I'm Ai-nyan! Ask me anything about ailia.
Close
Profile
An inference SDK for developers who want to run AI outside of Python — in embedded systems, mobile, and games. Includes audio, LLM, and tracking libraries for Unity, Kotlin, and C++, enabling end-to-end edge AI application development.
Experience AI powered by ailia SDK right now — object detection, pose estimation, super resolution, speech recognition, speech synthesis, LLM, and more.
Need to run AI in offline environments — factories, vehicles, and field sites without network access
Cloud APIs charge per inference — costs explode at scale
Sensitive data — images, audio, documents — cannot be sent outside the organization
Python AI libraries (Transformers, etc.) can't be used from Unity or Kotlin
Fully offline inference — no network required, runs entirely on device
License fee only — no per-inference or per-user charges
All processing stays on device — data never leaves your environment
ailia Tokenizer / Speech / LLM / Tracker available for Unity, Kotlin, and C++
Use Cases (Selected)
ailia SDK was adopted for AI inference in the Smart Smoothing feature, providing a cross-platform, GPU-accelerated inference engine.
View details →Achieved Smart Smoothing on Galaxy devices through a proprietary high-speed inference implementation using Vulkan on Android.
View details →Real-time driver state detection via a steering-wheel camera — no cloud connection required. Multiple AI models for face, gaze, and posture run simultaneously.
View details →Adopted for NTT Docomo's autonomous wheelchair initiative. Obstacle detection and privacy masking run in real time on the edge.
View details →AI implementation at a soccer experience facility in Tokyo Dome City. FaceSwap, video generation, and LLM dialogue are unified under ailia SDK for long-term stable operation.
View details →AI-powered features — including automatic metadata input and similar image search — were added to the DAM system "CIERTO" using ailia SDK.
View details →ailia TFLite Runtime was implemented on NSITEXE's RISC-V processor DR1000C, enabling AI inference in Non-OS and RTOS environments.
View details →Achieved high-speed inference using Intel® Xe Matrix Extensions (XMX) through collaboration with Intel.
View details →Speech recognition, speech synthesis, low-latency edge AI, and generative AI were all implemented with ailia SDK, powering an interactive pavilion that welcomed hundreds of thousands of visitors.
View details →Comparison with ONNX Runtime
Because every library is developed in-house, ailia SDK can be deeply optimized for specific devices and models — covering areas that general-purpose designs struggle with, such as direct Vulkan/Metal implementation and model-specific tuning.
| ailia SDK | OSS Runtime | |
|---|---|---|
|
GPU Inference via Native Vulkan & Metal
Android (Vulkan), Windows/Linux (Vulkan or CUDA), iOS/macOS (Metal)
|
 |
 |
|
Speech, TTS, LLM, Tokenizer & Tracking Libraries
ailia Speech / Voice / LLM / Tokenizer / Tracker — Unity, Flutter, C++ support
|
|
|
|
400+ Official Model Library
Validated models with sample code, ready to run immediately
|
|
|
|
Non-OS / RTOS Support
In-house libraries enable support for arbitrary OS environments
|
|
|
|
Official Unity Bindings + Model Collection
Pick a model and deploy instantly for games and XR
|
|
 |
|
Model-Specific Optimization
In-house libraries enable deep optimization for specific models and devices
|
|
|
△ = Supported but optimization is limited due to general-purpose implementation ※ Based on our own research
We officially provide a complete library suite — from LLM tokenization to object tracking — for non-Python environments including Unity, Kotlin, and C++.
ailia AI Library Suite
ailia SDK is the inference engine. Combine it with dedicated libraries for speech recognition, speech synthesis, and LLM — and you have a complete AI product with zero cloud dependency.
Offline speech recognition based on OpenAI Whisper. Supports 99 languages. Available for Unity, C++, and Flutter.
Offline TTS based on GPT-SoVITS. Clone a voice from just 10 seconds of audio. Available for Unity and Flutter.
Run GGUF-format LLMs on device. Supports Llama 3, Gemma 3, and Qwen. Available for C++, Unity, and Flutter.
Industry Use Cases
Select an industry to see how ailia SDK solves real-world challenges.
Manufacturing · Quality Inspection
Run anomaly detection models directly on an embedded board attached to a camera. No network needed — factory data stays on-premises.
Automotive · ADAS · Driver Monitoring
Analyze surrounding conditions and the driver in real time. Compatible with various companies’ SoCs, it supports implementing AI models in in-vehicle devices.
Models Used — ailia MODELS
Security · Surveillance Camera
Person detection and crowd counting run entirely on the camera device. No video upload required — only event metadata is transmitted, minimizing privacy exposure.
Models Used — ailia MODELS
Games · Media · Interactive AI
You can interact with NPCs, create dance games, and realize interactive image processing for users.
Models Used — ailia MODELS
Healthcare · Regulated Devices
Model encryption protects diagnostic AI IP. Fully offline inference ensures patient data never leaves the device — privacy protection and data residency requirements are built in from the start.
Models Used — ailia MODELS
Licensing & Pricing / Services
Evaluate for free. License when you're ready to ship. No per-inference charges — ever. For AI acceleration and optimization, see AI Computing.
Free · No credit card required
Full SDK access for 30 days
All 400+ models included
All platforms included
Internal testing only
Annual subscription (per app) and device-based licensing are all available.
Integration into production & commercial apps
Redistribution rights to end users
Email & meeting support
Dedicated team in Japanese & English
Custom quote · Free consultation
Model quantization, optimization & acceleration
Embedded environment (CPU/GPU constrained) support
End-to-end optimization including pre/post-processing
GPU parallel processing acceleration
SoC porting support
Expert engineers in both AI and embedded systems
Get Started
The evaluation version is free for 30 days — no API key, no credit card required. Once you've confirmed it works, we'll recommend the right license for your use case.
The ailia AI Showcase app lets you experience AI powered by ailia SDK — no installation needed. Includes object detection, pose estimation, super resolution, speech recognition, TTS, LLM, and more.
pip3 install ailia — the evaluation SDK is ready in seconds.
git clone https://github.com/ailia-ai/ailia-models.git - Get 400+ models and sample code. Run the model you need right away.
Object detection, pose estimation, speech recognition — all running offline.
Once you've validated it works, share your use case, device count, and app structure. We'll propose the best license — annual subscription or device-based. Consultation and estimates are always free.
Select your programming language to view the API docs.
