WhatsApp calls can now be recorded and transcribed — here’s how it works
Meta’s Calling API now supports server-side recording and speaker-separated transcripts, with a spoken consent announcement. What it means for sales and support teams.
For years, “what did we promise on that call?” has been one of the hardest questions for any team that sells or supports customers over WhatsApp. Calls happened on someone’s phone, and the only record was a memory. Meta’s WhatsApp Business Calling API now changes that: businesses can switch on recording and transcription for individual calls, directly in the API.
What Meta added
- Call recording — the audio of business-initiated and user-initiated calls, delivered after the call ends.
- Call transcription — a JSON transcript with the full text plus speaker-separated segments (business and customer), timestamps and confidence scores.
- A spoken announcement — before anything is captured, both people hear a notice built from the purpose you provide, in the language you choose.
Both features are opt-in per call. You enable them when you place a call, or when you accept an incoming one, by sending a short configuration with a purpose (up to 250 characters) and an announcement language. Recording and transcription are independent: you can use either or both, and when both are on, the customer hears a single combined announcement.
Recording starts only after the announcement has played. A customer who doesn’t want to be recorded can simply end the call — so your purpose text should be short, honest and specific (“to confirm your order details”).
How the files arrive
Shortly after the call ends — typically within a minute — Meta sends a webhook with a media ID, a SHA-256 hash for integrity checks and a temporary download link. The link expires after five minutes, and Meta keeps the file for seven days. After that, it’s gone. If you want to keep recordings, you need to download and store them yourself.
Transcripts detect the spoken language automatically. Support is broad but changes over time, and for unsupported languages the transcript can come back empty — so treat transcripts as a powerful convenience, not a legal record.
How ThreadWave handles it
- Turn recording and transcription on separately in settings, write your purpose and pick the announcement language.
- When the files arrive, ThreadWave downloads them, verifies the hash, converts audio to a compact format and saves everything encrypted on your computer.
- The recording player and a collapsible, speaker-by-speaker transcript appear right under the call in the customer’s thread.
- If a file arrives late or a download fails, ThreadWave retries in the background — well within Meta’s seven-day window.
“For quality and training.” · “To confirm the details of your order.” · “So we can follow up accurately on your request.”
Meta prices recording and transcription separately from calling, and availability can depend on your account — check Meta’s documentation and your billing before turning them on for every call.