SpeakLinker
WhatsApp voice messages, now readable.
SpeakLinker turns WhatsApp voice notes into text without asking anything of the person who receives them. Add SpeakLinker as a contact, forward it a voice message, and the complete transcription comes back in the same chat a few seconds later — in whatever language was spoken, detected automatically. No upload, no app, no sign-up.
Category
Voice & messaging
Stage
Live, free
Model
Free core, accounts for automations
Interfaces
WhatsApp, web app
The problem
Voice notes are the fastest way to send a message and one of the slowest ways to receive one. They cannot be skimmed, searched or read in a meeting, and a two-minute recording about a one-line decision still takes two minutes to hear.
Transcription tools exist, but nearly all of them ask the receiver to do the work: install an app, export the audio, upload a file, pick a language. For something people deal with dozens of times a day, every one of those steps is a reason not to bother.
How it works
The interface is the one people already use. Add SpeakLinker to your contacts, forward any voice note to it, and read the reply. There is nothing else to learn, and the language is identified from the recording rather than chosen from a list.
Behind that chat, the gateway only records and queues each message; downloading, audio conversion, speech-to-text and the reply all happen in workers that scale independently. Every stage is idempotent and retried with backoff, so a timeout never produces a duplicate reply, and if WhatsApp disconnects, replies wait rather than being lost.
WhatsApp is treated as one channel rather than the architecture. Everything past the gateway works with a normalised message, which is what lets the same pipeline later serve other messaging channels and calls.
Beyond transcription
People who create an account can link their WhatsApp number and attach automations to every transcript: translate it into a preferred language, summarise it as a short paragraph or bullet points, pull out action items, dates, phone numbers and addresses, or apply an instruction of their own. Results arrive back in WhatsApp, and can also be forwarded to an email address or posted to a webhook.
The account adds a searchable history and a retention period the user chooses — without making registration a requirement for the core service.
Privacy by default
Original audio is encrypted at rest, deleted automatically after a successful transcription, and never kept longer than 24 hours. Transcripts for people without an account are kept for around a week and then removed.
Sending the word DELETE to SpeakLinker erases stored history at any time. Credentials and keys are encrypted with AES-256-GCM, and data is never sold or used for advertising.
Capabilities
What it does
Forward and read
No upload and no app: forward a voice note in WhatsApp and the transcription comes back in the same chat.
Automatic language detection
Works with voice notes in almost any language, identified from the recording rather than chosen by the user.
Translation and summaries
Optionally translate every transcript into your language, or reduce it to a short summary or bullet points.
Extraction
Pull action items, dates, phone numbers and addresses out of a rambling voice note into something you can act on.
Email and webhooks
Send each transcript to an inbox, or POST it to your own HTTPS endpoint to feed a CRM or workflow.
Chat commands
Help, privacy, language and delete are all available as plain messages, without ever opening a website.
Built with
- NestJS
- BullMQ workers
- PostgreSQL
- Redis
- FFmpeg
- Whisper large-v3
- React
Next venture
Speak
Speak, and it types. In almost any app.