●OMNI — The gemini-omni-flash-preview endpoint retires on September 30. Nineteen days left, which makes this the last realistic window for a migration piece●MIGRATE — Its replacement, gemini-omni-1.1-flash, went GA on August 27. An extend task continues a clip, and two images let the model interpolate between them●RESOLUTION — video_config now takes a resolution of 360p, 720p, 1080p or 4k. Worth stating plainly that 1080p and 4K are produced by upscaling●SAMPLING — temperature, top_p and top_k have been folded into the deprecated list. If you want less variance, the lever is now your prompt and structured output●TRANSCRIBE — gemini-3.5-transcribe covers utterance-level language detection across 85+ languages, speaker diarization, word-level timestamps, and up to 1,000 custom terms●FLASH — Gemini 3.8 Flash reached GA on September 2, and Lyria 3.5 opened in public preview a day later, generating full songs from text and image input●OMNI — The gemini-omni-flash-preview endpoint retires on September 30. Nineteen days left, which makes this the last realistic window for a migration piece●MIGRATE — Its replacement, gemini-omni-1.1-flash, went GA on August 27. An extend task continues a clip, and two images let the model interpolate between them●RESOLUTION — video_config now takes a resolution of 360p, 720p, 1080p or 4k. Worth stating plainly that 1080p and 4K are produced by upscaling●SAMPLING — temperature, top_p and top_k have been folded into the deprecated list. If you want less variance, the lever is now your prompt and structured output●TRANSCRIBE — gemini-3.5-transcribe covers utterance-level language detection across 85+ languages, speaker diarization, word-level timestamps, and up to 1,000 custom terms●FLASH — Gemini 3.8 Flash reached GA on September 2, and Lyria 3.5 opened in public preview a day later, generating full songs from text and image input