MeetFree Technology
MeetFree 用一条端到端的流式处理链路，把现场语音实时转成双语字幕：语音识别与翻译一体完成，边说边译、低延迟，并可通过云端接口嵌入智能音箱、会议设备等第三方硬件。MeetFree turns live speech into bilingual captions through an end-to-end streaming pipeline: speech recognition and translation run together, sentence by sentence, with low latency — and can be embedded into third-party hardware such as smart speakers and meeting devices via a cloud API.
技术架构Architecture
音频从设备麦克风采集，经 WebSocket 实时上行，云端依次完成流式语音识别与实时机器翻译，译文随说随返回。双方无需安装任何东西，一台设备完成全部工作。Audio is captured on the device and streamed over a WebSocket; the cloud runs streaming speech recognition and real-time machine translation in sequence, returning the translation as you speak. Only one device is needed.
设备麦克风采集，前端语音活动检测（VAD）、自动增益与降噪。On-device mic capture with VAD, auto gain and noise handling.
WebSocket 双向长连接，音频帧流式上行，断线自动重连。Persistent WebSocket streams audio frames, with auto-reconnect.
实时语音转文字，结合上下文与术语表，边说边出结果。Real-time speech-to-text using context and glossary.
识别与翻译一体处理，双语译文逐句返回并显示为字幕。Recognition and translation run together; captions return sentence by sentence.
关键指标Key metrics
以下为面向实时商务对话的设计指标。具体延迟与识别表现会随网络、设备和环境噪音变化，建议在真实场景中实测校准。These are the design targets for live business conversation. Actual latency and recognition vary with network, device and ambient noise; we recommend validating in your own environment.
核心能力Core capabilities
采用实时流式语音识别模型，结果边说边返回，并结合上下文理解，减少短句误判。Real-time streaming ASR returns results as you speak and uses context to reduce short-utterance errors.
语音识别与翻译一体化处理，减少中间环节，降低整体延迟，原文与译文逐句对应。Recognition and translation run together to cut intermediate steps, pairing source and target sentence by sentence.
可定制行业术语、产品名、专有名词和数字表达，提升专业场景下的识别与翻译准确度。Custom industry terms, product names and numbers improve accuracy in professional settings.
前端语音活动检测（VAD）、能量阈值过滤、自动增益控制，并提供高噪音环境下的展会模式。On-device VAD, energy thresholds, auto gain and a high-noise exhibition mode.
会谈中的数字自动高亮，并对可疑数字进行标记，帮助核对价格、数量等关键信息。Numbers are highlighted and suspicious figures flagged to verify prices and quantities.
会谈结束可一键生成多语言 AI 纪要；手机、平板、电脑浏览器均可运行，无需安装。One tap generates multilingual AI summaries; it runs on phones, tablets and desktop browsers without install.
硬件集成方案Hardware integration
MeetFree 的实时翻译能力可通过云端实时接口嵌入第三方硬件。硬件厂商无需自建 AI 模型与算力，即可让设备具备多语言实时对话能力。MeetFree's real-time translation can be embedded into third-party hardware via a cloud API. Manufacturers add multilingual live conversation without building their own AI models or compute.
会议室、前台、酒店客房的音箱，直接开口即可跨语言对话。Speakers in meeting rooms, front desks and hotel rooms enable hands-free cross-language talk.
多语言会议终端、一体机、前台接待设备，实时显示双语字幕。Multilingual meeting terminals and kiosks display live bilingual captions.
手持翻译机、翻译耳机等可穿戴设备，获得持续升级的翻译能力。Handheld translators and earbuds gain translation that keeps improving.
开源代码Open source
MeetFree v1.5.0 核心版本采用 Fair Code 协议开源。您可以自由查看、自行部署与内部修改；不得公开再分发修改后的源码，亦不得基于本代码提供竞争性翻译 SaaS 服务。MeetFree v1.5.0 core is open source under the Fair Code license. You may view, self-host and modify internally; you may not publicly redistribute modified source or offer a competing translation SaaS.
代码完整公开，可下载后在自己的服务器上部署运行，填入您自己的 API 密钥即可使用。Full source is public; download and run on your own server with your own API keys.
硬件合作方可基于开源代码评估集成方案，进行内部测试与定制开发，无需担心技术黑盒。Hardware partners can evaluate integration, run internal tests and build custom versions — no black box.
允许个人与内部使用；禁止公开再分发修改版、禁止基于本代码提供竞争性翻译服务。Personal and internal use allowed; no public redistribution of modifications, no competing translation service.
典型应用场景Typical scenarios
跨语言会议、客户来访接待，设备直接承担实时口译。Cross-language meetings and visitor reception handled by the device.
酒店前台、展会服务台等多语言服务场景，降低人工成本。Multilingual service at hotel and exhibition counters.
随身翻译设备帮助海外商务人士应对现场沟通。Portable devices help overseas business people communicate on site.
面向多国客户的服务终端与自助设备，统一翻译能力。Service terminals and kiosks with unified translation.
合作流程How we work together
明确设备形态、语言、使用场景与网络环境。Device form, languages, scenarios and network.
评估音频链路、延迟目标与接口方案。Audio path, latency targets and API approach.
提供实时接口与文档，完成原型验证。Real-time API and docs for a working prototype.
在目标硬件与真实环境中联调，校准效果。Joint testing on target hardware and real settings.
完成集成，云端模型持续升级维护。Complete integration with ongoing cloud upgrades.
无论是智能音箱、会议设备还是翻译硬件，我们都可以根据你的设备形态与场景，评估实时翻译能力的集成方案。For smart speakers, meeting devices or translation hardware, we can assess an integration plan based on your device and scenario.
邮件沟通技术方案Email us about integration [email protected]