Sight-Terp: Advanced Computer-Assisted Interpreting

Sight-Terp Assist is the live interpreting module in the Sight-Terp Suite—a web-based CAI platform built for consecutive and simultaneous workflows. Assist acts as a digital boothmate: it captures speech, structures segments, and surfaces translations with visual cues so interpreters can scan, decide, and deliver faster under pressure.

Alongside Sight-Terp Evaluate (AI performance feedback) and Sight-Terp Practice (AI-generated training speeches), Assist completes an end-to-end interpreter workflow: train → interpret live → assess performance—all under one account at www.sightterp.net.

Key Features

• Microphone selection & quick setup — Choose your input device through a streamlined selector with no page refresh required. Switch microphones mid-session without breaking your flow.

• Live transcription & translation — Real-time ASR paired with neural machine translation across multiple languages, powered by Soniox and Microsoft Azure Speech.

• Entity highlighting — Automatically marks names, organizations, dates, numbers, and other entities in both the transcript and machine translation output for faster comprehension.

• Click-to-translate overlay — Highlight any text in the live transcript to instantly view its translation in a popup.

• Speaker diarization — Detects and color-codes different speakers in multi-party speech.

• Glossary upload — Import CSV/XLSX glossaries to improve ASR and MT accuracy with domain-specific terms, background context, and custom key-value metadata.

• Digital notepad — Built-in annotation workspace with tools for underlining, circling, and striking through notes.

• Fullscreen mode — Maximize the workspace for a distraction-free interpreting setup, including booth and remote configurations covered in the integrated Setup Guide.

• Credit-metered sessions — Assist sessions are billed in 3-minute blocks, with remaining credits visible in your account bar at all times.

New Update: A Fresher, More Reliable Suite

We’ve rolled out targeted upgrades across Assist and the broader Sight-Terp Toolbox to make the platform faster, cleaner, and more dependable in real interpreting conditions.

Microphone Selection

Select audio devices instantly using an improved, reload-free picker. Switching between microphones during a session is now more intuitive and responsive—whether you’re on a Bosch console, a USB interface, or a remote setup with virtual audio routing.

Dual-Side Entity Highlighting

Important entities—names, dates, numbers, organizations, and more—are now highlighted in both the source text and the machine translation. This synchronized highlighting supports quick scanning and reduces cognitive load when you’re working under time pressure.

Dark Mode with Adaptive Logo

The interface now offers a dark theme with a matching dark-mode header logo. Branding adjusts automatically based on the selected theme for a unified visual experience during long sessions.

Interface Refinements

A redesigned “Enlarge” button and improved spacing enhance readability and workflow clarity. The overall UI has been polished to improve visual hierarchy and reduce fatigue during extended interpreting tasks.

Accurate Trial & Account Information

A top information bar now displays your profile details along with a clearly formatted trial expiration date, pulled directly from Auth0 app metadata. You always know where you stand—credits, plan status, and trial window—without leaving your workflow.

Updated Subscription Page

A clean, branded subscription interface is now integrated within the app. Users can upgrade to Plus, Pro, or Elite tiers—or purchase one-time credit packs—directly from the pricing section via Paddle checkout. Credits never expire and work across all three apps.

Cleaner, More Secure Runtime

Superfluous debug logs have been removed for smoother operation and less console clutter. Assist now routes Soniox and Microsoft speech services through authenticated server-side proxies, keeping API credentials off the client. The result is a cleaner runtime and a more production-ready experience.

Sight-Terp Practice: Smarter Speech Generation & Playback

Practice has received a major reliability pass—especially for speech synthesis and browser playback.

Gemini-Powered Text & TTS

Practice now uses Google Gemini for both speech generation and neural text-to-speech, with 30 selectable voices and support for 33+ languages. Speeches are tailored by topic, difficulty (Very Easy through Professional/Technical), and intended use (consecutive vs. simultaneous interpreting).

More Reliable Audio Playback

We rebuilt the TTS playback pipeline to handle real-world browser behavior:

• Safari-compatible audio — Gemini returns raw PCM (audio/l16); the app now converts it to WAV before playback, fixing the “Hata”/error state Safari users were seeing.
• Retry strategy — Transient TTS failures (server errors, rate limits, empty responses) trigger up to 3 automatic retries with backoff, aligned with Google’s preview-model guidance.
• Duplicate-call protection — Guards against redundant consecutive synthesis requests so users don’t stack multiple players or download buttons.
• Playback status badge — A live indicator near the player shows Generating → Ready → Playing or Autoplay blocked, so users know exactly what’s happening.
• Auto-download with fallback — Generated speech downloads automatically when the browser allows it; a manual download button remains as backup.
• Built-in diagnostics — Console and server logs ([TTS] ...) help trace mime types, autoplay policy blocks, and provider errors during support.

Together, these changes make Practice feel less like a demo and more like a tool you can rely on for deliberate training sessions—on Safari, Chrome, or anywhere else.

The Full Sight-Terp Toolbox

Sight-Terp Assist is one pillar of a three-app suite designed for professional interpreters, students, and educators:

All three apps share a unified credit system:

• Assist: 1 credit = 3 minutes of session time
• Evaluate: 1 credit = 1 evaluation report
• Practice: 1 credit = 1 generated speech

Plans start free, with Plus from $5.90/month and flexible pay-as-you-go credit packs for users who prefer not to subscribe.

Built for Real Interpreting Work

Sight-Terp is not a generic transcription tool. It is built around how interpreters actually work: segmenting speech under pressure, scanning for entities and numbers, managing glossaries for specialized domains, and moving between preparation, live delivery, and post-session review.

Whether you are preparing for an exam, sharpening skills between assignments, or supporting a live booth with AI-assisted cues, the Sight-Terp Suite gives you one place to practice, perform, and evaluate.

Learn more at www.sightterp.net