Unverified85% confidenceFactTime unknown
传统语音处理管道采用ASR→LLM→TTS三段式架构，而OpenAI Realtime API采用端到端音频处理方式
1
Sources
85%
Confidence
Long-term
Relevance
5/31/2026
First Seen
Sources
OpenAI实时语音Demo活动：评选标准、奖励机制与开发者机会
twitterOpenAIDevs
Related Entities
Related Claims
Unverified现代实时语音系统通常采用流式ASR、流式LLM输出和流式TTS三者并行的架构，而非串行等待每步完成79% similarUnverified端到端语音架构的代表性工作包括Google的AudioLM、Meta的Voicebox以及OpenAI Realtime API79% similarVerified传统语音AI普遍采用级联式架构，先用ASR将语音转文字，再交给语言模型处理，最后用TTS合成输出77% similarUnverifiedOpenAI的GPT-Live语音模型采用端到端架构，原始音频波形直接输入模型，模型直接输出音频，中间不再经过独立的ASR和TTS模块76% similarUnverifiedOpenAI新模型采用端到端语音架构，直接以音频作为输入和输出，跳过中间文字转换环节74% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/13033API
curl https://kongchang.com/api/v1/knowledge/claims/13033MCP
get_claim(id=13033)