Nano Banana 2 Lite when the hero does not need Pro

Pro is for the still that has to survive a landing page, a paid unit, or a client who will zoom. Lite is for everything you will throw away or bury in a stack: reference sheets, thumbnail batches, slideshow frames, "is this composition even a thing." Paying Pro for those is not quality. It is a habit.

Nano Banana 2 Lite is the cheap, fast, text-to-image-only tier. The cheap-fast spec is the review: four-second-class turns, 4 credits, no image-to-image, no references, fixed 1K. What Nano Banana 2 actually does is the mid-tier. This post is the routing rule: when the still is not a hero, do not buy a hero model.

Three tiers, three jobs

The Nano Banana family on Versely is not a marketing gradient. Each rung is missing a verb the one above it has.

Tier	Credits (Versely)	Verbs	Output
Lite	4	Text-to-image only	Fast, 1K, 14 aspect ratios
Nano Banana 2	8 (4 when discounted)	Generate, edit, up to 14 references	Up to 4K
Pro	15 SD/HD, 30 at 4K	Generate and edit, hero headroom	Up to 4K

Lite cannot touch an existing image. That is the wall. If the brief starts with a photo, a pack shot, or a character still you need to hold, Lite is already the wrong counter. You are not "saving" by running it; you are generating something you will redo on 2.

Pro is not "2, but nicer." It is the still you cannot afford to look like a draft: key art, a print-scale hero, the one frame a campaign hangs on. It is also the wrong counter for a twelve-up contact sheet of wardrobe options.

The decision rule

Ask two questions, in order.

1. Does this job start from an image I already have?

If yes, Lite is out. Route to Nano Banana 2 (or Pro if the edit itself is the hero). Lite has no edit path and no reference slots. Character lock, product lock, "change the jacket, keep her face" — all of that is 2.

2. Does a single output have to survive scrutiny?

If no — it is one of many, it will be small, it will be a board, it will be thrown away — Lite.

If yes — one frame, full size, a stranger will judge it as the brand — Pro, or 2 if you need references more than you need the last quality step.

That is the whole router. Everything else is examples of those two answers.

What "not a hero" looks like in production

Reference sheets. Face, wardrobe, palette, three-quarter, product on white. These exist so the video model has something to lock to. They are not the ad. Lite (or 2, if you are conditioning on a real product photo) is the right spend. Pro on a sheet is paying hero rates to discover the character drifted — which is the expensive way to learn what a still should have taught you for less.

Thumbnail and variant stacks. A thumbnail is judged in a fraction of a second against a row of competitors. The winning move is many, then pick, not one perfect. Lite is built for that loop. Promote a winner to 2 only if it needs an edit or a 4K pass.

Slideshow frames. A carousel is a set. No single frame is the campaign. Batch on Lite off one prompt template. If a frame has to carry dense type, that is not a Lite job and not a Pro job either — that is a layout model.

Concepting. "Is this even the composition." You are not choosing a shippable still. You are choosing whether the idea exists. Lite, then stop. The keeper, if there is one, gets a second pass on the tier the final job actually needs.

Do not use Lite as a preview of Pro. Lite cannot edit and cannot take references; a composition that works on Lite can fail the moment you need the pack shot to stay the pack shot. Preview on Lite only when the final job is also a from-scratch still.

What you still pay Pro for

A hero still. Say it plainly so the budget meeting cannot pretend otherwise.

• The key visual on a landing page or a paid unit at scale.
• A print-adjacent frame, or anything you will actually 4K.
• An edit of a hero — not a sheet — where the last pass has to hold.

If you cannot name the URL or the insertion the still will live on, it is not a hero. It is a board. Boards go to Lite (from scratch) or 2 (from a reference). Pro is the exception you can point at.

The Lite spec and Nano Banana 2 capabilities pieces are the reviews. This decision only needs the brief: image in, or one still out that has to be right. Then pick the rung.

FAQ

Is Lite "good enough" for client work?

For client volume — options, frames in a set, internal boards — yes. For a single client-facing hero, no. The client will zoom. Lite is 1K, from-scratch only, and ranked below 2 on text-to-image for a reason. Send the hero to 2 or Pro. Do not send the twelve options there.

Can I generate on Lite and then edit on Pro?

You can take a Lite still and feed it into a higher tier as an image input, because 2 and Pro can edit. That is a valid promote-the-winner path when the Lite composition is the thing you liked. It is not a reason to start on Lite when you already know the job is an edit of a real photo. Start on the tier that has the verb.

Why not run everything on Nano Banana 2 and ignore Lite and Pro?

Because 2 is the middle on purpose. It is the right default when you need a reference or an edit and the still is not key art. It is twice Lite's list price for jobs Lite can finish, and it is not the 4K hero tier. Using 2 for everything is how the credit bill looks "reasonable" and still wastes the rungs.

Does Arena rank change this router?

No. Lite is 7th on text-to-image in the catalog snapshot; 2 is 3rd; Pro is 6th. Those numbers move. The verbs do not: Lite still cannot edit or take references. Route by the job, then look at the board if two models on the same rung are in play.