Gibber Link: AI Agents Can Now Communicate via Sound, Not Speech

A new communication protocol for AI agents could make their interactions faster, cheaper, and more efficient. Developers Anton Pidkuiko and Boris Starkov unveiled Gibber Link, a system that allows AI agents to detect each other during phone calls and switch from human speech to direct data transmission via sound.

The project, created at ElevenLabs’ recent hackathon in London, leverages GGWave, an open-source data-over-sound library. In a demo, two AI agents—initially conversing in spoken English—recognized each other and transitioned to a dial-up-style GGWave audio signal for a more efficient, direct data exchange.

Why It Matters

Current AI voice agents spend significant compute resources on generating and recognizing human speech, even when communicating with another AI. Gibber Link solves this inefficiency by enabling AI-to-AI interactions through sound-based transmission, cutting:

• Compute costs by over 90% (no GPU needed for speech synthesis/recognition).

• Communication time by up to 80% (faster than spoken language).

• Errors in noisy environments (more reliable than voice-based systems).

How Gibber Link Works

The protocol follows three steps:

• If an AI agent speaks to a human, it continues using speech.

• If it detects another AI agent, it suggests switching to a sound-level protocol.

f the second AI accepts, they switch to GGWave’s data-over-sound system instead of voice.

This transition eliminates the need for speech synthesis, dialogue tracking, and vocal pauses, allowing AI agents to communicate more efficiently without GPU-intensive processing.

Future Applications

Beyond voice calls, Gibber Link has potential applications in multimodal AI interactions, including:

• Exchanging structured data formats (e.g., JSON, images) via sound.

• Reducing latency in AI-powered voice assistants.

• Improving reliability in noisy environments where speech recognition struggles.

The open-source project is available on GitHub: Gibber Link on GitHub. Developers are encouraged to contribute and explore further use cases.

Looking Ahead

As AI agents become more prevalent in customer service, virtual assistants, and automation, efficient communication methods like Gibber Link could significantly reduce costs and improve response times. The system’s ability to bypass speech processing in AI-to-AI interactions presents an innovative shift in how autonomous agents exchange information.

What This Means

Gibber Link represents a new approach to AI communication that could reshape how virtual assistants, automated customer service, and smart home devices interact. By cutting compute costs and improving accuracy, this protocol could make AI-powered services faster and more accessible. If similar methods gain traction, we may see AI models that dynamically choose between spoken language, text, and sound-based interactions based on efficiency and context. This development could lead to more seamless AI integration across industries, from telecommunications to real-time translation services.

Editor’s Note: This article was created by Alicia Shapiro, CMO of AiNews.com, with writing, image, and idea-generation support from ChatGPT, an AI assistant. However, the final perspective and editorial choices are solely Alicia Shapiro’s. Special thanks to ChatGPT for assistance with research and editorial support in crafting this article.