Google Releases Gemini 3.5 Transcribe for AI Speech-to-Text
Google has launched Gemini 3.5 Transcribe, a new AI model designed to improve speech-to-text accuracy and speed. The model aims to produce cleaner, edited transcripts by removing filler words and corrections.
Google has introduced Gemini 3.5 Transcribe, an artificial intelligence model intended to enhance speech-to-text capabilities by automatically editing out filler words like "ums" and self-corrections. This new model, part of the Gemini 3.5 family, is designed to deliver polished text outputs from spoken language.
The company states that Gemini 3.5 Transcribe offers significant improvements in both speed and accuracy compared to its predecessor, Chirp 3. Google reports a 70 percent increase in speed from voice input to final transcribed text. Additionally, the live-speech error rate has been reduced to 5.5 percent, an improvement over Chirp 3's measured rate of 7.32 percent.
This technology is already integrated into Google's ecosystem, powering the Gboard "Rambler" feature on Pixel devices. The broader rollout aims to make voice input more efficient and reliable for users across various Google applications and services.
Google's ongoing development in AI speech recognition signals a continued effort to refine user interaction with its products, aiming for more seamless and less error-prone voice-based experiences.