Spoke Class

Features

Complete feature reference for Spoke Class.

Real-time Transcription

Live speech-to-text via Deepgram WebSocket:

  • 12 source languages: English, Chinese, Spanish, French, German, Japanese, Korean, Portuguese, Russian, Italian, Dutch, Hindi
  • Speaker diarization for separating professor and student contributions
  • Auto-punctuation for clean, readable transcripts
  • Low-latency streaming with interim and final results

Live Translation

Real-time translation during recording:

  • 6 target languages: Chinese, Spanish, Japanese, Korean, French, German
  • Side-by-side display of original and translated text
  • Context-aware translation using recent transcript segments
  • 800ms debounce for translation quality

Custom AI Extraction (Intents)

The most distinctive feature of Spoke Class — user-defined extraction prompts:

  • Create custom prompts to extract specific information
  • Link intents to courses so they automatically apply to new recordings
  • Examples: "List key definitions", "Extract homework deadlines", "Summarize main arguments"
  • Intents run during the background processing pipeline

Course Organization

  • Group recordings by course
  • Link multiple intents to each course
  • Browse recordings by date within a course
  • Course-level statistics and usage tracking

Background Processing

  • Asynchronous pipeline: upload → transcribe → extract → complete
  • Real-time progress tracking with percentage display
  • Automatic retry on failure
  • Status visible in the recordings list

Recording Management

  • Browse all recordings with status indicators
  • Full-text search across transcripts
  • Playback with synchronized highlighting
  • Export transcripts and extracted content

On this page