Gemini 3.5 Transcribe brings emotion detection and speaker ID to speech-to-text

Google has released Gemini 3.5 Transcribe, a speech-to-text model that advances accuracy alongside new capabilities for identifying speakers and detecting emotional tone in audio. The tool supports real-time streaming and can attribute speech to multiple speakers within a single recording. These features position the model as a potential competitor to existing transcription solutions across industries that rely on audio processing.

The addition of emotion detection and speaker identification represents a step beyond basic transcription toward more contextual audio understanding. Industries handling substantial audio workloads - from media to customer service to research - could integrate these capabilities to extract richer data from recordings without additional processing steps.

• Gemini 3.5 Transcribe combines improved accuracy with speaker attribution and emotion detection in a single real-time model.
• The feature set targets industries that process significant audio data, potentially shifting how transcription integrates into existing workflows.
• Competitive pressure in speech-to-text is intensifying as models add analytics capabilities beyond basic word-for-word conversion.