Meeting Minutes: From Recording and Transcription to Structured Minutes in One Go
💡 One-sentence summary: Use WorkBuddy to build a pipeline from recording to deliverable AI meeting minutes — automatic recording transcription, one-click structuring into five elements with the official prompt, and export ready to send to the group.
Most results for searching "AI meeting minutes" are transcription tools and template samples, but what really eats your time is never "turning the recording into words"—it's the step after transcription: fishing conclusions out of an hour-long play-by-play, matching action items, and chasing down owners. This lesson breaks the WorkBuddy minutes scenario into one pipeline: hand the recording over for transcription (existing notes or group chat discussions can also be thrown in directly), use the official minutes prompt to organize it into structured minutes in one click, refine through multiple rounds, then export and deliver—you only handle verification and forwarding.
The weekly report materials summarized in the previous lesson A1 Weekly Report Auto-Generation already included meeting minutes—this lesson turns that into a pipeline too. This is lesson 2 of Route 1 in "Scenario Practice," based on WorkBuddy v5.9.3, with content drawn from the official practice case "Organize Meeting Minutes," the feature description "Audio to Text," and the zero-cost skill "Local Whisper."
- For complete instructions on drag-and-drop upload, voice input, and chat record forwarding, see 2.3 Input Methods Explained; this lesson only covers how to use them in the minutes scenario
- Meeting held in Tencent Meeting? See C5 Tencent Meeting Management for how to retrieve recordings, transcripts, and smart minutes
What You'll Be Able to Do After This Lesson
- Hand a meeting recording file to WorkBuddy and get back a transcript with time ranges (single file ≤1GB, duration ≤5 hours, about 0.6 credits/minute)
- Use the official minutes prompt to organize transcripts, meeting notes, or group chat records into five-element structured minutes: meeting topic, key conclusions, action items, owners, deadlines
- Refine through multiple rounds until it's deliverable: specify the output format (Markdown, Word, table), correct misidentified names, compress the discussion into conclusions
- Send the minutes out: keep them in the workspace, upload to the cloud as a Tencent Docs smart document, or share to WeChat/WeCom
Why WorkBuddy Instead of Web-Based AI?
| Capability | Web-Based AI | WorkBuddy |
|---|---|---|
| Recording to text | ❌ An hour-long recording mostly can't be processed directly; you have to find another transcription tool first | ✅ Upload audio directly for transcription, returning text with time ranges, single file ≤1GB, ≤5 hours |
| Getting meeting materials in | ❌ Copy-paste piece by piece; long texts need to be chunked yourself | ✅ Drag documents in, long pasted text collapses automatically, WeChat chat records forwarded directly as attachments |
| Producing deliverable files | ❌ A block of text in the chat box; formatting is on you | ✅ Generates a minutes artifact that can be uploaded to the cloud or shared to WeChat/WeCom |
| Handling sensitive recordings | ❌ The whole recording gets uploaded, and it's unclear who it goes to | ✅ File processing is done locally by default; paid transcription shows a confirmation before upload, and sensitive meetings can use Local Whisper locally offline |
| Connecting to meeting tools | ❌ Export the recording yourself, upload the file yourself | ✅ Tencent Meeting users can check recordings, transcripts, and smart minutes with a single sentence |
Your Current Struggles
- An hour-long meeting with the recording in hand, yet you spend half an hour finding a transcription tool and another half hour fishing conclusions out of the play-by-play
- Taking notes while in the meeting, catching one sentence and missing the next, only to find after it ends that the most critical decision was never written down
- Conclusions scattered across the group: a half-hour meeting followed by two hours of group chat, scrolling through chat records until your eyes blur and still can't piece it together
- Finishing the minutes isn't the end: formatting into a document, uploading to the shared drive, @-mentioning every owner for confirmation—the whole routine is more exhausting than the meeting itself
When to Use This Technique
When you need to:
- Organize meeting recordings, chat records, or scattered notes into formal minutes—these are exactly the three material types listed in the official practice case
- Quickly produce actionable minutes after the meeting ends: action items with owners and deadlines, not just "the meeting concluded successfully"
- Use AI to organize privacy-sensitive internal meetings too, without carelessly uploading the recording
And you don't want to:
- Manually listen to the recording and type out a transcript
- Repeatedly copy-paste long texts into a web-based AI chat box
- Re-describe the minutes format from scratch every time
🎒 Preparation Before Starting
- Completed 1.4 Your First Task: From One Sentence to an Acceptable Artifact, know how to create tasks and find artifacts in the results area
- Have materials to organize on hand, pick one of three: a meeting recording file; meeting notes as text or a document (TXT, Markdown, Word, PDF); or chat records from a WeChat/WeCom group chat
- Have credits available when using paid audio-to-text: billed at about 0.6 credits/minute based on actual audio duration, and the first call in a free-model session shows a confirmation prompt (the trial version includes 500 base credits per month)
- For zero-credit transcription or sensitive materials: install the Local Whisper skill first (install name local-whisper); see D2 Zero-Cost Skills Selection for the full list of ten zero-cost skills
- 15 minutes; ideally, follow along with a real meeting you recently attended
Follow Along with Me
Step 1: Hand the Meeting Materials to WorkBuddy (Recording, Notes, or Chat Records)
Why: The ceiling on minutes quality depends on how complete the materials are. The official practice case groups minutes materials into three types—meeting recordings, chat records, and scattered notes. Step one is to gather what you have into the same task.
How to proceed (pick one of three based on your materials):
Option A: You Have the Recording
Drag audio files directly anywhere on the page (no need to aim at the input box); release to add them. You can drag up to 50 files.
Method B: You already have a text record
Paste directly into the input box, or drag in TXT, Markdown, Word, or PDF files. Pasted content over 3000 characters is automatically collapsed into an expandable entry—click to open and keep editing, so no need to worry about flooding the screen.
Method C: Conclusions scattered in a group chat
In WeChat or WeCom, forward that chat record to WorkBuddy; the client is launched automatically, a chat record attachment is generated in the input box, and it is processed together with the task.
You should see: Materials appear as cards in the attachment area of the input box, showing file names and status; nothing is sent automatically after adding, and you can keep adding text notes.
- My meeting materials are sitting in the input box as attachments or text, and I haven't hit send
Step 2: Have WorkBuddy turn the recording into text (audio-to-text)
Why: The official guidance is clear that meeting minutes are primarily text-based, so the raw recording must first be converted to text. WorkBuddy has built-in audio-to-text capability—no need to transcribe elsewhere and upload again; you can do it all in one task.
How to do it (after handing over the recording file, say in the input box):
Turn this recording into text.You can also paste an audio link instead of uploading a file. To get the key points in one step, use the other two phrasings from the official examples:
Summarize the core content of this meeting recording.Extract action items from the recording, and organize owners and deadlines.You should see: Transcription runs automatically in the background—no need to manually check progress; when done, it returns transcribed text with time ranges, so you can trace any conclusion back to which minute of the recording at a glance.
💡 This is the unique value of WorkBuddy
Audio-to-text is a paid capability, billed by actual audio duration at about 0.6 credits/minute; the first call in a free-model session pops up a confirmation prompt where you can choose to confirm, always allow for this session, or decline—only after confirmation is the audio actually uploaded, and declining cancels this transcription. Don't want to spend credits, or dealing with sensitive internal meetings? Install the Local Whisper skill (install name: local-whisper); once the model finishes downloading, transcription runs fully local and offline, and the recording never leaves your machine. See 2.6 Credits and Pricing for credit rules.
Formats and limits (check before transcribing):
| Item | Details |
|---|---|
| Audio formats | wav, mp3, m4a, flv, mp4, wma, 3gp, amr, aac, ogg-opus (.opus or .ogg), flac |
| File limits | Single file ≤1GB; audio duration ≤5 hours |
| Supported languages | Chinese, English, Cantonese, and some dialects |
- I have the transcribed text with time ranges, and spot checks match the recording
Step 3: Generate a five-element AI meeting minutes with the official prompt
Why: The gap between a raw transcript and minutes you can hand off is just a clear structure. This prompt from the official practice case covers the five elements and can be used directly as a template.
How to do it (the transcribed text, pasted records, or chat record attachment are all in the task—just send it):
Based on the meeting record I provided, organize it into formal meeting minutes, including the meeting topic, key conclusions, action items, owners, and deadlines.This is the original sample instruction from the official "Organize Meeting Minutes" case. If the material is freshly transcribed text, it matches up on its own; if the materials are multiple files, name which one to use in the instruction.
You should see: A structured set of minutes with all five elements—meeting topic, key conclusions, action items, owners, and deadlines—each conclusion traceable, no longer a meeting play-by-play.
- My minutes have all five elements, and every action item lists an owner and a deadline
Step 4: Polish the minutes over multiple rounds—specify the format, fix errors and omissions
Why: The first version often has misrecognized names and uneven levels of detail. The official usage advice is clear: first specify the output format (Markdown, Word, table, or plain text), then add constraints (such as the minutes structure); adjusting over multiple rounds is faster than nailing it in one shot.
How to do it (append as needed, like a chat):
Output as a Word document; organize the action items into a separate table: item, owner, deadline, notes.The transcription misheard 「Wang Qian」 as 「Wang Qin」—please correct all of them; keep only conclusions from the discussion, no need to repeat it sentence by sentence.You should see: The minutes are rearranged or exported as a document per your request; each revision gets closer to a version you can post directly to the group. If you over-edit, you can roll back to a previous round and start again—see 2.1 Conversation Advanced.
- I polished it at least one round: the format is specified and misrecognized names are corrected
Step 5: Export and deliver—save, upload to the cloud, share with attendees
Why: Finishing the minutes isn't the end—only when the right people see them and the right people claim their action items is it done. Deliverables are saved by default in the current session or your designated workspace, so they won't be lost; if collaboration is needed, send them out.
How to do it (pick one of three, or combine):
| Delivery method | How | Best for |
|---|---|---|
| Keep it local | Nothing to do—the minutes are already saved in the current session or the designated workspace | Archiving for yourself, editing later |
| Upload to the cloud | Upload the minutes deliverable to the cloud from the results area; Markdown is created by default as a smart document in Tencent Docs | Collaborative team editing, claiming action items via comments |
| Share to WeChat/WeCom | Share the deliverable with attendees from the results area | Posting directly to the meeting group |
The four tabs of the deliverables area (Deliverables, Web, Workspace Files, Changes) are fully explained in 1.4 Your First Task. For ways to use collection forms and comment summaries after the minutes are stored in Tencent Docs, see C3 Tencent Docs Connector.
You should see: Attendees receive the minutes file or an online document link in the group, and action item owners each claim their items—only at this point is the whole pipeline complete.
- The minutes have been delivered per my choice: the file is in the workspace, or attendees have received the link/file
📋 Magic Prompt
How do you write a meeting minutes template without missing anything? Lock down the five-element structure, then let the AI fill in the real substance. The three below cover the three use cases: "recording in one step," "materials already available," and "action items only." Placeholders are marked with 【】—swap in yours and they're ready to run.
📝 Recording to Minutes in One Click (all-in-one version)
Expected result: after uploading the recording, one instruction completes transcription + minutes—suited for those who don't want two steps.
## Role
You are a meeting minutes specialist, skilled at distilling conclusions and to-dos from recordings.
## Task
Turn this meeting recording into text and organize it into formal meeting minutes.
## Input Information
### Required Fields
- Meeting recording: 【dragged-in audio file, or audio link】
- Meeting topic: 【e.g., Q4 channel review meeting】
### Optional Fields
- Attendee list: 【e.g., Wang Qian, Li Nan, Zhang Wei?】
- Topics of special focus: 【e.g., budget allocation?】
## Output Specification
- Give the full transcribed text first (with time ranges), then the minutes
- The minutes include: meeting topic, key conclusions, action items, owners, deadlines
## Constraints
- ✅ Every action item must state an owner and a deadline; mark anything not mentioned in the recording as 「To be confirmed」
- ✅ Correct names against the attendee list
- ❌ Do not repeat the discussion process sentence by sentence
- ❌ Do not fabricate conclusions not present in the recording📝 Formal Meeting Minutes Generator (materials-already-available version)
Expected result: the complete version of the official practice case prompt—transcribed text, chat records, and scattered notes can all be fed in directly.
## Role
You are my office assistant, skilled at organizing scattered meeting records into formal minutes.
## Task
Based on the meeting record I provided, organize it into formal meeting minutes, including the meeting topic,
key conclusions, action items, owners, and deadlines.
## Input Information
### Required Fields
- Meeting notes: [pasted transcription, group chat discussion, or dragged-in TXT/Markdown/Word/PDF]
- Meeting topic: [e.g., product pricing review meeting]
### Optional Fields
- Output format: [Markdown / Word / table?]
- Minutes structure: [e.g., grouped by agenda item?]
## Output Specification
- All five elements included: meeting topic, key conclusions, action items, owners, deadlines
- Number key conclusions one by one; list all action items together at the end
## Constraints
- ✅ Use only information that appears in the notes; for anything not mentioned, write "not clarified in the meeting"
- ✅ Rewrite spoken phrasing into written language while preserving the original meaning
- ❌ Do not add your own opinions or suggestions
- ❌ Do not write boilerplate like "the meeting went very well"📝 Action Item Tracking Table (Post-Meeting Follow-Up Edition)
Expected result: no need to reread the full text—one instruction pulls out all action items and produces a to-do table you can post directly to the group chat, giving you grounds to chase progress.
## Role
You are a project assistant responsible for post-meeting follow-up on to-dos.
## Task
Extract all action items from these meeting notes and organize them into a trackable table.
## Input Information
### Required Fields
- Meeting notes: [transcription or meeting minutes]
### Optional Fields
- Owner mapping: [e.g., "Wang Qian = Wang Qian from Marketing"?]
- Deadline convention: [e.g., uniformly set to next Friday?]
## Output Specification
- Table columns: Item | Owner | Deadline | Notes
- Sort by deadline from nearest to farthest
## Constraints
- ✅ Every action item has an owner and a deadline; mark missing ones as "to be confirmed"
- ✅ Write items as verifiable actions, e.g., "submit the final draft of the channel policy"
- ❌ Do not mix pure discussion items into action items
- ❌ Do not fabricate ownersCheckpoints ✅
- The transcription has time ranges; spot-checking a few segments matches the original recording
- The minutes include all five elements: meeting topic, key conclusions, action items, owners, deadlines
- Every action item has a clear owner and deadline, with no vague wording like "to follow up"
- I have double-checked key information such as names, amounts, and dates against the recording/notes
- The minutes have been exported as needed (Word/Markdown/table) and saved to the workspace or uploaded to the cloud
- When the free-session credits confirmation pops up, I understand the three options: confirming uploads the transcription; "always allow" stops asking for the rest of this session; declining cancels; if I don't want to spend credits, I switch to Local Whisper
Pitfall Reminders
❌ Tossing the recording in and just saying "help me organize this"
Symptom: you get back a vague summary, or it asks you what format you want.
Cause: the official example prompts all come with a clear task—transcription, summarization, and to-do extraction each have their own phrasing; "organize this" leaves it to guess.
Correct approach: first state clearly what you want (transcribe / produce minutes / extract to-dos), then add structural requirements. Use the magic prompt above directly, replacing the placeholders with yours.
❌ Poor audio quality, people talking over each other—the transcription is riddled with errors
Symptom: names get mixed up, terms become unrecognizable, and the minutes go wrong along with them.
Cause: per the official documentation, speech recognition accuracy is affected by audio quality, language, and accent.
Correct approach: use the meeting software's built-in recording whenever possible—it picks up audio more cleanly than recording externally with a phone; the official zero-cost skill page also advises "provide clear audio whenever possible." After transcription, read through it first and call out obvious errors by name for correction in the next round of instructions (see step 4).
❌ Transcription fails or times out, and you just wait or re-upload repeatedly
Symptom: you wait a long time with no result, or get a message about a file problem.
Cause: per the official documentation—invalid audio, upload failure, or unsupported formats produce the corresponding error message; a transcription timeout or failure returns a failure explanation and will not treat blank content as the final result.
Correct approach: check three things as prompted: the format is on the supported list (wav, mp3, m4a, etc.), a single file is ≤1GB, and duration is ≤5 hours. Long audio takes longer to process, so keep your device online and wait patiently; the failure explanation usually includes the cause, so don't re-upload blindly.
❌ Clicking confirm all the way through without reading the confirmation dialog for sensitive meeting recordings
Symptom: recordings of internal strategy meetings or confidential reviews get uploaded wholesale, and you feel uneasy afterward.
Cause: paid audio transcription requires uploading the audio for processing (uploaded only after confirmation). WorkBuddy file processing is done locally by default, but this transcription pipeline is a cloud capability.
Correct approach: the confirmation dialog is there for you to gatekeep—if you don't want to upload, choose decline; the transcription is canceled and the audio is not uploaded. For sensitive meetings, switch to the Local Whisper skill for local offline transcription; once the model is downloaded, it no longer depends on the online transcription API, and the recording never leaves your device.
Advanced Tips (Optional)
🔧 Tencent Meeting users: check recordings, transcriptions, and smart minutes directly. If the meeting was held in Tencent Meeting, you don't need to export the recording manually. Use the instructions from the official practice case directly: "Help me check my recent meeting recordings," "Help me look at the transcription of the last meeting," "Help me get the smart minutes for this meeting"—you can even "search for 'technical proposal' in the meeting transcription." For the full Token setup process, see C5 Tencent Meeting Management.
🔧 Archive minutes into Tencent Docs. After authorizing the Tencent Docs connector, minute outputs uploaded to the cloud are created as smart documents by default, so the team can collaborate on edits directly and claim action items in comments; for the accompanying collection-form and comment-rollup tricks, see C3 Tencent Docs Connector.
🔧 Dictate a supplement on the way out of the meeting. For background you didn't capture during the meeting, just hold the voice shortcut after it ends and speak to WorkBuddy—on Mac, long-press left ⌥; on Windows, long-press ALT to invoke voice input; the recognized text fills the input box and can be further edited, then have it merge the supplement into the minutes.
Lesson Summary
✅ A one-stop, three-step workflow for AI meeting minutes: hand the material to WorkBuddy (recording / notes / chat history) → transcribe into text with time ranges → use the official prompts to organize it into minutes with the five elements.
✅ The official five elements of minutes: meeting topic, key conclusions, action items, owners, deadlines—lock down the structure and the rambling log disappears.
✅ Audio-to-text is a paid capability: about 0.6 credits/minute based on actual duration; the first call in a free session pops a confirmation, and the audio is uploaded only after you confirm.
✅ For zero credits and privacy, choose Local Whisper: install name local-whisper, local offline transcription, sensitive meeting recordings never leave your device.
✅ Three delivery options: keep it in the workspace, upload to the cloud as a Tencent Docs smart document, or share to WeChat/WeCom.
Key information recap:
| Item | This lesson's scope |
|---|---|
| Transcription billing | about 0.6 credits/minute, based on actual audio duration; the first call in a free session pops a confirmation |
| Audio formats | wav, mp3, m4a, flv, mp4, wma, 3gp, amr, aac, ogg-opus (.opus/.ogg), flac |
| File limits | single file ≤1GB; duration ≤5 hours |
| Supported languages | Chinese, English, Cantonese, and some dialects |
| Local transcription | Local Whisper skill (install name local-whisper), runs offline |
| Five elements of minutes | meeting topic, key conclusions, action items, owners, deadlines |
Next Lesson Preview
Once the minutes are done, the action items often include a line like "produce a proposal document"—that's exactly what the next lesson covers. Next lesson: A3 Word Document Generation: A Proposal in One Sentence, Refined Over Multiple Rounds Until You're Satisfied.
You will learn:
- The official annual party planning proposal prompt—get a first draft in one sentence
- The three-round polishing method: more detailed budget, add precautions, more formal tone
- How to refine document tasks like minutes and proposals round by round until they're ready to deliver