EazyRecording
LIVEEazyRecording 的目標不是錄下會議，而是讓每一場會議結束後，自動變成團隊知識庫的一部分。好的會議工具應該讓你忘了錄音這件事，但不會讓團隊忘記會議裡的決定、責任與脈絡。EazyRecording is not about recording meetings. Its goal is to make every meeting automatically become part of the team knowledge base. A good meeting tool should let you forget about recording, without letting the team forget the decisions, responsibilities, and context from the meeting.
這個專案的切法，是把 AI 從「摘要工具」往真正的工作系統推進：它不只自動錄音與產生 meeting note，還把 template、PM 修改回饋、團隊知識庫與工作流同步串起來，讓會議資料會被累積、改善、再利用。The angle here is moving AI beyond a summarizer and into a real work system: not just automatic recording and meeting notes, but templates, PM feedback, team knowledge capture, and workflow sync that make meeting data accumulate, improve, and get reused.
為什麼我做這個Why I built this
我做 EazyRecording，是因為管理團隊後越來越有感：會議最可惜的地方不是花了多少時間，而是會議裡產生的知識沒有留下來。決策、客戶需求、技術風險、責任分工和下一步行動，如果沒有自動進入團隊知識庫，過幾天就會變成某個人腦中的模糊記憶。講完就散，散了就要重講；重講，就是管理成本。I built EazyRecording because, after managing teams, I kept seeing the same problem: the most wasteful part of meetings is not the time spent in them, but the knowledge that disappears afterwards. Decisions, customer needs, technical risks, ownership, and next steps become vague memories unless they automatically enter the team knowledge base. When knowledge disperses, teams repeat themselves; repetition is management cost.
它解決什麼問題The problem it solves
多數 AI 會議工具只解決「產生一份摘要」；EazyRecording 想解的是更後面的斷點：人會忘記按錄音、錄音檔沒人重聽、AI 摘要每次都像從零開始寫、PM 還要手動修改，最後會議內容仍然沒有進到團隊真正工作的地方。Most AI meeting tools only solve the problem of generating a summary. EazyRecording goes after the next bottleneck: people forget to record, recordings are rarely replayed, AI summaries start from scratch every time, PMs still have to edit the notes manually, and the final meeting knowledge often never reaches the systems where the team actually works.
產品做了什麼What the product does
全自動會議錄音Fully automatic meeting recording
設計方向是讓會議開始時自動啟動錄音、背景安靜記錄，讓使用者忘了工具存在，但不會忘記會議內容。Designed to start recording automatically when a meeting begins and run quietly in the background, so users can forget the tool exists without losing meeting memory.
系統音訊 + 麥克風System audio + microphone
用 ScreenCaptureKit 同時錄系統音訊與麥克風，完整留下線上會議雙方聲音，並輸出適合會議保存的 AAC 音訊。Uses ScreenCaptureKit to capture system audio and microphone together, preserving both sides of an online meeting and exporting meeting-ready AAC audio.
即時逐字稿Live transcription
AssemblyAI streaming 讓會議進行中就產生逐字稿，支援繁中、英文、自動偵測與簡轉繁，讓內容在發生當下就變成可被理解的資料。AssemblyAI streaming creates live transcripts during the meeting, with Traditional Chinese, English, auto-detection, and Simplified-to-Traditional conversion.
會議紀錄 TemplateMeeting note templates
不同會議可以套不同 meeting note template：產品會議看需求與優先順序，工程會議看風險與 blocker，客戶會議看痛點、承諾與 follow-up。Different meetings can use different note templates: product meetings focus on requirements and priorities, engineering meetings on risks and blockers, customer meetings on pains, commitments, and follow-up.
PM Feedback LoopPM feedback loop
PM 修改最終 meeting note 後，系統根據修改結果改善下次輸出，逐步學會這個團隊怎麼寫決策、action items、風險與脈絡。After the PM edits the final meeting note, the system uses those edits to improve the next output and learn how the team writes decisions, action items, risks, and context.
自動進團隊知識庫Straight into the team knowledge base
會議結束後，自動整理摘要、決策、責任、風險、待追蹤事項與知識點，讓會議不只開完，而是真的被組織吸收。After the meeting, it organizes summaries, decisions, owners, risks, follow-ups, and knowledge points so the meeting is absorbed by the organization instead of merely ending.
AI / Agent 整合AI / Agent integrations
Live AI ChatLive AI chat
錄音進行中可以直接問 AI「剛剛討論了什麼？」、「目前有哪些 action items？」或「哪些地方還沒有結論？」讓 AI 使用當下逐字稿回答。During recording, users can ask AI what was discussed, which action items exist, or what is still unresolved, using the live transcript as context.
Meeting-to-Knowledge LayerMeeting-to-knowledge layer
AI 不只產生摘要，而是把會議拆成決策、任務、風險、背景脈絡與知識點，準備進入團隊知識庫與專案系統。AI does not merely summarize; it turns meetings into decisions, tasks, risks, context, and knowledge points ready for the team knowledge base and project systems.
從自動記錄走向 AI DelegateFrom auto-capture toward an AI delegate
長期目標是讓 AI 先幫我聽會議、整理立場、追問缺口，最後只把真正需要我判斷的部分交回來。The long-term goal is for AI to listen to meetings, organize positions, ask missing questions, and eventually bring back only the parts that truly need my judgment.
雙 STT Provider + Streaming SSEDual STT providers + streaming SSE
支援 AssemblyAI 即時串流與 OpenAI Whisper 會後轉錄設計；AI Chat 採 streaming SSE，讓回應不必等完整答案才出現。Supports AssemblyAI live streaming and OpenAI Whisper post-meeting transcription; AI chat uses streaming SSE so responses appear progressively.
品質不是口號Quality signals
Crash-safe recordingCrash-safe recording
先錄 PCM CAF 以降低 app 意外退出造成的音訊損失，再轉 AAC M4A。錄音工具最基本的信任是：你按下錄音，它不應該輕易弄丟內容。Records PCM CAF first for crash safety, then converts to AAC M4A. The basic trust of a recording tool is simple: once recording starts, it should not easily lose the content.
不是固定 prompt，而是 feedback loopA feedback loop, not a fixed prompt
PM 最終修改會回流成下次產出的參考，讓 AI 不是每次從零開始寫，而是越來越貼近團隊偏好的筆記格式。Final PM edits flow back into the next generation, so AI does not start from scratch every time and gradually matches the team’s preferred note style.
語言與中文細節Language and Chinese-text details
支援繁中、英文、自動偵測與簡轉繁，這是實際使用台灣會議時會遇到的品質細節。Supports Traditional Chinese, English, auto-detection, and Simplified-to-Traditional conversion — practical quality details for Taiwanese meetings.
會議輸出不留在 app 裡Output does not stay inside the app
設計包含 GitHub / 工作流同步，讓 action items、決策與專案脈絡進入真正執行工作的地方。The design includes GitHub and workflow sync so action items, decisions, and project context move into the places where work actually happens.