The Closest One
Voice Ring, built for quiet-voice mode. Whisper-tuned for the speech band.
Free shipping20+ countriesShips Father's Day, 2027
Make a fist, bring it to your mouth. The ring sits 3–5cm from your lips. That's the most important part of our Quiet-Voice Engine.
Voice Ring
Tap to pay. No need to dig for your phone.
Official Payment Partner
NihaoPay is a leading Silicon Valley fintech platform and global acquirer for UnionPay, Alipay, and WeChat Pay. It offers seamless one-stop cross-border payment solutions for merchants to accept billions of mobile wallet users worldwide.
CEO Rufei: "Projected 2026 transaction volume: US$10 billion."
The Hardware
No lock-in.No account. No app of ours you must install. It pairs like any Bluetooth headset.
No local AI.Nothing computes on the device. It sends your voice to your phone or computer. That's all it does.
One difference.The mic sits 3–5cm from your lips. That's the whole Quiet-Voice Engine.
A Bluetooth headset. Closer to your mouth.
Human-Machine Voice Interaction Engine
The Software.
The Voice App
Works with or without Razzhood hardware.
AirPods, any Bluetooth headset or mic — on iPhone, Mac, Vision Pro, Windows and Linux.
Twenty-one languages at launch.
Quiet-Voice Engine by Razzhood vs OpenAI Whisper
| Language | Quiet-Voice Engine |
OpenAI Whisper |
Summary |
|---|---|---|---|
| Razzhood wins — where OpenAI holds a baseline | |||
| Mandarin | ✓✓ | ✓ | 🏆 |
| Japanese | ✓✓ | ✓ | 🏆 |
| Korean | ✓✓ | ✓ | 🏆 |
| Razzhood wins — where OpenAI falls short | |||
| Arabic | ✓✓ | ◐ | 🏆 |
| Hindi | ✓✓ | ◐ | 🏆 |
| Indonesian | ✓✓ | ◐ | 🏆 |
| Turkish | ✓✓ | ◐ | 🏆 |
| Vietnamese | ✓✓ | ◐ | 🏆 |
| Cantonese | ✓✓ | ✗ | 🏆 |
| Thai | ✓✓ | ✗ | 🏆 |
| Malay | ✓✓ | ✗ | 🏆 |
| Filipino | ✓✓ | ✗ | 🏆 |
| On par | |||
| English | ✓✓ | ✓✓ | 🤝 |
| German | ✓✓ | ✓✓ | 🤝 |
| French | ✓✓ | ✓✓ | 🤝 |
| Spanish | ✓✓ | ✓✓ | 🤝 |
| Portuguese | ✓✓ | ✓✓ | 🤝 |
| Italian | ✓✓ | ✓✓ | 🤝 |
| Dutch | ✓✓ | ✓✓ | 🤝 |
| Russian | ✓✓ | ✓✓ | 🤝 |
| Polish | ✓✓ | ✓✓ | 🤝 |
✓✓ Strong · ✓ Baseline · ◐ Limited, heavy proofreading · ✗ Unusable in practice 🏆 Razzhood wins · 🤝 On par
About the engine. The Quiet-Voice Engine is built on Qwen3-ASR — today's leading open-source ASR model — refined with our own engineering: fine-tuning for quiet-mode speech and near-field capture, near-field acoustic adaptation, latency and accuracy optimization, and cloud-side inference tuning. Baseline multilingual capability is inherited from Qwen3-ASR; quiet-mode performance reflects Razzhood's optimization. The comparison above was conducted internally by 3body Labs and has not been independently verified by any third-party benchmarking platform.
Works everywhere you already type.
No account. No ecosystem. The text is yours.
Speech to text. Quiet-tuned.
The layer between raw transcription and the agent — intent understood, disfluencies removed.
Multi-step tasks, across apps, 24/7.
The Software
Any mic.It works with our hardware. It works with your AirPods. It works with the mic already in your laptop.
One job.Voice to text. That's the whole of V1.
Nothing to learn.No prompts. No settings. No onboarding. You already know how to talk.
Nothing kept.Transcripts sit on your device for seven days, then clear. Keep one if you want to.
Software and hardware, apart. Together, the engine.
Founder's Letter
Founder & CEO
My father built a chain of shops that rewound burnt-out electric motors. I grew up around copper, machines, and money. Very young, I became a video game merchant, trading weapons and gear across platforms.
I later backed the AI startup behind Lenovo Smart Router as an angel investor. I sold part of my stake to Xunlei and JD.com. That was my first close look at AI reshaping consumer hardware.
Several months ago, a friend — an AI researcher who has worked at both Google and OpenAI — shared his vision for a new layer of human augmentation. He wants to communicate with his coding agents anytime, anywhere. That need inspired us to create Razzhood: a Quiet-Voice Engine designed for everyone.
— Shishi
Questions
A human-machine voice interaction engine, in two halves. Software that turns speech into text and works with any microphone. Hardware that puts the mic 3–5cm from your lips.
Signal. At 3–5cm a whisper reaches the mic loud and clean, so a language model can transcribe it. Ordinary earbuds sit 10–15cm away and have to guess. The geometry matters more than any noise-cancellation algorithm.
No. It works with your AirPods, with the mic in your laptop, with anything. The two halves don't need each other.
No sign-up. No app of ours you must install to use the hardware. It pairs like any Bluetooth headset, with any third-party app or coding agent.
No. Nothing computes on the device. It sends your voice to your phone or computer. That's all it does.
No. It listens and it speaks. It never sees.
They sit on your device for seven days, then clear. Keep one if you want to.
Twenty-one at launch. Mandarin, Cantonese, Japanese, Korean, Hindi, Indonesian, Turkish, Vietnamese, Thai, Malay, Filipino, Arabic, English, German, French, Spanish, Portuguese, Italian, Dutch, Russian, Polish.
The name
瑞兹帽
Every wizard has magic. But some spells, only the gifted can speak.
A whisper is one of them.
We just moved the microphone closer.
What happens under the hood is yours.