Tantalum

Requirements

• Transcribing is very human resource intensive (an hour of audio requires 4–8 hours of effort), so need faster transcribing without losing accuracy
• Translation is very challenging, and is even more resource intensive
• Lack of human resources create significant delays getting transcriptions
• Need verbatim transcriptions and commercial solutions scrub filler words such as um, ah, mhmm, uh-uh, uh from their results
• All transcriptions need to be verified before they are sent to Crown, but there is no simple verification process to facilitate this policy
• Documents need to be translated to English and from English to foreign for legal docs like MLATs

Core capabilities

• Fast and accurate audio and video transcriptions and translations
• Fast and accurate text-to-text translation to and from English
• Batch audio file submission for voice texts, etc., carved from devices
• Low cost 24/7 transcribing
• Content extraction (e.g., from audio or video files)
• Transcriptions include filler sounds, affirmations and negations that reaches human-level transcription accuracy
• Auto language detection if language is unknown (60+ languages)
• Integrated, feature-rich editor for text verification with variable speed playback and a "Karaoke" mode that highlights the relevant word as it is spoken in the audio
• Diarization of up to 10 speakers in a mono-channel recording

Advanced features

• Text translation (contextual translation of non-English transcript)
• Automatic punctuation, capitalization and number formatting
• Auto identification of low confidence words
• Universal and customizable macros to speed up editing
• Playback of the video files while verifying the text
• Workflow management with job queueing and verifier assignment
• Track verification progress within the transcription
• User and administrator accounts with customizable permissions
• Integrated file activity logging
• Outputs machine readable JSON files along with text and PDF files
• Can integrate any ASR engine as new and better ones come along

File verification editor

• Side-by-side audio waveform and transcript text
• Variable speed playback (0.5x to 2.0x)
• Karaoke mode — word-by-word highlight synced to audio
• One-click insertion of speaker labels and timestamps
• Macro shortcuts for common editing patterns
• Low-confidence word flagging with color coding

User status view

• Real-time dashboard of assigned verification jobs
• Progress bars per transcription with percentage complete
• Priority queue with deadline indicators
• Quick-access links to pending, in-progress, and completed work
• Notification alerts for new assignments or returned revisions

Job adding view

• Drag-and-drop batch upload for multiple audio/video files
• Auto-detection of audio format, sample rate, and channel count
• Language selection with auto-detect fallback
• Speaker count estimation and diarization pre-configuration
• Priority and deadline assignment with calendar picker
• Verifier assignment with workload balancing

Exported transcript view

• Clean formatted text output with speaker labels and timestamps
• PDF export with official header and pagination
• Machine-readable JSON with full metadata and confidence scores
• Time-coded segment export for integration with case management systems
• Redacted and unredacted version control

Support

Annual maintenance provides access to:

• Portal (code updates, notifications, reporting, document status, etc.)
• Telephone support
• All updates (save those specifically restricted)
• Security notifications (through encrypted email)

Road map

Available on request.