Overview
Hush is an open-source noise suppression model by Weya AI, designed to clean real-time call audio by removing background noise and competing voices. It processes audio frames in under 1 ms on CPU, making it ideal for voice AI agents, ASR systems, and call centers. Trained on over 10,000 hours of real-world noisy audio, Hush ranks among the top speech-enhancement models on Hugging Face.
Persona.js is a lightweight, open-source AI chat UI library by Runtype. It is framework-free, backend-agnostic, and WebMCP-native, allowing agents to discover and execute tools exposed by the parent page. Persona supports floating, docked, or fullscreen layouts, streaming chat, voice, and theming, and can be integrated via npm or a simple script tag.
Feature Comparison
| Feature | Hush | Persona.js |
|---|---|---|
| Primary Function | Real-time noise suppression for voice AI calls | Lightweight AI chat UI library for websites |
| Target Audience | Voice AI developers, call centers, BFSI teams | Web developers building AI assistants, copilots, or chat widgets |
| Deployment | Cloud or data center (model size ~8 MB) | Client-side via npm or script tag, any backend |
| Real-time Performance | Processes 10 ms audio in <1 ms on CPU | Streaming SSE with pluggable parsers |
| Integration Complexity | Slots into existing Voice AI stack | Minimal setup: import, mount, point to SSE endpoint |
| Open Source | Yes (open-source model) | Yes (open-source library) |
| Key Differentiator | Focuses on cleaning noisy audio for ASR | WebMCP-native, agents discover page tools |
Pricing
Hush is free and open-source. Weya AI offers a 2-week pilot for enterprise evaluation, where they implement the model in one workflow and measure lift before full rollout.
Persona.js is free and open-source. The Runtype CLI and proxy services may have separate pricing, but the core library is freely available under an open-source license.
Pros and Cons
Hush
Pros:
- Real-time CPU processing, no GPU needed
- Trained on 10,000+ hours of real noisy audio
- Top-5 on Hugging Face Audio-to-Audio leaderboard
- Handles overlapping speakers and tough environments
- Lightweight model (8 MB)
Cons:
- Limited to audio enhancement only
- Requires integration with existing Voice AI stack
- Not a standalone chat or UI solution
Persona.js
Pros:
- Framework-agnostic, works with any stack
- WebMCP-native for agentic tool discovery
- Shadow DOM isolation prevents style leaks
- Multiple layout modes: floating, docked, fullscreen
- Extensive demo and integration examples
Cons:
- Requires SSE backend for streaming
- No built-in noise suppression
- Primarily a UI library, not a full agent framework
Verdict
Choose Hush if your primary need is cleaning noisy audio for voice AI agents in real-time calls. It excels at removing background noise and competing voices, improving ASR accuracy and call quality. Choose Persona.js if you want to quickly add a customizable, agentic chat UI to any website without framework lock-in. It is ideal for building copilots, support widgets, or full-screen assistants that can interact with page tools. Both are open-source and excel in their respective domains.
Hush
Persona.js