AI 分析
痛点：多说话人音频（财报电话会议、长会议、多人播客）的转录服务要么丢失说话人标签，要么把说话人分离（diarization）作为额外收费步骤，且时间戳对不齐。
目标用户：需要处理财报电话会议、长会议、多人播客转录的开发者、分析师、内容团队。
证据：The thing I kept running into with off-the-shelf providers was that on multi-speaker audio — earnings calls, long meetings, podcasts with several hosts — the transcripts either dropped speaker labels entirely or charged a premium for diarization as a separate step. Most providers I looked at were $0.12–$0.22/hour. No subscription. I started with $0.50 — roughly 6h 15m of audio.
变现：按音频小时预付费，无订阅，提供API调用额度包。
MVP：用Whisper做转录、pyannote做说话人分离，封装一个接收公开音频URL、返回带内联说话人标签和时间戳的REST API，支持Google登录和预付费额度，2周可上线基础版。
竞品：AssemblyAI、Deepgram、OpenAI Whisper API、Rev、Otter.ai、Fireflies.ai
摘要：作者开发JOBE Vox STT API，解决多说话人音频转录中说话人标签丢失或diarization额外收费的痛点，提供一次POST返回谁在何时说了什么。他提到竞品价格约$0.12–$0.22/小时，自己定价$0.50起步（约6小时15分钟音频，合$0.08/小时），无订阅。该帖是产品自荐，未展示第三方付费意愿或收入数据。