Thursday, August 27, 2026
Tech Beat
Aug 26, 2026, 7:19 PMArtificial Intelligence

Google Gemini 3.5 Transcribe Cuts Voice Input Latency 70%

Google's Gemini 3.5 Transcribe cuts speech-to-text latency by 70%, lowers errors to 5.5%, supports 85 languages, and expands across Google apps and Chrome.

Listen to this briefingAudio briefing

Summary

With Gemini 3.5 Pro still unreleased, Google announced Gemini 3.5 Transcribe on August 26, 2026. The model turns speech into polished text by removing fillers, revising spoken corrections and using custom vocabulary for specialized jargon. Google says it is about 70 percent faster from speech to final text than Chirp 3, with a 5.5 percent live-speech error rate versus 7.32 percent. It supports 85 languages and up to three speakers in prerecorded audio.

Gemini 3.5 Transcribe already powers Gboard’s Rambler, currently limited to Pixel 11. On August 26, it also reached the Gemini app on macOS, Antigravity with permissioned access to screen context and chat history, AI Studio’s build model for AI-assisted app development, and the Gemini API. Google plans to expand Rambler to more Gemini Intelligence devices later in 2026 and add Chrome support soon for voice input in any web field. Faster, cleaner dictation could benefit consumers and developers, but the model changes spoken wording, making it unsuitable where exact transcripts matter.

Positives

  • 70 percent faster processing reduces the wait between speaking and receiving final transcribed text.
  • 5.5 percent live-speech error rate improves on Chirp 3’s 7.32 percent.
  • 85 supported languages and three-speaker prerecorded transcription broaden the model’s practical reach.
  • August 26 access through macOS, Antigravity, AI Studio and the Gemini API extends the model beyond Pixel 11.
  • Custom vocabulary, filler removal and automatic correction can produce cleaner text containing specialized jargon.

Risks & concerns

  • 5.5 percent error remains possible, despite improving on Chirp 3’s 7.32 percent.
  • AI cleanup changes the speaker’s wording, creating risks when verbatim transcription is required.
  • Gboard’s Rambler remains limited to Pixel 11 until expansion to more Gemini Intelligence devices later in 2026.
  • Chrome support has no firm release date beyond Google’s promise that it is coming soon.
Primary sourceAI - Ars Technicahttps://arstechnica.com/ai/2026/08/google-announces-gemini-3-5-transcribe-for-ai-powered-speech-to-text/
Read full article
Editorial note: Tech Beat summarizes and analyzes third-party reporting. The source link is the authoritative article. This page does not reproduce the full source text.

More From The Wire

Artificial IntelligenceAug 27

OpenAI Brings ChatGPT Ads to India With 50 Brands, ₹725 Daily Floor

Artificial IntelligenceAug 27

Nvidia Nears $12.9 Billion Hugging Face Acquisition Amid Conflicting Reports

Artificial IntelligenceAug 27

OpenAI Expands Brazil Presence to Support Nationwide AI Adoption