Translation App Docs
Translation App technician handbook

Run a live translation session.

A short operator guide for understanding the session model, setting up the app, checking the live views, and monitoring the AI translation engine during an event.

Standalone and managed

The Translation App can run with two operational models. The live workflow is similar, but ownership of session metadata and AI credentials changes.

Standalone

Metadata stays on the machine

Session metainformation is saved locally, usually in the app database on the computer running the session.

  • AI credentials are configured locally.
  • The client can choose the AI providers they prefer.
  • Good fit when the installation must keep operational data on site.
Managed

Metadata is stored on our server

The server provides session storage and credentials. The operator does not need to configure API keys.

  • AI credentials are provided by the server.
  • Profiles and available providers are controlled centrally.
  • Good fit for hosted deployments and centrally managed teams.

Signal Flow

Think of a session as a simple signal path. The operator's job is to confirm each stage is receiving, producing, and delivering the expected signal.

1

Input

Choose the microphone, system audio, interface, or virtual audio device that carries the source speech.

2

AI engine

The selected profile turns speech into translated captions and, when enabled, translated audio.

3

Outputs

Each target language can be routed to local audio devices, browser receivers, captions, or monitoring views.

4

Monitoring

Watch connection state, signal, receivers, and service health while the event is live.

Views

The app is organized around a few operator views. Screenshots will be added here so this page can become a quick visual guide.

Dashboard

Start from the dashboard to create a new session, open an existing one, or review the main workspace entry points.

Session management

Use session management to find, open, and review sessions. In managed mode this reflects server-side session metadata; in standalone mode it reflects local metadata.

Session configuration

Configure the session title, languages, AI profile, credentials model, receiver look, and event-specific customisation before going live.

Live

Run the live session from this view: check routing, connect delivery, monitor input and output status, and keep the translation flow under control.

Usage

Review duration and service consumption for sessions when usage reporting is enabled for the workspace.

End-to-end setup

  1. Create or open the session.
    Confirm the session title, source language, and target languages. These are the main pieces the audience and operator will recognize during the event.
  2. Select the AI profile.
    Choose the profile licensed for the event and appropriate for the latency, quality, compliance, and customisation requirements.
  3. Choose the audio input.
    Select the microphone, sound card, system audio, or virtual audio device carrying the source speech. Speak or play test audio and verify that the input signal is active before connecting delivery.
  4. Choose the broadcast mode.
    Use local broadcast when the translated audio is going into an on-site audio system, booth setup, mixer, or virtual local device. Use cloud broadcast when listeners should join over the internet from their phones or laptops.
  5. Test one receiver or output.
    For cloud broadcast, open the receiver link on a phone, choose a language, connect, and confirm audio or captions arrive. For local broadcast, listen to the selected output device or mixer channel.
  6. Go live and monitor.
    Keep the status indicators visible and stop the session cleanly when the event ends.

Configuration

These are the main configuration areas an operator can adjust before a session. Some deployments may hide or lock options depending on the license and operating model.

AI profile

The profile selects the translation engine behaviour. The license decides which profiles are available.

  • Fast: prioritizes very low translation latency.
  • Intelligent: enables a more sophisticated translation experience and fuller customisation, with slightly higher delay.
  • Other profiles may be available for specific providers, languages, or deployment policies.

Customisation

Customisation gives the AI engine context for the event. It is useful for names, terminology, acronyms, product names, formatting preferences, and words that should be translated in a specific way.

  • The first section is for terminology and reference items that should guide recognition and translation.
  • The second section is for broader instructions, style, context, and event-specific notes.
  • Entries can be imported and exported so operators can reuse prepared material across sessions.
  • Use the expected table/list format shown in the editor. If material is not imported, entries can be added manually.
  • For English source documents, the first list can be populated automatically from documents; otherwise it must be prepared manually.

Receiver look

The receiver look controls what mobile and browser listeners see when they open the session link.

  • Set the audience-facing title and available language list.
  • Add an icon or logo when the deployment allows event-specific branding.
  • Add extra information for participants, such as event instructions, sponsor notes, venue details, or advertising.
  • The info field uses an HTML editor, so operators can format text, links, and simple rich content without editing code.

Broadcasting

Broadcasting decides how translated audio and captions are delivered.

  • Local: sends translated output to local devices, mixers, booth systems, virtual cables, or room audio. It does not add cloud delivery usage.
  • Cloud: broadcasts over the internet so users can join from phones or laptops. It adds app usage because internet delivery has to be handled by the service.
  • Local and cloud delivery can be combined when the event needs both an on-site system and mobile receivers.
  • For cloud delivery, check the technical notes for bandwidth requirements and network planning.

Audio routing

Audio routing maps the source and translated languages to the correct devices and delivery paths.

  • Select the input that carries the source speech: microphone, interface, system audio, or virtual device.
  • For local delivery, assign translated languages to local output devices or channels.
  • For cloud delivery, make sure the receiver/broadcast connection is active and test it from an external device.
  • Use a combination of local and cloud routing when the same event needs physical room audio plus online/mobile listeners.

Recording and usage

Recording is saved locally on the operator machine. Enable it only when the event workflow requires a local record of audio or captions.

  • Activate recording from the Recording view before or during the session.
  • Confirm the save location and make sure the machine has enough free disk space for the expected event duration.
  • Stop recording cleanly at the end of the session so files are finalized correctly.
  • Usage reporting, when enabled, shows duration and service consumption. Cloud delivery may add usage because internet broadcast resources are being used.

Monitoring

Monitoring is the live confidence layer. It should tell the operator whether source audio is present, the AI services are active, receivers are connected, and outputs are moving.

Operator rule: if audio is missing, check the path in order: input signal, AI engine status, output route, receiver connection, listener volume.

Technical points

Use this as the quick technical reference before a live event or when coordinating with venue IT.

Area Guidance Operator note
Operator workstation 5 Mbps up/down minimum; 10 Mbps+ recommended The higher figure gives more headroom for multilingual voice output, recording, cloud delivery, and network variation. Prefer wired Ethernet when possible.
Audience receivers Approx. 0.1-0.3 Mbps per listener per audio channel Captions-only use is much lower. Cloud delivery must be planned around receiver count, number of languages, and venue Wi-Fi quality.
Network access Outbound TCP 443; HTTPS and WSS Corporate networks and venue firewalls must allow HTTPS and WebSocket traffic to the app server and enabled provider endpoints.
Audio input Microphone, professional interface, system audio, or virtual device Microphone permission is required. Windows may expose duplicate device names through different audio APIs; choose the explicit device that matches the physical setup.
Local broadcast Local output devices, mixers, booth systems, virtual cables Local delivery does not add cloud-delivery usage. Confirm the target channel on the mixer or local output before going live.
Cloud broadcast Browser receivers via the configured cloud backend Cloud delivery adds app usage because internet delivery resources are used. Test at least one mobile receiver on the same network conditions expected during the event.
Latency Profile + network + language pair + output mode Latency and stability matter more than peak speed. Voice output and more sophisticated profiles usually add delay compared with captions-only or low-latency profiles.
Recording Local workstation storage Recording is optional and writes locally, commonly under the user's Documents app recordings folder. Check disk headroom before long events.
Data storage Standalone: local metadata; managed: server metadata Event audio and transcripts are not stored by the app server. They are stored locally only when recording is enabled. Managed mode stores operational data such as license, session metadata, configuration, update checks, and usage metering.
Credentials Standalone: local keys; managed: server-provided credentials Standalone local sensitive values are protected in the app configuration. Managed mode can deliver temporary or server-managed credentials according to deployment settings.
License controls Profiles, providers, languages, broadcast options, installations, quotas The license determines which profiles and session capabilities are visible to the operator.