← All posts
OmniCall

Can AI Translate a Live Phone Call Without an Interpreter?

August 23, 2026 · Axentra
Can AI Translate a Live Phone Call Without an Interpreter?

Yes. As of 2026, AI can translate a live phone call in real time — voice to voice, in both directions — without adding a third-party interpreter to the line. The caller speaks their language, the other party hears (or reads) it in theirs, and each reply is translated back, typically with a delay of a couple of seconds rather than instantly. This already ships on consumer phones — Samsung's Live Translate, Google's Pixel 10 Voice Translate and Live Translation in the iPhone Phone app — and in 911 tools such as Prepared's two-way audio translation; Motorola Solutions also announced an Interpreter Agent integrated into its CAD in June 2026. It does not replace a human on complex, high-stakes conversations — but it can remove the wait, the third party, and the handoff on many routine calls.

Below is what's genuinely possible today, how it works, where the honest limits are, and how to pick the right option for a 911 center.

How does real-time voice-to-voice call translation work?

The classic approach has three steps. Some systems translate after each turn — Samsung's Live Translate, for example, delivers translations after each person speaks — while newer streaming systems work as the person talks; Google has even built an end-to-end model that goes directly from speech to speech:

  1. Speech recognition (ASR): the caller's audio is transcribed to text as they talk.
  2. Machine translation: that text is translated into the target language.
  3. Speech synthesis (TTS): the translation is spoken aloud in a natural voice to the other party.

The result is a two-way, spoken conversation where neither person has to speak the other's language. Many systems also produce a written transcript — Pixel 10, for example, shows an on-screen transcript.

The key shift from older tools: it's voice-to-voice and round-trip, not just captions on a screen. It feels like a call, not a chat window.

How fast is it? Does the delay make it unusable?

Speed is the whole game: a translation that arrives ten seconds late breaks the rhythm of a real conversation. Today's best systems add a delay measured in seconds: Google describes its real-time speech-to-speech model as having only a 2-second delay, and MacRumors notes that Live Translation in the iPhone Phone app speaks the translation aloud with a slight delay. Expect short pauses, but the exchange stays inside a single live call instead of adding an interpreter line.

How accurate is AI phone call translation?

Accuracy is generally best for everyday language and clear audio, but it isn't flawless — an Android Authority hands-on with Pixel 10 Voice Translate on real calls found both good and bad.

That's why the responsible deployment keeps a human in the loop: the operator sees the live transcript, confirms critical details back ("you said the intersection of X and Y — correct?"), and escalates to a certified human interpreter when the stakes demand it.

Can AI replace a phone interpreter service entirely?

Honest answer: not entirely, and you shouldn't want it to. What it does replace is the friction — identifying a language, dialing an interpreter line, waiting for a connection, and running a slower three-party call. For routine multilingual calls, AI handles the conversation right away; for complex, legally sensitive, or ambiguous ones, you still route to a human. AI is the fast default and the human interpreter is the escalation path.

What languages can it handle?

Coverage varies by product. Samsung's Live Translate debuted with 13 languages and supports 20 on the Galaxy S25 series; Pixel 10 Voice Translate works between English and 10 languages; and Prepared, which launched its two-way audio translation for 911 starting with Spanish, also offered real-time translation of 911 call transcripts in more than 30 languages as of its 2024 announcement. For a US/Mexico operation, English⇄Spanish is the key pair. The practical question isn't just "how many languages" — it's whether the system detects the caller's language quickly and switches without a manual step.

Is real-time call translation secure and private?

This matters for 911, healthcare, and any regulated contact center. Ask any vendor:

Deployments that sit on top of the systems you already run help keep security, retention, and workflow under your control.

Why consider Axentra OmniCall for a 911 center

When a call comes in another language, a 911 center usually has three paths: a phone interpreter line (three parties and a wait), the caller's own phone translation (it depends on the citizen's device and the center doesn't control it), or an AI platform for 911. But translation is only one moment of the call: first it has to be answered, and afterward it has to be entered, classified, and dispatched.

Axentra OmniCall covers the whole call. It answers every 911 call on the first ring, 24/7, with natural conversation — no menu, no "press 1." It resolves and closes calls that aren't emergencies (pranks, questions, misdials) and records the reason. In real emergencies it is the operator's copilot: it translates both ways, transcribes, fills in the CAD, detects sounds and emotions, and guides the protocol. All of it on top of the phone system and CAD you already run, with a human always in charge.

Here's how it compares with today's most common workflow, an interpreter line plus manual entry:

CriterionAxentra OmniCallAlternative
Answers on the first ring, no menu✓ 24/7, natural conversation✗ Depends on an operator being free
Non-emergency calls✓ Resolves them, records the reason✗ Tie up an operator
Two-way translation✓ No interpreter; natural voice✗ Dial a line, three-party call
Live transcription✓ 97% accuracy✗ No live transcript
CAD entry✓ < 2 s; Tyler, Hexagon, Motorola or own✗ Manual typing
Sounds and emotions✓ Gunshots, screams, glass; fear, panic✗ Operator's ear only
Protocol guidance✓ Next question, urgency, dispatch; auditable✗ No live AI guidance
Existing phone system and CAD✓ Integrates, no replacement✓ Same
Human in charge✓ Of every emergency✓ Same

Bottom line: an interpreter line solves the language; OmniCall solves the language, the answering, and the data entry in a single workflow.

What about other AI platforms for 911? Prepared and Motorola Solutions also offer AI translation for 911, and they're worth evaluating. Axentra's real advantages lie elsewhere: it connects to the CAD you already run (Tyler, Hexagon, Motorola or the state's own system) without replacing it; it's built in Spanish for Mexico and Latin America, where the operator reads the translation in Spanish (and in English in the United States); it can run in the cloud or in the institution's own command center; and Axentra is one provider across video, 911, data, and citizen safety, with hardware and turnkey delivery. Ask every vendor to prove it on your own calls.

Want to test it in your 911 center? We offer two pilots: 10 recorded calls with known outcomes, replayed as if live, or a phone number with the system ready for your operators to try. Talk to us.

Operations that can’t run on guesswork?

See Axentra working in an environment like yours.

Talk to us →