WaveSpeed

Open-source, cross-platform application for running 1000+ AI models — image generation, video generation, face swap, digital human, motion control, and more. Features a visual workflow editor for building AI pipelines, Featured Models with smart variant switching, and 12 free creative tools. Available for Windows, macOS, Linux, and Android.

Playground Screenshot

Android App

The Android app shares the same React codebase as the desktop version, giving you access to the AI Playground, Featured Models, Creative Studio, and all 1000+ models from your phone.

Full AI Playground with multi-tab support and all input types including camera capture
Featured Models with smart variant switching
Model browser with search, filter, and sort
Creative Studio tools (face enhancement, background removal, image eraser, segment anything, media conversion)
History, My Assets, templates, and auto-save
18 languages, dark/light theme, Android 5.0+

Creative Studio

12 free AI-powered creative tools that run entirely in your browser. No API key required, no usage limits, completely free. Also available as a standalone web app at wavespeed.ai/studio — fully responsive, works on desktop, tablet, and mobile browsers.

Tool	Description
Image Enhancer	Upscale images 2x–4x using ESRGAN with slim, medium, and thick quality options
Video Enhancer	Frame-by-frame video upscaling with real-time progress and ETA
Face Enhancer	Detect faces with YOLO v8 and enhance with GFPGAN v1.4 (WebGPU accelerated)
Face Swapper	Swap faces using InsightFace (SCRFD + ArcFace + Inswapper) with optional GFPGAN post-processing
Background Remover	Remove backgrounds instantly — outputs foreground, background, and mask with individual downloads
Image Eraser	Remove unwanted objects with LaMa inpainting, smart crop and blend (WebGPU accelerated)
Segment Anything	Interactive object segmentation with point prompts using SlimSAM
Video Converter	Convert between MP4, WebM, AVI, MOV, MKV with codec and quality options
Audio Converter	Convert between MP3, WAV, AAC, FLAC, OGG with bitrate control
Image Converter	Batch convert between JPG, PNG, WebP, GIF, BMP with quality settings
Media Trimmer	Trim video and audio by selecting start and end times
Media Merger	Merge multiple video or audio files into one

WaveSpeed Creative Studio

Visual Workflow Editor

Node-based pipeline builder for designing and executing complex AI workflows. Chain any combination of AI models, free tools, and media processing steps into automated pipelines.

WaveSpeed Visual Workflow Editor

Features

AI Playground: Multi-tab playground with dynamic forms, batch processing (2-16x), mask drawing, LoRA support, abort control, and auto-randomized seeds
Featured Models: Curated model families with smart variant switching — auto-selects the best variant based on inputs and toggles (Seedream 4.5, Seedance 1.5 Pro, Wan Spicy, InfiniteTalk, Kling 2.6, Nano Banana Pro, etc.)
Model Browser: Fuzzy search, sort by popularity/name/price/type, favorites filter
Visual Workflow Editor: Node-based pipeline builder with 20+ node types
- Triggers (directory scan, HTTP API), AI tasks, 12 free tool nodes, processing (concat, select), group/subgraph, I/O nodes
- Run all / run node / continue / retry / cancel / batch runs (1-99x), real-time execution monitor with cost tracking
- Group/subgraph containers with exposed I/O, breadcrumb navigation, and workflow import
- HTTP API mode: expose workflows as REST endpoints via built-in HTTP server — works as a skill server for OpenClaw and other AI agents
- Directory batch processing: auto-execute per media file in a folder
- Prompt optimizer, guided tour, result caching, circuit breaker, cycle detection
- Cost estimation & daily budget, import/export (JSON + SQLite), multi-tab, undo/redo, customizable output naming
Free Tools: 12 AI-powered creative tools (no API key) — see Creative Studio above
Z-Image: Local image generation via stable-diffusion.cpp with model downloads, progress, and logs
Templates: Playground + workflow templates with presets, i18n search, import/export, and usage tracking
History & Assets: Recent predictions (24h), saved outputs with tags/favorites/search, auto-save to local folder
Media Input: File upload (drag & drop), camera capture, video/audio recording
18 languages, dark/light/auto theme, auto updates (stable + nightly), cross-platform (Windows, macOS, Linux, Android)
Cross-Platform: Available for Windows, macOS, Linux, and Android

Installation

Quick Download

Desktop

Mobile

Or browse all releases on the Releases page.

Platform Instructions

<details> <summary>Windows</summary>

Download .exe (installer) or .zip (portable)
Run the installer and follow the prompts, or extract the zip file
Launch "WaveSpeed Desktop" from Start Menu or the extracted folder

</details> <details> <summary>macOS</summary>

Download .dmg for your chip (Apple Silicon or Intel)
Open the .dmg file and drag the app to Applications
Launch the app from Applications

</details> <details> <summary>Linux</summary>

Download .AppImage or .deb
For AppImage: Make it executable (chmod +x *.AppImage) and run it
For .deb: Install with sudo dpkg -i *.deb

</details> <details> <summary>Android</summary>

Download the .apk file
Open the file on your Android device
If prompted about "Unknown sources", allow installation from this source
Install and launch the app
Requires Android 5.0 (API 21) or higher

</details>

Nightly Builds

Note: Nightly builds may be unstable. Use the stable releases for production use.

Development

Prerequisites

Node.js 20+
npm

Setup

bash
# Clone the repository
git clone https://github.com/WaveSpeedAI/wavespeed-desktop.git
cd wavespeed-desktop

# Install dependencies
npm install

# Install pre-commit hooks (requires pre-commit: pip install pre-commit)
pre-commit install

# Start development server
npm run dev

Scripts

Script	Description
`npm run dev`	Start development server with hot reload
`npm run dev:web`	Start web-only dev server (no Electron)
`npm run build`	Build the application
`npm run build:web`	Build web-only version (no Electron)
`npm run build:win`	Build for Windows
`npm run build:mac`	Build for macOS
`npm run build:linux`	Build for Linux
`npm run build:all`	Build for all platforms
`npm run dist`	Build and package for distribution
`npm run format`	Format code with Prettier
`npm run format:check`	Check code formatting

Mobile Development

The mobile app is located in the mobile/ directory and shares code with the desktop app.

bash
# Navigate to mobile directory
cd mobile

# Install dependencies
npm install

# Start development server
npm run dev

# Build and sync to Android
npm run build && npx cap sync android

# Open in Android Studio
npx cap open android

See mobile/README.md for detailed mobile development guide.

Project Structure

wavespeed-desktop/
├── data/templates/         # Preset workflow templates (AI generation, image/video/audio processing)
├── electron/               # Electron main process
│   ├── main.ts             # Main process entry
│   ├── preload.ts          # Preload script (IPC bridge)
│   ├── lib/                # Local generation (sdGenerator for stable-diffusion.cpp)
│   └── workflow/           # Workflow backend
│       ├── db/             # SQLite database (workflow, node, edge, execution, budget, template repos)
│       ├── engine/         # Execution engine (DAG runner, scheduler, cache, circuit breaker)
│       ├── ipc/            # IPC handlers (workflow, execution, history, cost, storage, http-server)
│       ├── nodes/          # Node handlers (AI task, free tools, I/O, triggers, processing, control)
│       ├── services/       # HTTP server, model list, retry, service locator, template loader
│       └── utils/          # File storage, hashing, save-to-assets
├── src/
│   ├── api/                # API client
│   ├── components/         # React components
│   │   ├── ffmpeg/         # FFmpeg components
│   │   ├── layout/         # Layout components
│   │   ├── playground/     # Playground components
│   │   ├── shared/         # Shared components
│   │   ├── templates/      # Template components
│   │   └── ui/             # shadcn/ui components
│   ├── hooks/              # Custom React hooks
│   ├── i18n/               # Internationalization (18 languages)
│   ├── lib/                # Utilities (fuzzy search, schema-to-form, smart form config, etc.)
│   ├── pages/              # Page components
│   ├── stores/             # Zustand stores
│   ├── types/              # TypeScript types
│   ├── workers/            # Web Workers (upscaler, face enhancer/swapper, background remover, image eraser, segmentation, ffmpeg)
│   └── workflow/           # Workflow frontend
│       ├── browser/        # Browser-only workflow API (web mode without Electron)
│       ├── components/     # Canvas, node palette, config panel, results panel, run monitor, prompt optimizer
│       ├── hooks/          # Workflow-specific hooks (undo/redo, group adoption, free tool listener)
│       ├── ipc/            # Type-safe IPC client
│       ├── lib/            # Cycle detection, free tool runner, model converter, topological sort
│       ├── stores/         # Workflow, execution, UI stores (Zustand)
│       └── types/          # Workflow type definitions
├── mobile/                 # Mobile app (Android)
│   ├── src/                # Mobile-specific overrides
│   ├── android/            # Android native project
│   └── capacitor.config.ts
├── .github/workflows/      # GitHub Actions (desktop + mobile)
└── build/                  # Build resources

Tech Stack

Desktop

Framework: Electron + electron-vite
Frontend: React 18 + TypeScript
Styling: Tailwind CSS + shadcn/ui
State Management: Zustand
HTTP Client: Axios
Workflow Canvas: React Flow
Workflow Database: sql.js (SQLite in-process)
AI/ML (Free Tools): @huggingface/transformers, onnxruntime-web, @tensorflow/tfjs, upscaler (ESRGAN)
Media Processing: @ffmpeg/core, mp4-muxer, webm-muxer
3D Preview: @google/model-viewer
Local Generation: stable-diffusion.cpp (via sdGenerator)

Mobile

Framework: Capacitor 6
Frontend: React 18 + TypeScript (shared with desktop)
Styling: Tailwind CSS + shadcn/ui (shared)
Platform: Android 5.0+

Configuration

Launch the application
Go to Settings
Enter your WaveSpeedAI API key
Start using the Playground!

Get your API key from WaveSpeedAI

API Reference

The application uses the WaveSpeedAI API v3:

Endpoint	Method	Description
`/api/v3/models`	GET	List available models
`/api/v3/{model}`	POST	Run a prediction
`/api/v3/predictions/{id}/result`	GET	Get prediction result
`/api/v3/predictions`	POST	Get prediction history
`/api/v3/media/upload/binary`	POST	Upload files
`/api/v3/balance`	GET	Get account balance

The built-in workflow HTTP server also exposes:

Endpoint	Method	Description
`/api/health`	GET	Health check
`/api/workflows/{id}/run`	POST	Trigger a workflow execution via API
`/api/workflows/{id}/schema`	GET	Get workflow input/output schema
`/schema`	GET	Get active workflow schema
`POST /` (any path)	POST	Run the active workflow

Add an HTTP Trigger node to your workflow to define the API input schema (each field becomes an output port), and optionally add an HTTP Response node to customize the response. Start the server from the workflow canvas — it listens on a configurable port (default 3100) with CORS enabled.

This turns any workflow into a callable REST endpoint, making it easy to integrate with OpenClaw or other AI agent frameworks as a skill server. For example, an OpenClaw agent can call GET /api/workflows/{id}/schema to discover the workflow's input/output contract, then POST /api/workflows/{id}/run with the required fields to execute the pipeline and receive results.