Google today introduced Gemini 3.5 Transcribe, a new speech-to-text model designed for more accurate, context-aware voice interactions. According to the company, the model better handles challenges that traditional speech recognition systems struggle with — such as background noise, specialist vocabulary, and self-corrections — and converts raw audio into polished, formatted text.
Availability and APIs
Gemini 3.5 Transcribe is available to developers via the Gemini API in Google AI Studio and via the Gemini Enterprise Agent Platform. The model is offered through two distinct API modes:
- Real-time streaming: continuous, bidirectional streaming with sub-second latency using the Live API and the gemini-3.5-transcribe-live model.
- Pre-recorded audio processing: transcribes recorded audio (meetings, call logs, etc.) with speaker attribution and word-level timestamps via the Interactions API using the gemini-3.5-transcribe model.
Google has already integrated the model into consumer-facing products such as the Gemini app on macOS and the Rambler feature on Android, and plans further integrations across other Google surfaces.
Key features and performance
- Smart transcription: the model handles self-corrections (e.g., "let’s meet Tuesday—no, Wednesday"), removes filler words (such as "ums" and "ahs"), and auto-formats the output.
- Function calling: Gemini 3.5 Transcribe can delegate complex tasks to other Gemini models via function calls (for example, image generation or file analysis); Google notes this functionality is currently available in the Gemini macOS app.
- Improved accuracy: Google cites Artificial Analysis measurements showing an average Word Error Rate (WER) of 4.0% for streaming use and 2.6% for non-streaming use cases. The company also reports a 70% improvement in time to final transcription compared with the prior Chirp 3 model. On the FLEURS benchmark across selected top languages and locales, the model delivered 5.50% WER in streaming mode and 5.04% WER in non-streaming scenarios.
- Custom vocabulary: the model adapts to specialized jargon and unique spellings provided by developers.
- Global language support: automatic detection and transcription for more than 85 languages, with handling of regional accents and dialects.
- Multi-speaker identification: for pre-recorded audio, the model attributes speech and provides word-level timestamps for up to three speakers (support for 3+ speakers is experimental).
Use cases and integrations
Google positions Gemini 3.5 Transcribe as suitable for voice agents, real-time captioning, and post-call analytics pipelines. Several developer platforms using the Gemini Live API — including Agora, Fishjam, LangChain, LiveKit, Pipecat, Vercel, and Vision Agents — have integrated real-time media streaming support so developers can focus on user experience.
Concrete integrations and product features highlighted by Google include:
- Gboard on Android: the Rambler feature converts spoken thoughts into well-formatted text, filters filler words, and supports voice-driven edits.
- Google Antigravity: with user permission, the model can use screen context and chat history to improve transcription accuracy for file names, agent notes, and active documents.
- Google AI Studio: access to 3.5 Transcribe in Build mode for voice-driven code and app development.
- Gemini app on macOS: transcribes natural speech into cleaned, formatted text and enables voice commands that combine with screen context to run complex workflows; background calls to other Gemini models allow tasks like file summarization, cross-app text repurposing, or image generation at the cursor via voice.
- Chrome: Google says voice-to-type in any web field will be coming soon.
Early feedback and rollout
Developers and companies such as Vivo, Intellitek Health, and Lingopal have reportedly given positive feedback on latency, accuracy, and language coverage. The Live API is already being used by multiple platforms to build voice-driven interfaces.
Availability:
- For developers: public preview in the Gemini API via Google AI Studio and Google Antigravity.
- For enterprises: public preview via the Gemini Enterprise Agent Platform, with Gemini Enterprise for Customer Experience coming soon.
- For general users: the Gemini app on macOS in English, Rambler on Android in select countries and languages, and a forthcoming integration in Chrome.
Google presents Gemini 3.5 Transcribe as a step toward making voice interactions feel more natural and productive by combining improved transcription accuracy, context awareness, and deeper integrations across Google products.
