50+
You wake up to 50 voice notes.
The family group. Two clients. A 4-minute ramble from your cousin. Some urgent, most noise.
Send a voice note — get the summary, the tasks, the answer. Then ask your own history anything. Encrypted transport. AI on your terms.
50+
You wake up to 50 voice notes.
The family group. Two clients. A 4-minute ramble from your cousin. Some urgent, most noise.
90m
And lose 90 minutes replaying them for what was asked.
Scrub, rewind, replay. Voice is how everyone you know talks — and the one thing your phone can't skim.
6
Six promises slip before 10am.
The viewing you said you'd confirm. The quote you owed before noon. Buried four threads deep by lunch.
→
Telbox listens before you do.
Every voice note arrives with its summary, its tasks, and a reply drafted in your phrasing — ready before the first sip.
Voice notes that work for you
Every voice note is transcribed, summarized, and slotted into a brief before you open the app. Tap a line in the brief, hear the exact clip it came from. Code-switching kept intact, accents preserved.
"Habibi, I want twenty units CIF Nairobi by today, same papers as last shipment. Send me the quote قبل الظهر."
7:02Yes. Voice note 14 Apr, 0:23. He said the building has "two bays underground, valet only after 9pm."
Echo · your daily voice journal
Echo is a thread with yourself. Record a thirty-second reflection on the drive home, in the lift, before bed — what happened, what's on for tomorrow. No contacts needed, no audience.
The moment you stop talking, Telbox turns the entry into a clean transcript and summary. Streaks build the habit day by day, and weekly themes show you what your week actually sounded like.
Echo entries join your history immediately. "What did I say about the Nairobi quote?" "When did I last mention the gym?" Ask, and Telbox answers from your own entries, with sources.
What Telbox does
Dialect-tuned transcription for Arabic, English, and the half-and-half way people actually trade. Voice quality holds up over LTE, jitter, and the occasional crackling FaceTime audio.
Fifty voice notes compress to four decisions. Every brief is evidence-linked: tap a line, hear the clip it came from, see the exact second the speaker said it.
Replies in your phrasing, not a generic AI tone. Telbox learns the way you actually open and close, the way you say "yes" to a client versus a vendor, and drafts ready to send.
Ask anything across every voice note you've ever sent or received. "Did we ever talk about parking?" surfaces the clip from twelve weeks ago in under a second.
Every clip you send is Ed25519-signed end-to-end. A deepfake of your voice can't sneak into the thread, and the public verifier means anyone can audit the chain of custody.
Calls, on Telbox
E2E encrypted on the WaveLink stack. A transcript and a next-action summary the moment you hang up — your calls join the history you can ask. Live Arabic ↔ English translation and captions: coming soon.
Get the betaCommit to 20 units CIF Nairobi · today? Confirm by voice to draft the quote.
Noise_NK handshake + AEAD per-frame seal. Server never sees plaintext.
Arabic ↔ English streamed mid-call, dialect-aware. In development; not in the current build.
Hang up. Get a signed transcript and a next-action list before you stand up.
Every participant cryptographically identified. Deepfakes can't sit in.
LAN → direct UDP → DERP relay. Always the lowest-latency leg.
Privacy posture
The transport layer never bends. AI processing tiers up. You always know which one you're on.
Every message, voice note, and call routes end-to-end encrypted between your device and the recipient. Telbox never sees plaintext in transit. True on every tier.
AI processing routes through the same Gemini API endpoint on every tier, free and paid alike. Google is contractually forbidden to train on your prompts or responses under the Google Cloud DPA.
Free runs the same no-training AI with monthly usage limits. Telbox Plus — coming soon — lifts the monthly AI limits, with priority routing and unlimited /ask memory.
For people
Voice + calls, encrypted. AI on your terms, with monthly limits.
Lifts the monthly AI limits. Priority queue.
Voice clone, calendar, real-time call translation.
SSO, audit, legal hold, retention. 5-seat minimum.
For businesses
Three platform plays the consumer app makes possible. Each one ships on the same WaveLink transport and no-training AI fabric Telbox runs in production.
$99/seat/month
Asynchronous voice for sales, customer success, and call-center teams. Every commitment, objection, and win condition indexed, transcribed, briefed, and signed. The Gong playbook applied to voice notes.
$50k–$500k ACV
License Telbox's no-training AI and WaveLink transport as a B2B2C infrastructure layer. Fintech, healthcare, legal, government. Auth0 for voice.
$199/month
A premium AI assistant that takes voice direction and reports back in voice. For founders, executives, and anyone whose calendar runs on the phone.
Talk to us about pilots, custom residency, or platform partnerships.
Email info@telbox.aiFor developers
A cryptographically verified AI participant inside an end-to-end encrypted messenger — built by talking, reaching the world through MCP both directions, shipped with Python & TypeScript SDKs.
Per-agent Ed25519 keys. A tap-to-verify badge, not a spoofable label.
Describe it; the @build compiler emits a typed, validated agent.
Consume external tools; be called from Claude Code, Cursor, Zed.
In TestFlight now · App Store coming soon
Telbox is free, and the App Store launch is coming soon. Want in early? The beta is open now — free, iOS 17+. Email us with your TestFlight Apple ID and we'll send the invite.
Get the beta