Speech Recognition & AI
Custom Phrase Auto-Generation
Upload a PDF, text file, or paste raw text — VoicePing automatically extracts domain-specific vocabulary and suggests custom phrases to improve recognition accuracy. The system identifies words that the ASR model is likely to misrecognize and adds them to your custom phrase dictionary. Supports all languages.Improved Hallucination Suppression
Reduced common transcription artifacts (e.g., repeated filler text) through improved training data filtering and inference tuning, resulting in cleaner transcription output across all languages.Better Short Utterance Detection
Improved bilingual mode stability for very short utterances — brief responses are now more reliably detected in the correct language.Custom Phrase Capacity Increased
The maximum number of custom phrases per workspace has been increased from 500 to 1,000, with minimal impact on recognition speed.Meeting Log & Transcript Improvements
Unified Transcript View
The meeting log list (with keyword search introduced in January) has been redesigned into a single unified view, adding:- Date range filtering, tag filtering, and creator filtering
- Sortable results with shareable filter URLs via query parameters
- Available on both web and mobile
Bulk Transcript Download
Select and download multiple transcript files directly from search results — ideal for enterprise data export workflows.Speaker Reassignment
Expanded the speaker merge feature (introduced in January) with a full reassignment UI:- Reassign speaker labels across different transcript groups
- Bulk relabel all segments from one speaker to another
Copy-to-Clipboard
One-click copy button on the meeting log page to copy the full transcript text including summary, for quick pasting into external tools.SRT Download
Meeting transcripts can now be downloaded in SRT (subtitle) format for use in video editing and subtitle workflows.Meeting Log Language Default Fix
The meeting log language selector now correctly defaults to the user’s display language instead of always defaulting to English.Mobile App Updates
- Bilingual mode fix: Critical mobile bilingual mode bug resolved via expedited app store update
- Bilingual mode UI: Improved button layout for smaller mobile screens
- Unified transcript list: Mobile now uses the same search and filter interface as the web app
- Nine new languages: All newly added ASR languages available on mobile
- Offline transcription speed: Significant performance improvements for on-device transcription on both iOS and Android
Desktop App Updates
- Full-screen mode: Audio extraction (transcription) mode now supports full-screen for a better presentation experience during live events
- Configurable message duration: Minimum message duration reduced to 2 seconds for faster subtitle display in real-time transcription
Subtitle & Caption Improvements
- Line-break fix: Resolved issues where subtitle text broke awkwardly across lines in embedded video files
- Subtitle positioning: Moved subtitle text slightly up to avoid overlapping with the system taskbar
- English font size fix: Corrected font sizing for English text, numbers, and special characters in subtitle rendering
UI/UX Improvements
- Group and project CSV export: Export group and project member assignments as CSV for department cost allocation and bulk management
- User feedback collection: Enhanced signup flow to collect persona and feedback data, including a free-form comment field
- Japanese localization: Completed Japanese translations for transcript list headers, filter labels, and navigation components
- Email preference management: New email preference system accessible from email footers — manage marketing and system email subscriptions without logging in
Bug Fixes
- Fixed translation display for Chinese: Translations from certain languages to Chinese were not showing due to a language code mismatch — now resolved
- Fixed endless summary status: Meeting log summaries could get stuck in “creating” state — improved timeout handling
- Fixed tokenizer loading: Extended vocabulary tokens were not loading correctly for new languages, causing empty transcription results
- Fixed Hindi mono-language mode: Hindi transcription sometimes returned English even with bilingual mode disabled — now correctly forced to Hindi