Google has released a new transcription feature called Gemini 3.5 Transcribe that the company reports can convert speech to text across 85 languages while automatically correcting verbal stumbles and filler words.
What Happened
Gemini 3.5 Transcribe is designed for use cases including meeting notes, interviews, and content creation workflows. The system processes audio input directly and outputs clean, punctuated text without requiring manual editing of common speech irregularities such as hesitations or repeated words, according to information shared by Google. The feature integrates with Google's broader Gemini ecosystem and is accessible via API.
Why It Matters
For developers building applications that rely on accurate transcription, the ability to handle multilingual input in a single pass reduces complexity in pipelines that previously required separate language detection and correction steps. For end users, the auto-correction of verbal stumbles addresses a common pain point in automated transcription: raw output often contains disfluencies that require manual cleanup before the text can be used.
The Bottom Line
Gemini 3.5 Transcribe expands Google's portfolio of on-device and cloud-based AI tools. Availability and pricing details were not fully detailed in the initial announcement.