AI Models
The models behind Visxy
What each model is good at, where it differs from the others, and what it costs to run.
AI Models
What each model is good at, where it differs from the others, and what it costs to run.
The fast lane of Google's Nano Banana family. Built for trying Visxy's long 9:16 prompts — not for print-size files or long edit chains.
Credits come with a Generator plan. The exact cost is shown on the Generate button before anything is spent.
Lite is the model you open when you are still deciding whether a prompt is any good — not whether the file is finished. Google ships it as Gemini 3.1 Flash Lite Image. On Visxy it is the cheap, throwaway pass through the image prompt library: long portrait briefs, camera language, wardrobe, light, mood.
We have actually run it here. A typical Visxy 9:16 fashion-editorial prompt — subject, lens, set, fabrics, colour — came back as a usable still. That is the job. If you already know the frame is a keeper, you are on the wrong page.
Trying looks. The Visxy library is mostly 9:16 portraits with a paragraph of direction, not a five-word vibe. Lite will take that paragraph. Use it to see whether the pose, the crop, and the lighting read before you spend Nano Banana 2 or Nano Banana Pro on the same text.
It is also the right model for small assets you will never crop into: a story frame, a thumbnail, a sticker-like subject on a simple ground. Google built it for volume and speed, not for a file you will print or inspect at arm's length.
What it is not good at is the last mile. Tiny type you intend to read, a label that has to match a real product, a face you must lock across several references, a banner you will hang large — those fail in ways that look "almost right." Move those to Banana 2 or Pro instead of prompting harder here.
Lite is not Banana 2 with a smaller slider. It is a different Gemini image model (Flash Lite, not Flash). Nano Banana 2 and Nano Banana Pro let you pick a larger output; Lite has no size control at all. That is the whole product: one class of file, quickly.
It shares the ultra-wide frames with Banana 2 (the ones Pro does not offer). That does not make Lite the banner model. A panoramic from Lite is a sketch of a banner. If the wide frame is the deliverable, generate it on Banana 2.
Against Pro the gap is honesty, not branding. Pro is for frames that have to survive a close look. Lite is for frames you might delete. If you catch yourself writing the exact lettering a poster should say, stop and open Pro.
Paste a Visxy library prompt as written. These models read sentences, not keyword soup. The prompts that work here already name the subject, then the camera, then the light, then the clothes and the room. Cutting that down to "handsome man, studio, cinematic" is how you get a generic stock face.
Stay on 9:16 if the prompt came from the image library. auto is fine when the subject matters more than the crop; it is a poor default when you already know you want a portrait.
One idea per prompt. "She turns, then walks, then the city lights up" is three pictures. Lite will mash them.
Do not ask it to invent readable micro-copy. If the library prompt specifies a headline, treat the Lite result as a layout sketch and re-run the same prompt on Banana 2 or Pro.
A generation on Visxy is not the two-second playground clip some hosts advertise. The one Lite still we completed here took about a minute door to door, including queue and storage. Plan for that when you are iterating.
There is no public Lite Edit page. Editing is this model with a picture attached. Describe the change, not a new brief: "make the backdrop seamless paper, keep the pose" works; "an editorial studio portrait" reads as start over.
Keep the edit local — colour, sky, backdrop, a wardrobe swap. Lite is not the multi-reference, multi-turn editor. If you need a face or a product to survive several attached stills, that is Nano Banana 2. If you are correcting lettering on packaging, that is Nano Banana Pro. Detail that is not in the source will be invented.
No. Lite is Gemini 3.1 Flash Lite Image; Banana 2 is Gemini 3.1 Flash Image. Lite has no output-size picker. Banana 2 is a different model that can go larger, not a setting on this page.
Yes — that is the point of this page. The long 9:16 portraits in the library are the prompts we have actually run. If a result is only "close," move the same text to Banana 2 rather than rewriting the prompt for Lite.
Skip it when the file has to be large, when type inside the picture matters, or when you must keep a face or product consistent across reference stills. Those are Banana 2 and Pro jobs.
Yes, by attaching a reference to this model. There is no separate Lite Edit article. Keep the instruction to a single change. Long sequential edits belong on Banana 2; label-level corrections belong on Pro.