Gemini Omni 1.1 Flash goes GA: video extend, image interpolation

Google's production video model extends existing clips and morphs between two images, reportedly near $0.03/sec; the preview endpoint sunsets Sept 30. Plus Cursor, Codex, Copilot.

Nowline AUG 30 1:00 AM banner

Top AI stories from the last hour

Top AI stories from the last hour

Copy markdown

  • Extend a clip, or morph between two frames

    gemini-omni-1.1-flash is now generally available on the Gemini API. A new video-extension task continues an existing clip at its end, and image_to_video interpolates a transition between two input images. In practice: stitch longer sequences and build image-to-image motion without a preview flag.

  • Four resolutions, but 4K is upscaled

    Outputs support 360p, 720p (default), 1080p, and 4K. Multiple hands-on reports note the 4K (and part of 1080p) is upscaled rather than natively rendered, so treat it as a finishing pass, not true 4K detail.

  • The price: reportedly about 3 cents a second

    Google didn't post pricing in the release notes, but several write-ups peg draft generation near $0.03 per second and say scenes now extend to roughly 40 seconds. Cheap enough to iterate on shot ideas before committing to a final render.

  • Migrate off the preview endpoint by Sept 30

    gemini-omni-flash-preview is deprecated September 30, 2026. If you built on the preview over the summer, point your calls at gemini-omni-1.1-flash now to avoid a hard cutoff.

  • Elsewhere: Cursor spins up an app from a prompt, no repo

    Cursor's new "Start from Scratch" lets a Cloud Agent launch with no GitHub connection: it auto-creates an Origin repo and can publish straight to a live URL via Vercel. Prompt-to-deployed-app in one flow. Note the clock, though: Cursor loses OpenAI-model access on Nov 12.

  • Elsewhere: Codex 0.150 adds @-mentions and Interrupt hooks

    OpenAI's Codex CLI 0.150.0 brings @ mentions to reference tasks, clickable Markdown links, task auto-titling, and new Interrupt hooks; a 0.150.1 patch fixes retained-image handling in token budgets. Security fixes for Windows, Unix, MCP, and Bedrock are included.

  • Elsewhere: Copilot code review defaults to 'Balanced' Sept 28

    GitHub is switching automated code-review effort from Lite to Balanced by default on Sept 28, so reviews go deeper out of the box. Prefer the lighter, faster pass? Select Lite manually before then. The same update reopens Copilot Business/Enterprise sign-ups on Sept 1.