One desktop ASR workspace for real-time interpreting
InterpretBank ASR is a real-time speech recognition assistant inside the installable InterpretBank desktop app for macOS and Windows. It supports simultaneous and dialogic interpreting with live transcription, terminology support, and on-screen reading tools designed for professional workflows.
The same desktop application can run offline models locally on your computer or cloud models under GDPR-compliant processing. InterpretBank does not store audio or transcripts and does not use your data for AI training.
For on-the-go use, the InterpretBank WebApp includes a simple speech recognition feature. It is useful for quick mobile transcription, but it does not provide the desktop ASR workspace, glossary matching, number highlighting, or advanced interpreter controls.
Cloud and offline at a glance
Cloud model: broader language coverage and highest quality through secure server-side processing.
Offline model: local speech recognition for maximum confidentiality because audio stays on the interpreter's computer.
Technical guide: ASR documentation
Desktop ASR features
- Low-latency transcription with selectable cloud or offline speech recognition models.
- Integration of own glossaries to get automatic translations in real time.
- Terms can be added to the session glossary on the fly in-session.
- Loaded glossary can be viewed instantly at any time.
- Numbers automatically highlighted in running transcription.
- Automatic translation of any portion of transcript by highlighting it.
- Font size and line spacing freely adjustable in the reader view.
- For most languages support includes matching of all forms of a term (for example plurals).
Cloud model languages
Albanian, Arabic, Armenian, Bulgarian, Catalan, Chinese, Croatian, Czech, Dutch, English, Finnish, French, German, Greek, Hungarian, Italian, Japanese, Korean, Norwegian, Polish, Portuguese, Russian, Slovenian, Spanish, Turkish, Vietnamese.
Offline model
- On-device speech recognition with maximum confidentiality: no audio leaves the interpreter's computer.
- Fast local processing with selected language models.
- Real-time glossary-based terminology assistance.
- Automatic highlighting of key items such as numbers and terms.
- Indefinite use included in subscription or perpetual license with valid PRO pack (no time-based usage restriction).
- Average 8% WER in typical conditions.
Offline model languages
English, French, German, Chinese, Russian, Spanish, Italian, Arabic, Japanese.
Summary: use the cloud model for maximum coverage and quality, or the offline model when local processing is required.
Price models.
InterpretBank ASR is available inside the installable macOS and Windows app with these model options:
- Cloud: time-based with a credit system to be purchased upfront; from 3 Euros/hr. Highest cloud transcription quality, GDPR-compliant processing, and broad language coverage. You can buy ASR Premium credits here (requires an active InterpretBank Subscription or Perpetual License with valid PRO pack).
- Offline: included in active subscription or perpetual license, with no time restrictions for use. Ideal when maximum confidentiality is required because processing remains local on the computer.
Data retention: InterpretBank never saves audio or transcription data and never uses it for AI training. Cloud processing is handled in real time under GDPR-compliant conditions. Offline processing stays on-device.
For technical details on how to use InterpretBank ASR, refer to our handbook here.
Desktop ASR model comparison
| Criteria | Desktop ASR Cloud Standard |
Desktop ASR Cloud Premium |
Desktop ASR Offline Model |
Typical competitor ASR tools |
|---|---|---|---|---|
| Quality | ✅ Good quality | ✅ Highest quality | ✅ Good local quality | ⚠️ Often generic quality claims not tuned for interpreting tasks |
| Language coverage | English, French, Spanish, German, Italian, Albanian, Arabic, Armenian, Dutch, Bulgarian, Catalan, Chinese, Cantonese, Croatian, Czech, Finnish, Greek, Japanese, Hungarian, Polish, Portuguese, Slovenian, Norwegian, Korean, Turkish, Vietnamese | English, French, Spanish, German, Italian, Albanian, Arabic, Armenian, Dutch, Bulgarian, Catalan, Chinese, Cantonese, Croatian, Czech, Finnish, Greek, Japanese, Hungarian, Polish, Portuguese, Slovenian, Norwegian, Korean, Turkish, Vietnamese | English, French, German, Chinese, Russian, Spanish, Italian, Arabic, Japanese (multilingual model coming soon) | ⚠️ Coverage may be broad but terminology support is usually generic |
| Security | ✅ High security, no retained audio/transcript | ✅ High security, no retained audio/transcript, EU servers | ✅ Maximum privacy: no audio leaves device | ❌ Policies are often unclear or not built for strict interpreting confidentiality |
| Processing mode | Server-side cloud processing | Server-side cloud processing | Fully on-device local processing | Usually cloud-first with limited local alternatives |
| Interpreter workflow integration | ✅ Full InterpretBank integration (glossaries, transcript tools, layout controls) | ✅ Full InterpretBank integration (glossaries, transcript tools, layout controls) | ✅ Full InterpretBank integration (glossaries, transcript tools, layout controls) | ❌ Often standalone dictation/transcription apps without interpreter-specific workflow support |
| Use duration model | Unlimited usage | Credit-based | Unlimited usage | ⚠️ Commonly limited by usage caps, tier locks, or changing fair-use rules |
| Best fit | Everyday cloud transcription on desktop | Maximum coverage and quality | Maximum privacy and local control | General meetings, not high-stakes interpreting workflows |