Google has made two new media models available to developers: Nano Banana 2 Lite (official name Gemini 3.1 Flash Lite Image) and Gemini Omni Flash. The company positions these releases as low-cost, high-throughput options intended to enable fast image and video workflows inside apps and services.
Capabilities (input/output)
- Nano Banana 2 Lite: a multimodal model that accepts text and image input and produces images up to 1k resolution, plus text output. Google reports roughly four seconds to generate an image.
- Gemini Omni Flash: a multimodal model that accepts text, image, and video input and returns 720p video with synchronized audio, up to 10 seconds per clip. Google did not disclose generation times for Omni Flash.
Availability and integrations
Both models are available through the Gemini API, Google AI Studio, and the Gemini Enterprise Agent Platform, as well as the Gemini app and Google Flow. In addition, Nano Banana 2 Lite is accessible via AI Mode in Google Search, NotebookLM, Google Photos, Stitch, and Google Ads; Gemini Omni Flash is also available through YouTube Shorts. The developer availability of Gemini Omni Flash follows its consumer-facing rollout by about six weeks.
Pricing
- Nano Banana 2 Lite: $0.034 per 1k-resolution image.
- Gemini Omni Flash: $0.10 per second for 720p video.
Performance and human evaluations
Google reports favorable human-rater results for both models, and both rank highly on Arena.ai community boards. Arena.ai Elo scores reported in Google’s materials are:
- Nano Banana 2 Lite: 1,250 Elo for text-to-image (fifth place).
- Gemini Omni Flash: 1,527 Elo for video generation (first place) and 1,347 Elo for video editing (second place, behind ByteDance Seedance 2.0 at 1,377 Elo).
In Google’s internal human-rater comparisons, Gemini Omni Flash ranked first for overall preference and instruction-following in video editing (504 examples) and on Meta’s public MovieGenBench (1,003 prompts), and tied for first on VBench I2V (355 image-text pairs).
How they work and known limits
Google published model cards for each release describing base architectures and known limitations, but withheld parameter counts and details of the training datasets.
- Nano Banana 2 Lite is a multimodal transformer trained on text and images and runs on the cost-efficient Gemini 3.1 Flash-Lite base model.
- Gemini Omni Flash is a multimodal transformer trained on text, image, audio, and video.
Gemini Omni Flash can produce a native-audio 720p clip from a text prompt, a starting image, or reference images. Its editing capability is conversational: when used via Google’s Interactions API the model preserves session history so each instruction revises the prior clip rather than fully regenerating it, supporting up to three sequential edits.
The image-to-video workflow is a primary integration: a Nano Banana 2 Lite image can be passed as a starting frame to Gemini Omni Flash through the same Gemini API or conversationally in the Google AI Studio web interface.
Background and timeline
The "Nano Banana" nickname began as a placeholder codename when Gemini 2.5 Flash Image launched in August 2025. Google has since extended the family with qualifiers: Nano Banana Pro in November 2025, Nano Banana 2 in February 2026, and now Nano Banana 2 Lite. Gemini Omni Flash first appeared publicly at Google I/O on May 19, initially available to Gemini app subscribers, Google Flow users, and on YouTube Shorts.
Why this matters
Google frames these releases as evidence that generative media is now cheap and fast enough to run at runtime inside apps rather than only as slower, curated production steps, changing unit economics. Ten-second clips can be chained into longer pieces, and developers can automate generation directly inside their apps. Use cases include high-volume digital advertising and social media content production.
Caveats
These models are not positioned for high-resolution, Hollywood-level production: their 1k image limit and 720p/10s video cap restrict professional film use. Their speed and cost advantages may be less significant for hobbyist users, but they can unlock value when combined with automation, interactivity, personalization, and agentic workflows.
Most of Google’s comparative figures come from internal benchmarks and human evaluations; independent confirmation is still limited.
Summary
Nano Banana 2 Lite and Gemini Omni Flash are Google’s latest steps to provide developers with inexpensive, fast multimodal media generation. They are designed to streamline workflows that move from image generation into video, and to support high-throughput automated use cases in apps and services.
