Create Video
Blogs

Nano Banana 2.1 Pricing: API Costs and Migration Guide

5 min read

Nano Banana 2.1 gives creative teams a new image-generation option and an immediate migration deadline. Google released gemini-nano-banana-2.1 to general availability on October 6, 2026, and announced October 29, 2026 as the shutdown date for the previous Nano Banana 2 API model, gemini-3.1-flash-image. Both dates come from the Gemini API release notes.

For PixVerse creators building reference images before generating video, the useful question is how much an approved asset costs. Developers also need to check whether an existing pipeline still works after changing models. This guide separates verified API prices from our planning recommendations; it does not report a migration benchmark.

Nano Banana 2.1 API prices

The following USD amounts cover image output only, as listed on Google’s Gemini Developer API pricing page, checked October 6, 2026.

Output resolution Standard per image Batch per image
1K $0.0336 $0.0168
2K $0.0504 $0.0252
4K $0.0756 $0.0378

These are not all-inclusive request prices. Standard input costs $1.50 per million text, image, or video tokens; text and thinking output costs $7.50 per million tokens. Batch rates for those categories are $0.75 and $3.75 respectively. Search grounding has separate billing. See the full pricing terms when estimating a workload with references, conversation history, or search.

Our arithmetic example: 100 Standard 2K outputs have an image-output subtotal of $5.04; the equivalent Batch subtotal is $2.52. Add the other applicable usage charges to estimate the complete API bill.

Budget for approved images

Suppose your campaign requires 20 approved opening frames. If you generate 100 candidates to select those 20, divide the total billed generation spend by 20, rather than by 100. This is a planning example, not an observed acceptance rate.

A useful internal metric is:

Cost per approved image = total generation spend ÷ approved images

Track the reason each candidate was rejected: incorrect product shape, illegible copy, inconsistent character details, or unsuitable composition. That record helps distinguish a prompt problem from a resolution problem. Increasing resolution is a poor first response to an incorrect subject or layout.

Our recommended starting workflow is to explore composition at 1K, review a small shortlist, and request higher-resolution finals only where delivery requirements justify them. Treat a regenerated final as a new candidate and inspect it again before using it as a video reference. Budget the subsequent video-generation stage separately.

When Batch makes sense

Google’s Batch API guide describes asynchronous processing with a target turnaround of 24 hours. That target should not become a promise to a customer who needs an immediate image.

We recommend Standard requests for interactive creative sessions, where someone is choosing the next edit after seeing each result. Consider Batch for independent thumbnail concepts, planned campaign variants, or evaluation sets that can wait.

For an automated campaign pipeline, give each request a stable identifier and record the corresponding output, review decision, and cost. Budget time to inspect results and rerun selected requests before delivery. A cheaper output is useful only when the workflow can accommodate its arrival time.

Migrate before October 29

Google’s deprecation notice explicitly directs users of gemini-3.1-flash-image to gemini-nano-banana-2.1. Our recommended migration process is:

  1. Find model references. Inspect application settings, scheduled jobs, evaluation tools, and deployment configuration for the old ID.
  2. Prepare representative cases. Include your real product references, required copy, usual aspect ratios, and multi-turn edits. Define what makes each output acceptable before running the comparison.
  3. Check request compatibility. Review resolutions, tools, and other options against the new model documentation. Explicitly inspect any workflow that previously requested 512-pixel output rather than assuming it carries over.
  4. Measure the complete workflow. Record billed usage, errors, completion time, and approved outputs. Compare cost per accepted asset alongside visual review.
  5. Move production traffic and monitor. Complete the switch before the shutdown date, including background jobs. Plan recovery around an available model and configuration that you have validated.

Changing an ID is only one part of migration. Keep the prompts and reference assets used in evaluation so that future changes can be assessed against the same creative requirements.

What this means for PixVerse users

PixVerse has an official Nano Banana 2 availability announcement. That announcement concerns Nano Banana 2; it does not confirm Nano Banana 2.1 integration. Google’s API dollar prices also do not establish PixVerse credit costs.

If you create through PixVerse, check the model selector and current in-product pricing before generating. If you call Google’s API directly, manage that API migration separately. For the next creative stage, our image-to-video guide explains how to plan motion from a still image.