modellix

Capabilities

Modellix is a unified API platform that enables agents to access 100+ AI models for generative tasks. Agents can generate images from text, create videos from text or images, edit existing images, and perform specialized tasks like virtual try-on, image outpainting, and text-based image translation. The platform supports asynchronous task processing with full request/response logging and transparent pricing.

Skills

Text-to-Image Generation

• Generate images from text prompts using models like Qwen Image Plus, Wanx, Seedream, and Hailuo
• Supports multiple artistic styles and high-quality image generation
• Endpoint: POST /api/v1/text-to-image/{provider}/{model_id}/async
• Example: curl -X POST https://api.modellix.ai/api/v1/text-to-image/alibaba/qwen-image-plus/async -H "Authorization: Bearer YOUR_API_KEY" -d '{"prompt": "A cute cat playing in a garden"}'

Text-to-Video Generation

• Create videos from text descriptions with cinematic quality
• Models include Wan 2.6 T2V, Seedance, and Hailuo 02 T2V
• Supports automatic dubbing and custom audio file uploads
• Endpoint: POST /api/v1/text-to-video/{provider}/{model_id}/async
• Charged by video duration (USD/sec)

Image-to-Image Editing

• Edit existing images through text instructions
• Capabilities include style transfer, watermark removal, image expansion, detail enhancement, and object addition/removal
• Models: Wanx 2.1 Image Edit, Qwen Image Edit Plus, Seedream 4.0+ I2I, Seededit 3.0
• Endpoint: POST /api/v1/image-to-image/{provider}/{model_id}/async

Image-to-Video Generation

• Generate videos from image references combined with text prompts
• Supports first-and-last-frame video generation (KF2V models)
• Models: Wan 2.6 I2V, Seedance I2V, Hailuo 02 I2V
• Endpoint: POST /api/v1/image-to-video/{provider}/{model_id}/async
• Charged by video duration (USD/sec)

Specialized Image Tasks

• Virtual Try-On: Generate try-on images from portrait and clothing photos (AI Try-On, AI Try-On Plus)
• Image Outpainting: Extend images with free expansion and rotation support
• Image Translation: Translate text in images across 11 languages while preserving layout
• WordArt: Create artistic text with semantic deformation or texture effects
• Image Parsing: Segment model and clothing images for preprocessing

Async Task Management

• Submit tasks and receive immediate task_id response
• Query task status and results using: GET /api/v1/tasks/{task_id}
• Task results include resource URLs, metadata, and timing information
• Results retained for 24 hours after completion
• Response includes status (pending/success/failed), duration, and generated resources

API Authentication & Key Management

• Create, view, and delete API keys through Modellix console
• Authentication via Bearer token: Authorization: Bearer YOUR_API_KEY
• API keys displayed only once after creation
• Support for Email OTP, Google, and GitHub authentication

Rate Limiting & Quotas

• Global limit: 1000 requests/minute
• Per API key limit: 100 requests/minute
• Per model limits subject to provider constraints
• Response headers include: X-RateLimit-Limit, X-RateLimit-Remaining, X-RateLimit-Reset
• Concurrent task limits enforced per team

Error Handling

• Unified error response format with code and message fields
• HTTP status codes: 200 (success), 400 (bad request), 401 (unauthorized), 404 (not found), 429 (rate limit), 500 (server error), 503 (service unavailable)
• Error categories: Invalid parameters, Missing required parameter, Invalid format, Value out of range, Authentication failed, Resource not found, Rate limit exceeded, Concurrent limit exceeded
• Retryable errors: 429, 500, 503 with exponential backoff strategy

Request/Response Logging

• Complete logging of all API calls including input parameters, output parameters, timestamps, response time, and cost consumption
• Financial-grade transaction ledger system for every user
• Real-time visibility into recharge records and consumption details
• Supports business optimization and audit compliance

Workflows

Basic Image Generation Workflow

1. Register and log in to Modellix console
2. Create API key at https://modellix.ai/console/api-key
3. Submit text-to-image request: POST /api/v1/text-to-image/alibaba/qwen-image-plus/async
4. Receive task_id in response
5. Poll task status: GET /api/v1/tasks/{task_id}
6. Retrieve generated image URL from result when status is “success”
7. Download image before 24-hour expiration

Video Generation with Custom Audio

1. Prepare text prompt describing desired video
2. Submit text-to-video request with optional audio parameters
3. Receive task_id
4. Monitor task progress via polling or streaming interface
5. Retrieve video URL and metadata when complete
6. Extract video duration for cost calculation (charged per second)

Image Editing Pipeline

1. Prepare source image and editing instructions
2. Submit image-to-image request with image URL and text prompt
3. Receive task_id
4. Query task status until completion
5. Retrieve edited image from result resources
6. Optionally chain multiple edits by using output as input

Error Recovery with Retry Logic

1. Attempt API request
2. Check response code
3. If 429: Extract X-RateLimit-Reset header, wait until reset time
4. If 500/503: Apply exponential backoff (1s, 2s, 4s) with max 3 retries
5. If 400/401/404: Fix request parameters and retry
6. Log error details including timestamp, parameters, and error message

Integration

MCP (Model Context Protocol) Integration

• Remote MCP server available at: https://docs.modellix.ai/mcp
• Compatible with Cursor, Claude Desktop, and other MCP clients
• Search tool for Modellix documentation with filter parameters (version, language, apiReferenceOnly, codeOnly)
• Enables AI applications to search documentation during response generation

OpenAI Integration

• Use Modellix as remote MCP server with OpenAI models
• Configure MCP tool in OpenAI API calls with server URL and label
• Allows GPT models to access Modellix documentation and capabilities

Multi-Provider Support

• Alibaba (Qwen, Wanx, Wan models)
• ByteDance (Seedream, Seedance, Seededit)
• MiniMax (Hailuo, Image-01)
• Additional providers continuously added

Context

Async Processing Model

All Modellix API calls are asynchronous. Agents must submit a request, receive a task_id, then poll the task status endpoint to retrieve results. This enables handling of long-running generation tasks without blocking connections.

Pricing Structure

• Text-to-image and image-to-image: Charged per image (USD/img)
• Text-to-video and image-to-video: Charged per second of video (USD/sec)
• Transparent pricing with granular unit costs for different models and parameters
• More competitive rates for select core models compared to official pricing

Result Retention

Generated results are stored for 24 hours after task completion. Agents must download or save results within this window or they will be permanently deleted.

Rate Limiting Strategy

Team-level rate limiting applies to all API keys under the same account. Agents should implement client-side concurrency control and monitor X-RateLimit-Remaining header to avoid frequent limit triggers. Exponential backoff is recommended for retryable errors.

Enterprise Reliability

• AWS Singapore infrastructure with redundant deployment
• 15-year enterprise IT service experience
• Rolling upgrades and strict SLA operational standards
• Support in Mandarin, Cantonese, and English via email and Discord community

For additional documentation and navigation, see: https://docs.modellix.ai/llms.txt