How Encrypted Voice and Video Calls Stay Private

Encrypted voice and video calls stay private by locking your audio and video on your own device before a single packet leaves it, so only the person on the other end can unlock them. The method that does this for live calls is called DTLS-SRTP, and it behaves very differently from the encryption that guards your text messages.

This guide explains how call encryption actually works, why the media has to travel peer-to-peer, and how a call on your phone can be routed so the other party never learns your IP address.

Text encryption and call encryption are not the same

Most people first meet end-to-end encryption (E2EE) through messaging, where a message is encrypted, parked on a server, and decrypted when the recipient opens it. Timing is not critical: a message can wait.

A live call is the opposite. Audio and video are generated dozens of times a second and have to arrive in milliseconds, or the conversation turns into a laggy mess. That real-time pressure is why calls use their own encryption toolkit rather than the one built for chat.

Signaling versus media

Every call has two separate streams:

• Signaling is the setup chatter: "I want to call you," "I'm ringing," "here's how to reach me," "hang up." It is small and infrequent.
• Media is the payload: the encoded audio and video, flowing continuously while you talk.

Keeping these two ideas apart is the key to understanding call privacy, because they can travel by completely different routes and still both be protected.

What DTLS-SRTP does

Real-time media is carried by a protocol called RTP. On its own, RTP is not private. SRTP (Secure RTP) is the encrypted version: every packet of audio and video is scrambled so anyone intercepting it sees noise.

But encryption needs keys, and both devices must agree on them without leaking them. That is the DTLS handshake. The two devices run a cryptographic negotiation and derive shared keys that exist only on the endpoints. The server that helped set up the call never holds them. That combination — DTLS to agree keys, SRTP to encrypt the stream — is why the pairing is written DTLS-SRTP, and it is the same core approach used across modern secure calling.

Why call media goes peer-to-peer

Here is the part that surprises people: for the audio and video to be fast, the two devices usually connect directly to each other, not through a central server. This is "peer-to-peer" (P2P) media.

Routing every packet through a company's servers would add distance and delay, and — more importantly for privacy — it would place the provider in the middle of your live conversation. A direct path is faster and keeps the media off the middleman's machines.

The trade-off is a networking reality: when two devices talk directly, each one can see the other's IP address. An IP address is roughly the return address for your internet connection; it can hint at your rough location and your internet provider. For a call with a close friend that may be fine. For a call with a stranger, a source, or a support contact, it may not be.

How a call can hide your IP address

The fix is a relay. Instead of connecting straight to each other, one or both devices send their encrypted media to a relay server that forwards it on. The relay shuttles scrambled packets back and forth, so the far end sees the relay's address instead of yours.

Two things are worth being clear about:

1. A relay changes the path, not the privacy of the content. The media stays end-to-end encrypted; the relay only ever sees ciphertext, never your picture or your voice.
2. Relaying adds a little delay, which is why many apps only relay when it helps — for example, on mobile networks, where hiding your IP matters more and direct connections are harder to establish anyway.

So the honest summary is: peer-to-peer is great for speed, a relay is what protects your IP, and good call design chooses between them deliberately.

How Graphite handles encrypted voice and video calls

Graphite is an end-to-end-encrypted messenger built by PrivacyPortal, and its calling follows exactly the model above.

• Media is encrypted with DTLS-SRTP. Your audio and video are protected end to end; the keys live on the two devices, not on any server.
• Media travels peer-to-peer. The live stream takes the fast direct path rather than being funneled through Tor, which would be too slow for real-time calls.
• On mobile, calls are routed so the other person does not see your IP. When you call from a phone, the media path is arranged so your address is not exposed to whoever you are speaking with.
• Signaling rides over Tor. Graphite runs over Tor by default — fail-closed, to a bundled hidden service — which hides your IP and network for the setup and messaging side of a call. The call media uses the DTLS-SRTP peer path instead, a deliberate split so calls stay both private and usable.

Crucially, none of this asks for a phone number or an email. Your identity in Graphite is a 12-word recovery phrase generated on your device; there is no account sitting on a server to leak. If you want the deeper version of the network story, see how to hide your IP when messaging and Tor messaging explained.

How this compares across popular apps

Encryption of call content is now common; the bigger differences are what each app knows about you and how you sign up. Those details shape how private a call really feels.

App	Phone number to register?	Message content E2EE by default?
Graphite	No (12-word phrase)	Yes
Signal	Yes	Yes
WhatsApp	Yes	Yes (cloud backups separate unless you enable encrypted backups)
Telegram	Yes	No — only opt-in Secret Chats are E2EE
Session	No	Yes (onion-routed, open-source)
Threema	No	Yes (paid, open-source clients)

A note on fairness: Signal is E2EE by default, open-source, and minimizes metadata, but it does require a phone number to register (usernames can hide that number from other users). WhatsApp encrypts message content by default with the Signal Protocol but is owned by Meta and ties you to a number. Telegram's ordinary cloud chats are not end-to-end encrypted — only its opt-in Secret Chats are — so it should not be described as encrypted by default. Session and Threema, like Graphite, let you skip the phone number entirely.

The quick mental model

• Signaling sets up the call and can be hidden over a network like Tor.
• Media is the audio and video, encrypted with DTLS-SRTP, and usually sent peer-to-peer for speed.
• Peer-to-peer exposes IPs; a relay is what hides your address, at the cost of a little latency.
• Your identity — whether you had to hand over a phone number at all — is a privacy question that exists before the call even starts.

Try encrypted calling without an account

If you want encrypted voice and video calls without handing over a phone number or an email, Graphite is free and in public beta on Windows (a portable .exe with bundled Tor), Android (it works on GrapheneOS and de-Googled phones with no Google Play Services), and the web. No phone number, no email, and a no-logs relay that only ever holds sealed ciphertext.