Setup: Set ATLAS_CLOUD_API_KEY in environment or .env file. Get your key at atlascloud.ai.

Output

All tools save output to ./output/ by default. Use -o or --output to specify a custom path.

Models

Models are automatically downloaded on first use and cached locally:

Tool	Model	Size	Cache Location
Upscale	RealESRGAN_x4plus	~64MB	`~/.cache/realesrgan/`
Face Enhance	GFPGANv1.4	~348MB	`~/.cache/gfpgan/`
Face Enhance	CodeFormer	~376MB	`~/.cache/codeformer/`
Background Remove	u2net	~176MB	`~/.u2net/`
Object Erase	LaMa	~200MB	`~/.cache/lama/`
Face Swap	buffalo_l + inswapper	~500MB	`~/.insightface/`
Smart Segment	FastSAM-s	~23MB	auto-downloaded by ultralytics

Total first-run download: ~1.5GB. All subsequent runs use cached models.

Tips

GPU Acceleration: All tools automatically use CUDA/MPS if available, falling back to CPU
Batch Processing: Most tools accept a folder path for batch processing
Memory: Face swap and segmentation may need 4GB+ RAM for large images
First Run: First execution downloads AI models — subsequent runs are instant

Workflow Examples

Combine local processing with cloud AI generation:

# 1. Generate a product image with AI
uv run scripts/ai-generate.py image "Minimalist perfume bottle, studio lighting" --model bytedance/seedream-v5.0-lite --size 2048*2048

# 2. Upscale to 4x resolution
uv run scripts/upscale.py ./output/seedream-v5.0-lite_*.png --scale 4

# 3. Remove background for e-commerce
uv run scripts/bg-remove.py ./output/*_x4.png --alpha-matting

# 4. Generate a product video
uv run scripts/ai-generate.py video "A perfume bottle rotating slowly" --model kwaivgi/kling-v3.0-pro/text-to-video --duration 5

# 5. Add watermark to the video
uv run scripts/media-process.py watermark ./output/text-to-video_*.mp4 --image logo.png

安装命令

一键复制安装选择包管理器,点击右侧复制,粘贴到终端回车即可

npx skills add xramjetx/free-image-and-video-generation-skill/SKILL.md -y

查看通用安装教程

查看 Skill 原文

兼容工具

Claude Code
Cursor
Cline
Codex
Windsurf
Gemini
OpenCode
Roo Code
AAide
+Augment
GitHub Copilot
OpenClaw
KKiro
Junie
WWarp
Goose
Trae
Kilo Code
OpenHands
Amp
Zencoder
CContinue
AIAider
CodeBuddy
CodeGeeX
Replit Agent

安全审计

83 / 100

社区

未声明 license/许可。
未声明 version。
未限定 allowed-tools，默认拥有全部工具权限。
检出高风险片段：pipe_curl_to_shell

元信息

来源: github
xramjetx/free-image-and-video-generation-skill
English

分类与标签

分类 AI 智能