FUTO Swipe logo

FUTO Swipe.

Type naturally with AI that keeps your data on your device

Open-source swipe typing models that run efficiently on-device. Includes layout-agnostic encoder, decoder & 1M swipe training dataset. Small footprint, private & fast.

Rank
▲ #31
Votes
117
Platform
Web / Mobile
Launched
Recently
FUTO Swipe screenshot

More About FUTO Swipe

FUTO Swipe

FUTO Swipe is an open-source family of AI models and algorithms that delivers fast, accurate swipe typing while preserving user privacy. Built to power FUTO Keyboard and freely available for developers, it brings enterprise-grade mobile typing to anyone who values data ownership and transparency.

Product Highlights

  • Privacy-First Design: Runs entirely offline on your device with no data sent to external servers, eliminating the privacy risks of cloud-based keyboard solutions.
  • Triple-Model Architecture: Combines a universal encoder, language-specific ContextLM, and layout-specific decoder to achieve top-4 accuracy with only ~4% fail rate on test data.
  • Lightning-Fast Performance: Compact footprint of under 2.5 million total parameters enables millisecond inference even on low-end Android devices.
  • Open & Accessible: Released under the FUTO Model License with full attribution requirements, plus a GPL-licensed C++ inference library for easy integration.
  • Sustainably Trained: Developed using only a single workstation GPU, demonstrating that cutting-edge AI doesn't require massive compute resources or environmental cost.

Use Cases

  • Private Mobile Typing: Ideal for users who refuse to compromise personal data for convenience, offering big-tech accuracy without surveillance.
  • Developer Integration: Perfect for keyboard app creators, accessibility tools, and embedded systems needing reliable swipe input without licensing headaches.
  • Research & Education: Valuable for NLP researchers studying on-device language models, efficient architectures, and privacy-preserving AI systems.

Target Audience

FUTO Swipe serves privacy-conscious Android users seeking alternatives to data-harvesting keyboards, alongside developers and researchers who need production-ready, open-source swipe typing technology they can audit, modify, and deploy freely.