We've tested most text-to-speech models and ElevenLabs' models are some of our favorites. Flash is ElevenLabs' fastest model, and offers a broad range of voices and 30+ languages — ideal for real-time voice agents and available to all Layercode users.
Report
6 views
Latency is everything when building voice AI. From day one, We chose to build our entire platform on top of Cloudflare. The global nature of Cloudflare’s network ensures minimal latency to every user, wherever they are.
We chose Cloudflare Workers to power the core of our voice AI pipelines. This ensures our users experience zero cold starts and reliable performance when handling voice data. In addition, Durable Objects and DO SQL provides the consistent execution and storage environment required for our hosted voice AI agents and many internal features.
Report
14 views
Nailing accurate speech-to-text transcription is the first — crucial — step to any agent powered by a voice pipeline.
We've tested a lot of STT models and found Deepgram to consistently beat the alternatives when it comes to speed and accuracy. We're excited to be able to offer their speech-to-text transcription to all users building voice agents with Layercode.
Report
13 views