Quick Start

Get up and running with WaveSpeedAI in minutes.

From advertising campaigns to short films, e-commerce visuals to social media content — creative professionals and developers worldwide trust WaveSpeedAI for fast, high-quality AI generation.

For policies, billing details, and FAQs, see User Guide & FAQ.

1. Create an Account

Sign up at wavespeed.ai with your Google or GitHub account.

New accounts receive $1 trial credit — available for most models (some models excluded).

2. Explore the Platform

Main Pages

Resources

3. Choose Your Path

Web Interface (No Coding Required)

The easiest way to get started — no API key needed:

1. Go to wavespeed.ai/models
2. Select a model
3. Enter your prompt and click Run

Playground Tabs

Each model page has four tabs:

• Playground — The main interface where you configure parameters and run the model.
• JSON — Shows your current input as JSON format, useful for debugging or copying to your code.
• API — Provides ready-to-use API code (cURL, Python, etc.) based on your current parameter settings. Great for developers who want to quickly integrate.
• History — View your previous generations for this specific model.

Toolbar

In the top-right corner of each model page, you’ll find:

• Add to favourite — Click the star icon to save this model to your favorites for quick access later.
• Schema — View the model’s detailed input/output schema, helpful for understanding available parameters.
• LLMs — Some models support multiple LLM backends. Use this dropdown to switch between them.

Prompt Enhancer

Located next to the prompt input field, the Prompt Enhancer automatically improves your prompt for better results. Note: This feature has an additional cost per use.

Run Button & Batch Mode

The Run button displays an estimated cost before you click. Remember, the final charge may vary slightly.

Batch Mode OFF:

Batch Mode ON:

• Batch Mode — Enable this to generate multiple outputs at once. When enabled, billing is calculated per image/output generated.
• Safety Checker — A content moderation filter that screens outputs. This is enabled by default on the web interface to ensure compliance with content policies.

API & Developer Tools

For programmatic access, choose your integration method:

4. For API Users

Get Your API Key

1. Go to API Keys
2. Enter a memorable name and click Generate

Important: API keys require a top-up to activate. Keys generated without a top-up will not work.

Security Best Practices: Treat your API key as a secret. Store it in environment variables, never expose it in frontend code, and rotate it regularly.

For more details, see our Authentication Guide.

Code Examples

Python SDK

import wavespeed
 
output = wavespeed.run(
    "wavespeed-ai/flux-dev",
    {"prompt": "A cat wearing a space suit"},
)
print(output["outputs"][0])

cURL

curl -X POST "https://api.wavespeed.ai/api/v3/wavespeed-ai/flux-dev" \
  -H "Authorization: Bearer YOUR_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{"prompt": "A cat wearing a space suit"}'

5. Popular Models

Image Generation

Video Generation

Audio & Music

Advanced Tools

With 700+ models covering image, video, audio, 3D, and more — WaveSpeedAI is your all-in-one platform for AI generation.

Browse all models at wavespeed.ai/models.

6. LoRA Support

WaveSpeedAI supports both training custom LoRA models and using existing LoRA models in your generations.

Train Your Own LoRA

Create personalized AI models without coding:

1. Upload Dataset — Compress 10-20 diverse images into a .zip file
2. Set Trigger Word — Define a unique identifier (e.g., “p3r5on”) for your model
3. Configure Parameters — Adjust steps, learning rate, and LoRA rank
4. Train — Click “Run” and wait for completion
5. Use or Download — Access via URL or download the .safetensors file

Note: Training time depends on dataset size and parameters (e.g., steps). Actual training often takes longer than the estimated time displayed. Please be patient. If a timeout or system error occurs, refunds are processed automatically.

Use Existing LoRA Models

Apply LoRA models from Hugging Face or Civitai in your API calls:

• Format: Only .safetensors files are supported
• Path: Use owner/model-name (Hugging Face) or direct URL (Civitai)
• Scale: Controls LoRA influence strength (default: 1)
• Trigger Word: Include in your prompt to activate the LoRA style

Compatibility: LoRA must match the base model exactly — same model name, version, and parameter size.

Learn more: Train LoRA Guide | Use LoRA Guide

7. Need Help?

• Billing & Account: See User Guide & FAQ
• Technical Support: support@wavespeed.ai
• Live Chat: Click the chat icon in the bottom-right corner of our website