Live Translate App Overview
The interpreter your compliance team can actually approve — company-governed, speech-to-speech translation for the frontline moment two people need to clear a language barrier on the floor.
What is the Live Translate App?
Live Translate is a turn-based, speech-to-speech interpreter built for frontline conversations. Two people share one phone (or pair two phones with a short code); each picks their language and holds a button to talk, and the other hears the message spoken back in their own language across 15 languages.
What sets it apart isn’t the translation — it’s the governance around it. Consumer translate apps can’t be used with sensitive workplace information (patient data, customer PII, incident details, HR conversations), so they get prohibited — and then used anyway. Live Translate runs through your organization’s own translation provider, keeps every conversation ephemeral (nothing is stored), and stays inside your tenant boundary — so you can sanction one tool instead of banning the rest.
Core Value Proposition:
- 🔒 Company-governed — runs through your configured translation provider, inside your tenant; sanctionable where consumer apps are banned
- 🕊️ Private by default — every turn is transcribed, translated, and spoken back in real time; nothing is persisted
- 🗣️ Speech-to-speech — the listener hears the translation spoken in their language, not just reads text
- 💳 No license to buy — included for every tenant; its AI usage draws from your platform AI/Voice credits as you use it
Key Concepts (How It Works)
If you only read one section, read this — it explains the model behind the screens, and clears up the three things people get wrong about Live Translate.
What a “turn” is (and why nothing is stored)
Live Translate has no conversation database. Each time someone speaks, that single utterance — a turn — runs through three steps and then disappears:
- Speech in → the recorded turn is transcribed (OpenAI Whisper).
- Translate → the transcript goes to your configured translation provider.
- Speech out → the translation is synthesized and spoken back in the listener’s language.
Turns are deliberately run uncached, straight through the provider — there is no stored transcript, no saved audio, and no conversation record on the server. The on-screen conversation log lives only in that browser tab for the length of the session. This is what makes the “nothing is stored” promise literally true at the code level, not just a policy. (Cost-tracking counters — how many transcriptions/translations ran and how long they took — are recorded, but never the words themselves.)
How the translation provider is chosen — “your provider,” precisely
“Your organization’s own provider” resolves in a definite order, so you always know what engine a turn ran through:
- If your admin has connected a dedicated machine-translation engine as an enabled translation integration (for example AWS Translate or DeepL, BYOK — bring-your-own-key), every turn uses it. This is faster and cheaper, and keeps translation on the vendor you chose and signed terms with.
- Otherwise, the platform-managed default engine is used.
- If neither is available, Live Translate falls back to the platform LLM so translation still works with zero configuration.
Because the provider is resolved per business, Live Translate uses the same engine you’ve configured for the rest of the product — it never quietly routes sensitive speech through a different consumer cloud. When a specific provider is configured, its name is even shown in the privacy line on the screen.
Turn-based vs. hands-free vs. “full-duplex” — three things people conflate
This is the distinction users most often get wrong, so here it is precisely:
- Push-to-talk (turn-based) is the baseline in both modes: you hold a button, speak, release, and the other person hears the translation. Available on one device and two.
- Hands-free (continuous) is an option in two-device (paired) mode only. Toggle it on and you stop holding the button — the app listens, detects when you start and stop speaking (voice-activity detection), and automatically sends each utterance on the trailing pause. It even suppresses your mic while the other person’s translated audio is playing, so it doesn’t translate the reply back and loop. It feels continuous, but under the hood it’s still auto-segmented turns.
- Full-duplex, always-on call interpretation — where both people talk over each other in a live stream, like a simultaneous interpreter on a phone call — is not available yet; it’s a later phase. Whisper is batch, not streaming, so today’s “continuous” is the hands-free VAD mode above, not true streaming duplex.
So: “Is it real-time?” → Yes, turn by turn. “Is it hands-free?” → Yes, in two-device mode. “Can we both talk at once like a live interpreter?” → Not yet.
Single device vs. two paired devices
- Single device (
One devicetab): two people share one phone. Person A and Person B each pick a language; each holds their own “Hold to speak” button. Everything happens on that one device — there’s no code, no pairing, no network of two phones. - Two devices (
Two devicestab): each person uses their own phone. One taps Start a session and shares a 6-character code; the other joins with it. Turns are relayed between the two phones over a live channel, and each phone plays the translation in its owner’s language. This is the mode that adds the hands-free toggle.
They solve the same problem in two physical situations — one shared handset, or two people who each keep their own phone.
Who can do what — the glossary is the only admin surface
Live Translate itself has no per-role gating beyond app access: any employee who can see the app can start or join a session, on either mode. There is no “manager-only” interpreter.
The one admin-only surface is Company terminology (the glossary), reachable from the “Company terminology” button that appears in Live Translate only for admins. That’s where an admin governs how words translate (see the next section). Regular users translate; admins tune the vocabulary.
How the company glossary shapes every translation
The glossary — managed at Company terminology (/admin/translation) — lets an admin pin how specific words come out, and it applies to Live Translate turns automatically (and to the rest of the product’s translation surfaces too). Two kinds of rule:
- Mapping — a source word in one language must render as a specific target word when translating into another language (e.g. a product name, an SKU, a site code, an internal term).
- Do-not-translate — a source word must pass through untouched in any language (brand names, equipment models, drug names).
It works provider-independently: before a turn is translated, matched terms are swapped for inert markers the engine won’t touch (protect), the text is translated by whatever provider is active, then the markers are swapped back to your chosen term (restore). Because it wraps around the provider, the same glossary behaves identically whether you’re on AWS Translate, DeepL, or the LLM fallback. Admins can build the glossary by hand, import/export CSV, or adopt a starter pack (e.g. Warehouse & Manufacturing Safety, Retail & Store Operations, Healthcare & Facilities — all English → Spanish).
At a Glance
| 📊 Availability | 🌐 Languages | ⚙️ Modes | 📱 Mobile Ready |
|---|---|---|---|
| Every tenant, no license | 15 out of the box | Single + paired two-device | ✅ Yes |
Perfect For:
- 🏭 Frontline managers & associates — clearing a quick instruction on the floor across a language barrier, no app-switching
- 🧰 Field & service workers — communicating with a customer or site contact one-on-one in the moment
- 🛡️ Compliance & IT leaders — sanctioning a governed interpreter for sensitive conversations instead of policing consumer apps
How It Works
One Turn, End to End
┌─────────────────────────────────────────────────────────────────────────┐
│ ONE LIVE TRANSLATE TURN │
├─────────────────────────────────────────────────────────────────────────┤
│ ┌──────────────┐ ┌──────────────┐ ┌──────────────┐ │
│ │ HOLD & SPEAK│──────▶ │ TRANSCRIBE │──────▶ │ GLOSSARY │ │
│ │ (your lang) │ │ (Whisper) │ │ protect → │ │
│ └──────────────┘ └──────────────┘ └──────┬───────┘ │
│ │ │
│ ▼ │
│ ┌──────────────┐ ┌──────────────┐ ┌──────────────┐ │
│ │ HEAR IT │◀────── │ SPEAK IT │◀────── │ TRANSLATE │ │
│ │ (their lang) │ │ (TTS) │ │ your provider│ │
│ └──────────────┘ └──────────────┘ └──────────────┘ │
│ │
│ Nothing persisted — the turn exists only long enough to be spoken. │
└─────────────────────────────────────────────────────────────────────────┘
Two-Device Pairing
Person A's phone Person B's phone
┌───────────────┐ 6-char code ┌───────────────┐
│ Start session │ ────────── A1B2C3 ───────────▶ │ Join session │
└───────┬───────┘ └───────┬───────┘
│ live session channel │
└───────────────────┬──────────────────────────────┘
▼
┌──────────────────────┐
│ Each turn relayed & │
│ played in the OTHER │
│ person's language │
└──────────────────────┘
Guardrails: same-business members only · exactly two participants ·
a third person holding a leaked code is rejected · session expires in 2h.
Key Features
🎙️ Speech-to-Speech Interpretation
The listener hears the translation spoken aloud in their language — the whole point of an interpreter, not a text box.
| Feature | Description |
|---|---|
| Push-to-talk turns | Hold to speak, release to send; the other side hears it spoken back |
| Spoken playback | Each turn is synthesized as audio in the listener’s language, not just shown as text |
| On-screen conversation log | Both the original and the translation appear in a running log for the session (in-tab only) |
| 15 languages | English, Spanish, French, German, Portuguese, Italian, Chinese, Japanese, Korean, Arabic, Hindi, Russian, Vietnamese, Tagalog, Polish |
Use Case: A shift manager holds the button and says “Please move the pallets to bay 4 before your break.” The associate hears it in Spanish, holds their own button, and replies — the manager hears the answer in English.
💡 Pro Tip: In single-device mode, set Person A’s language to your own and Person B’s to the language you’re bridging before you start — the “speaks” dropdown on each side is what tells the app which direction to translate.
📱 One Device or Two
| Mode | How it works | Best when |
|---|---|---|
| One device | Two people share a phone; each holds their own on-screen button | You’re standing together and one phone is enough |
| Two devices (paired) | One starts a session, shares a 6-character code, the other joins; each keeps their own phone | Each person has their own phone, or you want hands-free |
Use Case: Two associates on opposite ends of a line each open Live Translate on their own phone; one shares the code, and they hold a back-and-forth without passing a device around.
💡 Pro Tip: Two-device mode is the only one with the hands-free switch — flip it on and you can stop holding the button entirely; the app sends each utterance automatically when you pause.
⚡ Hands-Free (Continuous) Mode
Available in two-device paired mode. Instead of holding a button, toggle “Hands-free — translate continuously as you speak.” The app uses voice-activity detection to auto-detect the start and end of each utterance and sends it on the trailing pause.
| Behavior | Detail |
|---|---|
| Auto-segmentation | Speak naturally; a short pause closes the utterance and sends it |
| Blip filtering | Coughs and clicks below a minimum length are discarded |
| Runaway cap | A single utterance is capped so it can’t record indefinitely |
| Echo suppression | Your mic is muted while the other person’s translated audio plays, so it doesn’t translate the reply back |
💡 Pro Tip: Hands-free needs the other person to have joined first — if you flip it on before your peer connects, it politely waits and turns itself back off.
🔒 Governance & Privacy
The features that make Live Translate approvable, not just usable.
| Feature | Description |
|---|---|
| Your provider | Turns run through your configured translation integration (AWS Translate / DeepL, BYOK), the platform-managed default, or the LLM fallback — resolved in that order |
| Ephemeral by design | Turns are uncached and never persisted; no stored transcripts or audio |
| Tenant boundary | Paired sessions admit only members of the same business, and only the two participants — a third person with a leaked code is rejected |
| Provider named on screen | When a specific translation provider is configured, its name appears in the on-screen privacy line |
| Usage rate limits | Per-user hourly caps protect against runaway cost (interpret/relay turns and session creation are each capped) |
📖 Company Terminology (Glossary)
Admin-managed at Company terminology (/admin/translation). Governs how words translate across Live Translate and the rest of the product.
| Capability | Description |
|---|---|
| Mapping terms | Force a source word to render as a specific target word in a given language |
| Do-not-translate terms | Keep a word untouched in every language (brands, models, drug names) |
| Provider-independent | Enforced by wrapping any engine (protect → translate → restore) — identical behavior on AWS, DeepL, or LLM |
| CSV import / export | Bulk-load and back up your glossary |
| Starter packs | Adopt curated frontline term sets (Warehouse Safety, Retail Ops, Healthcare & Facilities — EN → ES) |
| Quality loop | Flagged translations queue for an admin, who can turn a flag into a glossary term in one step |
💡 Pro Tip: Start with a starter pack for your industry, then add your own site codes and equipment names as do-not-translate terms — those are exactly the words a generic engine mangles.
📊 Admin Analytics (Content-Free)
At Company terminology → Analytics (/admin/translation/analytics), admins get a usage and governance picture without any conversation text — upholding the “nothing stored” promise while still giving an employer what consumer translators can’t.
Standard Analytics:
- Translation, transcription, and spoken-playback counts (7 / 30 / 90-day windows)
- Top target languages by volume
- Success rate and average latency
- Provider breakdown (which engine handled the traffic)
Governance Signals:
- Active glossary term count
- Pending and resolved quality flags
User Roles & Permissions
| Role | Capabilities |
|---|---|
| Employee | Open Live Translate; start or join single-device and paired sessions; use push-to-talk and hands-free; translate across all 15 languages |
| Manager | Everything an employee can do (there is no manager-only interpreter gating) |
| HR / Admin | All of the above, plus Company terminology: create/import/export glossary terms, adopt starter packs, resolve quality flags, and view content-free analytics |
App access itself is governed by platform Access Management on the Live Translate marketplace app — an admin controls who sees the app; everyone who sees it can translate.
How We Compare
Live Translate competes less with translation quality and more on governance: whether the tool is something a compliance team can sanction for sensitive workplace conversations, inside the platform your workforce already uses.
| Feature | MangoApps Live Translate | Google Translate (consumer) | Pocketalk | Consumer AI translators |
|---|---|---|---|---|
| Runs inside your workforce platform (no separate app/device) | ✅ | ❌ | ❌ | ❌ |
| Uses your own configured translation provider (BYOK) | ✅ | ❌ | ❌ | ❌ |
| Ephemeral / nothing stored by default | ✅ | ⚡ | ✅ | ⚡ |
| Company glossary / terminology enforcement | ✅ | ❌ | ❌ | ⚡ |
| Two-device paired mode (shared code) | ✅ | ⚡ | ⚡ | ❌ |
| Content-free admin usage analytics | ✅ | ❌ | ✅ | ❌ |
| Included with the platform (no per-device/seat fee) | ✅ | ✅ | 💰 | ⚡ |
| Legend: ✅ Included | ❌ Not Available | 💰 Paid Add-on / separate purchase | ⚡ Limited or conditional |
Notes: Pocketalk is a strong privacy-first competitor — it’s HIPAA-compliant, keeps history off by default, and offers an enterprise admin panel — but it’s a separate device or standalone app with its own vendor cloud, not something that runs inside your platform through a provider you chose. Google Translate collects non-personal usage data on the consumer app; governed retention/DPA controls require the separate Google Cloud Translation / Translation Hub products.
Why MangoApps Live Translate?
- 🔗 Inside the platform — no new app to roll out, no device to buy; it’s in the same Apps grid as everything else
- 🛡️ Governed, not just private — your provider, your tenant, your glossary, and a content-free audit trail
- 💳 No per-seat cost — included for every tenant; usage rides your existing AI/Voice credit pool
Getting Started
For Employees
- Open Apps → Live Translate (or search “translate” / “interpreter” in global search).
- One phone? Stay on the One device tab — set each person’s language and hold the “Hold to speak” button to talk.
- Two phones? Switch to the Two devices tab. One person taps Start a session and reads out the 6-character code; the other types it in and taps Join.
- Pick your language, hold to speak — the other person hears it in theirs. Grant the microphone permission when your browser asks.
For Managers
- Use it exactly as an employee would — there’s no separate manager surface.
- On the floor, prefer two-device mode with hands-free for a natural back-and-forth without passing a phone around.
For Administrators
- Click Company terminology from inside Live Translate (visible to admins) to open the glossary.
- Add do-not-translate terms for your brands, equipment models, and site codes, and mappings for internal terms — or adopt a starter pack for your industry.
- (Optional) Connect a dedicated translation provider (AWS Translate / DeepL) as an enabled translation integration to route turns through your own engine.
- Check Analytics for content-free usage, language mix, and glossary health.
Best Practices
- ✅ Seed the glossary before rollout — do-not-translate your brands, SKUs, equipment, and site codes so the engine stops mangling them.
- ✅ Pick two-device + hands-free for real conversations — it’s the closest thing to a natural exchange today.
- ✅ Set languages first — the direction of translation is driven by each side’s language dropdown; confirm it before the first turn.
- ✅ Tell your compliance team it’s governed — the selling point over consumer apps is your provider, your tenant, nothing stored — make sure the people who banned Google Translate know this exists.
- ✅ Watch the analytics — rising volume in a language pair is a signal to build out that pair’s glossary.
Frequently Asked Questions
Q: Does it store the conversation?
A: No. Every turn is transcribed, translated, and spoken back in real time, and nothing is persisted — turns run uncached, and there’s no conversation database. Only content-free cost counters (how many calls, how long) are recorded, never the words.
Q: Which translation engine does it use?
A: Your organization’s configured provider. If an admin has connected a dedicated machine-translation engine (AWS Translate or DeepL) it uses that; otherwise the platform-managed default; and if neither is available it falls back to the platform LLM. It’s the same engine you use elsewhere in the product.
Q: Is it real-time, continuous conversation translation?
A: It’s turn-based by default (hold to speak). Two-device mode adds a hands-free toggle that auto-detects when you start and stop speaking and sends each utterance automatically — so it feels continuous. True full-duplex, always-on call interpretation (both people talking over each other) is a later phase.
Q: How do two people on separate phones connect?
A: One taps Start a session and shares the 6-character code; the other joins with it. Only members of the same business can join, and only the two participants — a third person holding a leaked code is rejected, and the session expires after two hours.
Q: Does it cost extra?
A: There’s no per-seat license — it’s available to every tenant. Its AI usage (transcription, translation, and spoken playback) draws from your platform AI/Voice credit pool as you use it.
Q: Can I make it use our company’s own words for things?
A: Yes — an admin manages the Company terminology glossary. Add “do-not-translate” terms (brands, models) and “mapping” terms (force a word to a specific translation), and they apply to every Live Translate turn regardless of which engine runs.
Related Resources
- Apps Overview — the full catalog of MangoApps Workforce apps
- Comms Hub Overview — internal communications that can also be translated for a multilingual workforce
- Lone Worker Overview — another frontline-first safety app for deskless teams
Live Translate — the interpreter your compliance team can finally say yes to.