Meta Private Message Summaries vs ThreadRecap | ThreadRecap
WhatsApp now summarizes your unread messages for you. So do you still need anything else?
Short answer: sometimes. The longer answer is worth reading, because the two tools solve different problems and mixing them up costs you either money or useful context.
The headline everyone keeps missing
Meta's Private Message Summaries are not a chat analyzer. They are an inbox helper. They summarize a handful of unread messages inside WhatsApp, on your device, so you can get back into an active conversation faster.
That is a good feature. It is also a very narrow feature.
ThreadRecap is a different job entirely: take an exported conversation, often weeks or months of it, sometimes 10,000+ messages with voice notes, and produce a structured record you can reference, share, or defend.
Same surface (WhatsApp). Different job.
Why the confusion keeps happening
The timing matters here. Meta rolled out Private Message Summaries at roughly the same time that third-party WhatsApp analysis tools started gaining attention. That coincidence made the two look like competing answers to the same question. They are not. The Meta feature answers "what did I miss in the last hour?" The export-based approach answers "what was the full arc of this conversation and what do I need to prove it happened?"
The word "summary" does a lot of heavy lifting in both cases, which is part of the problem. A one-paragraph narrative of your last 30 unread messages and a structured multi-section report covering 60,000 messages are both technically summaries. Treating them as interchangeable because they share a label leads to real mistakes — mostly people discovering too late that the native feature produced nothing they can actually use outside of WhatsApp.
The 30-second decision
Ask yourself one question: do I need this conversation to still exist tomorrow, outside of WhatsApp?
If no, use Meta's native summary. It is free, fast, and private by design.
If yes — because you are building an evidence report, documenting decisions for a team, or keeping a paper trail for a dispute — you need an export-based tool. Meta's summary vanishes with the context. It never leaves WhatsApp.
What Meta's Private Summaries actually do
Runs on-device ("Private Processing"), so your messages stay on your phone.
Summarizes unread messages in a chat you already have open.
Returns a narrative paragraph, not structured output.
Works on text. Voice notes are transcribed separately and not integrated.
No export, no archive, no downloadable record.
It is a great answer to "I have 80 unread messages in the family group, what happened?"
It is a terrible answer to "I need to prove what we agreed on three months ago."
The on-device model ceiling in practice
Running a language model entirely on a mobile device is a genuine engineering achievement. The privacy benefit is real: if nothing leaves the phone, no server can be compromised or subpoenaed. But physics and memory impose hard limits. The model that fits in device RAM is necessarily smaller and less capable than a cloud-hosted model. In practice this means Meta's on-device model has a practical scope ceiling — it cannot meaningfully analyse long chats with hundreds of messages, nuanced topics, and mixed voice/text content.
This is not a criticism of Meta's engineering. It is a fundamental trade-off. A model small enough to run on a mid-range Android phone cannot hold the same context window or reasoning depth as a model running on data-centre hardware. For catching up on the last few dozen messages in a casual group chat, that trade-off is entirely reasonable. For reconstructing six months of business negotiations with interspersed voice notes, it is not the right tool.
WhatsApp's native voice note transcription is a separate thing
One detail that trips people up: WhatsApp does have a native voice note transcription feature, but it is distinct from Private Message Summaries. The two features do not communicate. If someone sends you a voice note and you read it as a text transcript, that transcript is not fed into the unread-message summary. The summary works on text messages only. If your chat is heavy with voice notes — which is common in Brazilian, Portuguese, and many Latin American usage patterns where voice messaging is the primary mode of communication — the native summary is missing a significant portion of the conversation's actual content.
What ThreadRecap does differently
Reads a full WhatsApp export (`.zip` with text, images, voice notes).
Works on chats of any length — 200 messages or 20,000.
Transcribes voice notes with Whisper and merges them into the timeline.
Produces structured output: summary, decisions, action items with owners, open questions, and quoted evidence.
Saves the report — you can come back to it, share it, export it to PDF, or feed it to a lawyer.
Supports dispute-specific goals (business partner, small claims, landlord/tenant, workplace, freelancer).
Different tool for a different job.
How ThreadRecap handles voice notes end to end
ThreadRecap's voice note pipeline is worth understanding specifically because it is the capability most absent from Meta's native feature. WhatsApp stores voice messages in `.opus` format on Android and `.m4a` on iOS. Both formats are supported. When you opt in to voice note transcription, the audio files are encrypted in transit and sent to OpenAI Whisper for transcription. Whisper achieves approximately 95% accuracy on clear audio, which covers the vast majority of real-world voice messages recorded in reasonably quiet environments. Once transcription completes, the audio is deleted from the processing server. The resulting text transcript is timestamped and inserted into the conversation timeline in chronological order alongside the text messages, so the analysis treats the full conversation as a unified record rather than treating voice and text as separate streams.
This integration is the technical difference that matters most for dispute use cases. A business negotiation conducted partly over voice notes and partly over text is common. If the analysis only covers the text portion, it may systematically miss commitments, clarifications, or disagreements that were communicated verbally.
Scale and format
ThreadRecap handles chats of any length, up to 60,000+ messages or a 2 GB ZIP file. For reference, a two-year group chat in an active work or family group can easily reach tens of thousands of messages. Meta's on-device model cannot approach that scale. The structured output ThreadRecap produces — timeline, quoted messages, transcribed voice notes — is designed to be suitable for mediation or small claims court submissions, not just personal reference.
Side by side
Meta Private Summaries
ThreadRecap
Input
Unread messages in open chat
Full exported chat (any size)
Scope
Narrow — catching up
Broad — documenting, deciding, disputing
Voice notes
Separate native transcription
Integrated into analysis timeline
Output
Narrative paragraph
Structured report with sections
Persistence
Ephemeral
Saved, exportable, shareable
Privacy model
On-device, no server
Text parsed locally, only text sent to AI
Cost
Free
5 free credits, then pay-per-use
Best for
"What did I miss?"
"What was agreed? What do I do next?"
When to use which (real scenarios)
You have 120 unread messages in a work chat and a standup starts in 10 minutes.
→ Meta's native summary. Get the gist fast.
A supplier is denying the payment terms you agreed on via WhatsApp two months ago.
→ ThreadRecap. You need a timeline with quoted messages and transcribed voice notes you can send to a lawyer.
Your family group has 500+ messages this week. You want a laugh and a highlights reel.
→ Neither. Use ThreadRecap's Group Awards for the playful output, or Meta's summary if you just want the catch-up.
You're a project manager reviewing three weeks of WhatsApp discussion before a go/no-go meeting.
→ ThreadRecap. You want decisions, action items, and open questions — not a paragraph.
Your tenant is disputing a deposit deduction. You have eight months of chat with them.
→ ThreadRecap. The dispute summary goal gives you a structured record suitable for mediation or small claims.
A note on team and professional use
Project managers, paralegals, and operations leads increasingly use WhatsApp for real work communication, especially in markets where it functions as the default professional messaging layer. For these users the native summary is useful on a daily basis for keeping up with volume, but it does not produce anything that can live in a project management system, a legal file, or a client record. ThreadRecap's structured output — with explicit sections for decisions made, action items with owners, and open questions — maps cleanly onto how professional documentation actually needs to be organised. The ability to re-run an analysis on the same export with a different goal is also relevant here: a conversation that you first process for meeting notes may later need to be re-processed for dispute evidence if a relationship breaks down.
The privacy question most people get wrong
Yes, Meta runs Private Summaries on-device. That is genuinely good.
But on-device has a ceiling: the model has to be small enough to run on your phone. That limits what it can do. A long chat with hundreds of messages, nuanced topics, and voice notes is beyond what any on-device model can handle meaningfully.
ThreadRecap takes a different privacy path: your chat file is unzipped and parsed in your browser. Photos and videos never leave your device. Voice notes are sent for transcription only if you opt in, encrypted in transit, and deleted after processing. Only the text goes to the AI.
Both are reasonable privacy models. They make different trade-offs. Neither is pretending the other is unsafe.
What "ephemeral" actually means for your records
The ephemeral nature of Meta's Private Message Summaries is worth making concrete. The narrative paragraph the feature generates is not stored in WhatsApp's servers, is not indexed anywhere, and is not retrievable after you close the summary view. If you take a screenshot, you have a screenshot — not a structured record, not a searchable document, not something with timestamps attached to individual messages. For routine inbox management that is fine. For anything you might need to reference in two weeks, or anything that touches a financial or legal question, ephemeral is the wrong property for the output to have.
What to do today
If you want to catch up on a live chat, just use WhatsApp's native summary — it is already built in.
If you want to document, decide, or dispute, export the chat and run it through ThreadRecap. Start with the dispute summary flow if there is a disagreement involved, or with meeting recaps if it is a team conversation.
Either way, the point is to match the tool to the job. Meta built a great inbox assistant. ThreadRecap is built for the cases where "assistance" is not enough and you need a record.
If the conversation matters enough to remember, it deserves more than a paragraph.
Further reading:The complete WhatsApp evidence guide covers admissibility, how to prepare an export for use in a dispute, and jurisdiction-specific notes for Brazil, UK, US, and Portugal.
