[{"data":1,"prerenderedAt":4},["ShallowReactive",2],{"article-doc:docs\u002Fconcepts\u002Fstablestream":3},"---\ntitle: StableStream\ndescription: LansonAI's approach to maintaining readable continuity while live speech continues to evolve.\n---\nStableStream is LansonAI's approach to maintaining readable continuity while live speech continues to evolve.\n\nReal-time speech recognition produces changing information.\n\nA naive interface can expose those changes directly:\n\n```text\nwe should probably\nwe should probably ship\nwe should probably ship the\nwe should probably ship this\nwe should probably ship this next\nwe should probably ship this next week\n```\n\nIf earlier words are repeatedly replaced or rearranged, the interface may technically be updating quickly while becoming difficult to read.\n\nStableStream addresses this boundary between **model updates** and **human-readable output**.\n\n## Ready to read, not racing to display\n\nThe fastest possible token is not always the most useful token.\n\nA live interface must balance:\n\n- response latency\n- linguistic uncertainty\n- correction quality\n- visual stability\n- reading continuity\n\nStableStream is designed around that balance.\n\nIts goal is to allow new information to arrive continuously without making previously presented information unnecessarily unstable.\n\n## Recognition stability and visual stability\n\nThese are related but different concepts.\n\n**Recognition stability** describes how confident the system is that speech has been interpreted correctly.\n\n**Visual stability** describes how much already-presented content moves or changes on screen.\n\nA good live speech experience needs both.\n\nStableStream operates at this boundary.\n\n## Application behavior\n\nApplications consuming a live stream should distinguish between content that is still evolving and content that has become stable.\n\nThe API documentation describes the exact events and state transitions available to clients.\n\nThis page describes the underlying principle:\n\n> Live text should evolve without forcing the reader to repeatedly reconstruct what they have already understood.\n",1790059118950]