On-device AI that scales.
Local inference. Multi-Model orchestration. Edge/Cloud hybrid execution.
All in One SDK.
Easy Integration
From input to output
Chain models together or run them standalone. Define complex workflows in simple YAML.
pipeline: voice-agent
steps:
- model: whisper-tiny
task: transcribe
- model: gpt-4
task: reason
fallback: cloud
- model: kokoro-82m
task: synthesizefinal result = await Xybrid.runPipeline(
'voice-agent',
input: audioBytes,
);Get started in seconds
Run a single model or chain them into complex pipelines.
Same YAML config works across all platforms
Same YAML
name: voice-assistant
stages:
- id: transcribe
model: whisper-tiny
route: device
- id: think
model: gpt-4o-mini
route: cloud
input: ${transcribe.text}
- id: speak
model: kokoro-tts
route: device
input: ${think.text}Code
// Load the pipeline from assets
final loader = Xybrid.pipeline('assets/voice-assistant.yaml');
final pipeline = await loader.load();
// Create an audio envelope
final envelope = Envelope.audio(
audioBytes: audioBytes,
sampleRate: 16000,
channels: 1,
);
// Run the pipeline and play the result
final result = await pipeline.run(envelope: envelope);
await audioPlayer.play(result.audio);Everything you need for edge ML
From speech recognition to text-to-speech, run powerful ML models anywhere.
Edge-First
Run ASR and TTS models directly on-device for low latency and offline support.
Cloud Fallback
Seamlessly route to cloud APIs when device capabilities are insufficient.
Pipeline DSL
Chain models together with YAML pipelines: ASR → LLM → TTS in one config.
Cross-Platform
Flutter, iOS, Android, and Rust SDKs with unified API.
NPU Acceleration
Leverage CoreML, QNN, and Metal for maximum performance.
Telemetry
Track inference metrics and device capabilities across your fleet.
Ready to get started?
Join developers building the next generation of voice-enabled applications.