The latest articles on DEV Community by Kamal (@kamal_llm_manuplator). https://dev.to/kamal_llm_manuplator https://media2.dev.to/dynamic/image/width=90,height=90,fit=cover,gravity=auto,format=auto/https:%2F%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Fuser%2Fprofile_image%2F4012045%2F5d6d8ad7-f724-45d4-803f-ef76633aef9e.png https://dev.to/kamal_llm_manuplator en Kamal Fri, 03 Jul 2026 02:58:18 +0000 https://dev.to/kamal_llm_manuplator/nano-banana-2-lite-and-gemini-omni-flash-whats-actually-new-in-googles-gemini-api-3hng https://dev.to/kamal_llm_manuplator/nano-banana-2-lite-and-gemini-omni-flash-whats-actually-new-in-googles-gemini-api-3hng <p>Google added two new models to the Gemini API today: Nano Banana 2 Lite (image generation) and Gemini Omni Flash (video generation + editing). Neither is the Gemini 3.5 Pro release people have been waiting for, so it's easy to miss. Here's what's actually in them.</p> <p><strong>TL;DR</strong></p> <ul> <li>Nano Banana 2 Lite: <code>gemini-3.1-flash-lite-image</code> = text-to-image in ~4s, $0.034/1K images</li> <li>Gemini Omni Flash: <code>gemini-omni-flash-preview</code> = video gen + conversational editing, $0.10/sec </li> <li>Both are built to be chained: generate an image fast, then animate it into video </li> <li>Neither model is positioned as a quality upgrade = both are cost/speed plays</li> </ul> <p><strong>Nano Banana 2 Lite</strong></p> <p><strong>Model ID: <code>gemini-3.1-flash-lite-image</code></strong></p> <ul> <li>Text-to-image output in about 4 seconds</li> <li>$0.034 per 1K-resolution image</li> <li>Positioned as the direct replacement for the original Nano Banana (<code>gemini-2.5-flash-image</code>) - if you're on that model, this is a drop-in upgrade</li> <li>Available in Google AI Studio, Gemini API, Gemini Enterprise Agent Platform, and consumer surfaces (Search AI Mode, Gemini app, Photos, NotebookLM, Flow, Google Ads)</li> </ul> <p><strong>Gemini Omni Flash</strong></p> <p><strong>Model ID: <code>gemini-omni-flash-preview</code></strong></p> <ul> <li>Public preview in Google AI Studio and the Gemini API</li> <li>Conversational editing - refine a generated video using plain-language instructions instead of re-prompting from zero</li> <li>Multimodal referencing - combine text, image, and video inputs to keep a scene consistent</li> <li>$0.10 per second of video output (same rate as Veo 3.1 Fast)</li> </ul> <p><strong>Known limitations right now</strong></p> <ul> <li>Generations capped at 10 seconds</li> <li>No audio reference uploads yet</li> <li>No scene extension yet</li> <li>Video references under 3 seconds are accepted by the API schema but not correctly processed yet</li> <li>Character consistency across scene changes/pans still has rough edges</li> </ul> <p>Google says longer durations are coming. The part worth paying attention to: chaining them </p> <ol> <li>Generate an image with Nano Banana 2 Lite (fast, cheap)</li> <li>Pass that image as a reference into Omni Flash</li> <li>Omni Flash animates it into a video</li> </ol> <p>Both models are optimized for throughput and cost, not for topping a quality benchmark. If you're running high-volume image or video generation and speed/price matter more than peak output quality, these are worth testing. If you need top-tier quality, Nano Banana Pro is still the model for that. Has anyone here built the chained image-to-video workflow yet? Curious how the multi-turn editing holds up in practice.</p> aivideos nanobanana googleomniflash googlegemini