●SHUTDOWN — gemini-robotics-er-1.6-preview retires today, August 31. Any code still pointing at that preview model stops working from here on●DEPRECATION — September 30 is the next date to watch: the gemini-omni-flash-preview endpoint goes away and needs swapping for gemini-omni-1.1-flash, which reached GA on August 27●VIDEO — The Omni 1.1 Flash GA adds video extension through the extend task, and interpolation by passing two images to image_to_video so you can fix the first and last frame up front●VIDEO — A resolution parameter in video_config now accepts 360p, 720p as the default, 1080p, and 4k, with the note that 1080p and 4K outputs are produced by upscaling●SPEECH — gemini-3.5-transcribe reached GA on August 26 with utterance-level language detection across 85+ languages, speaker diarization, word-level timestamps, and up to 1,000 custom vocabulary terms●SPEECH — gemini-3.5-transcribe-live streams both ways over WebSockets on the Live API, with interim and finalized transcription events, a Smart transcription mode, and several VAD settings●SHUTDOWN — gemini-robotics-er-1.6-preview retires today, August 31. Any code still pointing at that preview model stops working from here on●DEPRECATION — September 30 is the next date to watch: the gemini-omni-flash-preview endpoint goes away and needs swapping for gemini-omni-1.1-flash, which reached GA on August 27●VIDEO — The Omni 1.1 Flash GA adds video extension through the extend task, and interpolation by passing two images to image_to_video so you can fix the first and last frame up front●VIDEO — A resolution parameter in video_config now accepts 360p, 720p as the default, 1080p, and 4k, with the note that 1080p and 4K outputs are produced by upscaling●SPEECH — gemini-3.5-transcribe reached GA on August 26 with utterance-level language detection across 85+ languages, speaker diarization, word-level timestamps, and up to 1,000 custom vocabulary terms●SPEECH — gemini-3.5-transcribe-live streams both ways over WebSockets on the Live API, with interim and finalized transcription events, a Smart transcription mode, and several VAD settings