deepface.dev
API for face recognition
Last verified:
What is deepface.dev?
deepface.dev is a managed API built on top of the open-source DeepFace project, designed for teams that need face verification, embedding generation, and vector comparison without managing GPU infrastructure. It provides a hosted solution for face recognition tasks that unifies verification, embeddings, and vector comparison with multiple model backends and detectors.
Key features include face verification to compare two images and decide if they match, embedding generation for single face images, vector comparison for image and vector workflows, detection and alignment for cleaner inputs, and customer-owned face search capabilities using embeddings with your own vector database. The platform accepts controlled image inputs using files and base64 by default, and offers both direct REST API access and agent-ready MCP runtime for AI agents.
The platform is designed for AI agents and teams who need commercial-safe face recognition capabilities. AI agents can sign up anonymously, create their own dashboard session, and generate free standard API keys or dedicated MCP keys. Human users can stay in control with persistent billing, usage tracking, and key management through a self-serve dashboard. It supports multiple approved models including FaceNet, FaceNet512, OpenFace, Dlib, and SFace, with detector options including OpenCV, RetinaFace, MTCNN, SSD, dlib, MediaPipe, YOLOv8, YOLOv11, and YOLOv12.
deepface.dev is particularly suitable for developers building face comparison workflows, applications requiring face verification without infrastructure management, AI agents needing face recognition capabilities, and teams needing commercial-safe licensed backbones with explicit allowlists. The platform excludes VGG-Face due to non-commercial restrictions and requires separate licensing for ArcFace models based on InsightFace weights.
deepface.dev pricing
Pricing model: Freemium
deepface.dev offers usage-based pricing with credits per request that fall as volume scales. Anonymous AI accounts get 10 free calls per month with 1 free standard API key. Full non-anonymous accounts get 50 free calls per month. Adding a saved card once unlocks another 50 free calls for 100 total per month before paid usage. Free-call allowances apply while the account credit balance is at or below zero; after exhaustion, requests stop unless credits are added. Users can top up accounts as needed for more credits. Enterprise plans offer custom contracts with higher rate limits, dedicated queue capacity, SLAs, and privacy controls.
deepface.dev pros
- No GPU infrastructure management required
- Anonymous signup available for quick testing
- 100 free calls per month with saved card
- Both REST API and MCP runtime support
- AI agents can self-serve and create their own keys
- Multiple approved model backends (FaceNet, OpenFace, Dlib, SFace)
- 9 detector options including YOLOv12 and RetinaFace
- Customer-owned face search with Supabase pgvector integration
- Self-serve dashboard for usage, billing, and key management
- Commercial-safe model allowlists only
- Usage-based pricing that decreases at scale
- Live API demo tester on homepage
- Machine-readable docs for AI agents
- Enterprise plans with custom contracts and SLAs
- Face detection and alignment for consistent embeddings
- Controlled image inputs with files and base64 support
- Persistent billing and account upgrades without rework
deepface.dev cons
- VGG-Face excluded due to non-commercial restrictions
- ArcFace requires separate commercial license
- Free calls only apply while credit balance is at or below zero
- Anonymous accounts limited to 10 free calls per month
- Usage-based pricing may become expensive at high volume
- Standard API key limited to 1 per anonymous account
- Requires adding saved card to unlock full 100 free calls
- Enterprise pricing requires custom contracts not publicly listed
Frequently asked questions about deepface.dev
What is deepface.dev?
deepface.dev is a managed API built on top of the open-source DeepFace project for face verification, embedding generation, and vector comparison without managing your own GPU infrastructure. It provides a hosted solution with REST API and MCP runtime support.
How do I sign up as an AI agent?
AI agents can sign up anonymously at deepface.dev/signup?agent=1 to create a free account session. After signup, open the dashboard and create a standard API key for direct REST or a dedicated MCP key for https://deepface.dev/mcp.
How many free calls do I get?
Anonymous AI accounts get 10 free calls per month. Upgrading to a real non-anonymous account increases this to 50 free calls per month. Adding a saved card once unlocks another 50 free calls for 100 total per month before paid usage.
What API endpoints are available?
The public API surface includes POST /verify to compare two images, POST /represent to generate embeddings, POST /compare to compare image and vector pairs without storing embeddings, and GET /capabilities to inspect the current model and detector allowlists.
Which models are approved on deepface.dev?
Approved models include FaceNet, FaceNet512, OpenFace, Dlib, and SFace. VGG-Face is excluded because its released weights are non-commercial. ArcFace models based on InsightFace weights require a separate commercial license.
What detectors are supported?
Approved detectors include OpenCV (default), RetinaFace, MTCNN, SSD, dlib, MediaPipe, YOLOv8, YOLOv11, and YOLOv12. You can use GET /capabilities to read the current allowlist before automating requests.
Can I use deepface.dev for customer-owned face search?
Yes, for face search across a customer-owned database, you can use embeddings with your own vector database. The platform provides a guide for customer-owned face search with Supabase pgvector integration.
What is the MCP runtime?
The MCP runtime is available at https://deepface.dev/mcp and provides agent-ready access for AI agents. You can create a dedicated MCP key from the account dashboard after signing in for MCP client connections.
Is enterprise pricing available?
Yes, enterprise plans include custom contracts with faster rate limits, higher rate limits and concurrency ceilings, dedicated queue capacity, tuned timeout policies, custom contract terms and privacy controls, and direct support for rollout and traffic planning.
What happens after I exhaust my free calls?
After your free allowance is exhausted, requests should stop passing unless credits are added. Free-call allowances apply while the account credit balance is at or below zero. You can top up your account as needed for more credits to continue using the API.