Cloudflare

Deepgram's low-latency STT and TTS models are served via Cloudflare Workers AI, with inference running in more than 300 edge locations worldwide. Pairing Deepgram's low-latency audio models with Cloudflare's edge network gives developers real-time voice AI without fighting cold starts or regional slowness.

The pattern: build end-to-end voice agents on a single platform by capturing audio, streaming to Deepgram via WebSockets, orchestrating logic and LLMs in Workers, and serving results at the edge. The CDN, serverless compute, and transcription endpoints all live inside one platform instead of being stitched together across vendors.

Every interaction benefits from Cloudflare's built-in TLS termination, DDoS protection, and caching for secure, reliable performance.

If you are building voice features on Cloudflare and want a production speech provider, Deepgram's models are reachable directly from Workers. The partnership announcement below has the technical details.

• Why the Deepgram + Cloudflare tie-up actually solves real problems
• Cloudflare partner page
• Cloudflare Workers documentation
• Deepgram documentation
• Contact Deepgram