Gemini 3.5 Transcribe

Google's monthly AI roundup blog for August 2026 explicitly names Gemini 3.5 Transcribe among the month's biggest AI updates, alongside Gemini 3.7 Flash and the Pixel 11 series of devices. The roundup, published September 1, 2026, confirms the model as a distinct August release within the Gemini family rather than a th The supplied excerpt is a broad monthly recap rather than a focused technical deep-dive on Gemini 3.5 Transcribe, and it bundles model news with consumer hardware (Pixel 11) and product rollouts such as Gemini in Chrome on Android and hands-free voice tools across Google Workspace and Gemini Live. The Gemini app offici

Google introduced Gemini 3.5 Transcribe on August 26, 2026, as a speech-to-text model that turns raw audio into formatted text with speaker attribution and word-level timestamps on pre-recorded files. Google's announcement describes it as "our most precise speech-to-text model yet, designed for intelligent voice intera The model is offered via two separate APIs with a key split: `gemini-3.5-transcribe-live` provides continuous bidirectional streaming with sub-second latency through the Live API, while `gemini-3.5-transcribe` handles pre-recorded audio, meetings, and call logs through the Interactions API. Speaker attribution on the p

Gemini 3.5 Transcribe is Google's first Gemini-branded model built specifically for speech-to-text, replacing the prior pattern of either using Google Cloud's Speech-to-Text API (the Chirp line) or handing audio files to a general Gemini chat model and asking it to transcribe. Per the supplied page excerpt, Google desc The eesel write-up notes a documentation inconsistency worth flagging: the models overview page tags Gemini 3.5 Transcribe as "New Stable," while the launch blog post says it is "in public preview" for both developers and enterprises, and it is already shipping in consumer products including Rambler on Gboard for Andro

Google shipped Gemini 3.5 Transcribe on August 26, 2026, announcing it in a blog post titled "Intelligent transcription with Gemini 3.5 Transcribe" as "our most precise speech-to-text model yet, designed for intelligent voice interactions." The release landed less than two weeks after Google was still previewing Gemini According to the supplied page excerpt, the non-streaming endpoint posts a 2.6% average word error rate, with support for more than 85 languages and a feature set that pulls diarization, word-level timestamps, and custom vocabulary biasing into a single API call. Google's developer documentation describes the model as

Google released Gemini 3.5 Transcribe as a dedicated speech-to-text model, shipping it as two distinct endpoints rather than one. The `gemini-3.5-transcribe` endpoint handles pre-recorded files through the Interactions API, while `gemini-3.5-transcribe-live` provides bidirectional streaming through the Live API, and th The release targets real-time voice interfaces and recorded audio, with automatic language detection covering more than 85 languages including mid-sentence code-switching. The model is API-only with no open weights and no self-hosted path, positioning it as a managed-service decision; solo developers and startups can b

Traceary, a third-party Vercel changelog aggregator, records the August 26, 2026 entry "Gemini 3.5 Transcribe now available on AI Gateway" with the description noting support for recorded and live audio via google/gemini-3.5-transcribe. The same rolling changelog surfaces adjacent same-day releases including Ling 3.0 F As an aggregator Traceary's content is derivative of vercel.com/changelog and the supplied excerpt is truncated mid-sentence, so it functions only as corroborating evidence of the August 26, 2026 release date and the recorded/live audio capability split. It should not be treated as a primary source for pricing, technic

Ars Technica reported on August 26, 2026 that Google announced Gemini 3.5 Transcribe as a new speech-to-text model in the Gemini 3.5 branch, designed to streamline voice input by editing out disfluencies such as "ums" and "uhs" and outputting polished, AI-refined text. The article cites Google's claim that the new mode According to Ars Technica, Gemini 3.5 Transcribe is already live in the Rambler feature of Gboard (currently limited to Pixel 11 phones) and has begun powering voice input in the Gemini app on macOS starting the same day as the announcement. Google stated that Rambler will expand to "more Gemini Intelligence devices" l

Vercel's official "Speech to Text" modality docs page names google/gemini-3.5-transcribe as a supported transcription model for pre-recorded audio alongside openai/whisper-1 and openai/gpt-4o-transcribe, with use cases including voice notes, call recordings, and podcasts. It documents the AI SDK integration using exper For live audio, the doc describes experimental_streamTranscribe, which opens a WebSocket via AI Gateway and streams partial transcripts back, naming google/gemini-3.5-transcribe-live among the supported streaming models alongside openai/gpt-realtime-whisper and spacexai/grok-stt. It also outlines a browser-safe pattern