Features

Complete feature reference for Spoke Translator.

Real-time Transcription

Deepgram WebSocket streaming with optimized connection:

• Auto-detect source language or select from 37+ languages
• Pre-connect WebSocket for minimal startup latency
• Auto-start recording when the translate screen opens
• Noise handling optimized for real-world environments
• Auto-punctuation for readable output

Real-time Translation

Instant translation alongside transcription:

• 37 target languages — the broadest language support across all Spoke apps
• Low-latency delivery via streaming Server-Sent Events
• Context-aware translation using recent conversation segments
• 800ms debounce for translation quality
• Side-by-side display of original and translated text

Smart Replies

AI-powered conversation suggestions:

• Context-driven suggestions based on the conversation history
• Responses generated in the source language
• Personalized using your user profile (full name, nickname)
• Toggle on/off during recording
• Runs via Supabase Edge Function

Speaker Inference

AI identification of conversation participants:

• Uses conversation context to identify speakers
• Integrates your user profile for self-identification
• Participant names from speaker settings enhance accuracy

Session Recovery

Automatic checkpoint system for reliability:

• Recording state checkpointed to local filesystem
• Transcripts and translations saved incrementally
• Resume interrupted sessions on app reopen
• No data loss from unexpected interruptions

Recording Features

• Auto-start — recording begins when the screen opens
• Keep-awake — screen stays on during active recording
• Phase tracking — idle → recording → paused
• Haptic feedback on interactions
• Low credit warning with real-time balance tracking
• Translation history — browse past sessions

History

• Browse all translation sessions
• View full transcripts with translations
• Search past sessions by content