OpenAI Shipped Your Voice Stack at $0.25/Min. Vapi Went Enterprise. The Infra Layer Abandoned Agencies in Eleven Days.

Short summary

OpenAI's gpt-realtime-2 ($0.25–0.35/min) unified speech-to-speech reasoning, collapsing the wrapper orchestration layer. Infrastructure platforms (Vapi, Synthflow) abandoned agencies and moved to enterprise customers ($500M+ valuations). Agencies now face three paths: DIY with Pipecat ($40–60K dev cost), overpay legacy wrappers at 7–10x price hikes, or adopt application-layer platforms designed for agencies.

• •OpenAI gpt-realtime-2 unified speech-to-speech reasoning in one API call, eliminating the need for stitched STT+LLM+TTS infrastructure that wrapper platforms profited from
• •Infrastructure layer (Vapi, Synthflow, Retell) pivoted upmarket to enterprise (Amazon Ring, $230M BPOs, Zillow, Deutsche Telekom), abandoning the 3–15 client agency segment
• •Agency operators face three paths forward: DIY on Pipecat v1.0 open source, stay with legacy wrappers accepting 7–10x price increases, or migrate to agency-focused application platforms

Generated with AI, which can make mistakes.