PRACTICAL GUIDE · PROMPTTOVISUAL
Nano Banana 2.1: What Changed, API Model ID and Image Pricing
Nano Banana 2.1 is Google’s image-generation and conversational-editing update, released October 6, 2026. This external model guide explains confirmed changes, output-only pricing and migration limits. It is not integrated into PromptToVisual.
EXTERNAL MODEL GUIDE · CHECKED OCTOBER 7, 2026
Google model documentation · PTV Nano Banana 2 · PTV 2 Lite
Change poster text with our current editor · Combine two images · Product photography
What changed, and is it available?
Google lists gemini-nano-banana-2.1 as generally available from October 6, 2026. Its release notes identify improvements in image quality, instruction following, text rendering and character continuity across turns. These are official claims, not our benchmark results. Source check: October 7, 2026.
2.1, Nano Banana 2, Lite and Pro are not interchangeable
Nano Banana 2.1 updates Nano Banana 2 (gemini-3.1-flash-image). Google positions it as an efficient alternative to Nano Banana Pro (gemini-3-pro-image), not a renaming of Pro. Nano Banana 2 Lite (gemini-3.1-flash-lite-image) is a separate efficiency-focused model. Choose the exact model ID, not only the family nickname.
Confirmed capabilities and limits
The model card lists 1K, 2K and 4K images, with 1K as the default. It describes improved text and infographic layouts, and fixes to tiling on 1:4, 4:1, 1:8 and 8:1 outputs at 2K/4K. It accepts up to 14 reference images, describing continuity for four characters and fidelity for ten objects. These are supported-workflow limits, not guarantees of identity or exact lettering.
API and access
The stable API model ID is gemini-nano-banana-2.1. Google links access through AI Studio and its API documentation. The model card supports image/text output and Batch, but not Flex or Priority. Account access, quotas and safety checks still apply; availability in Google’s API does not mean availability in another provider or in PromptToVisual.
1K, 2K and 4K: output cost, not the full request
Checked October 7, 2026, Google’s Standard pricing lists USD 0.0336 for a 1K output image, USD 0.0504 for 2K and approximately USD 0.113 for 4K. Batch lists USD 0.0168, 0.0252 and 0.0567 respectively. Standard input text/image/video costs USD 1.50 per million tokens; output text/thinking costs USD 7.50 per million. Image output is USD 30 per million tokens. Inputs and generated reasoning/text add to image-output cost.
Optional search and the 4K accounting discrepancy
The pricing page lists 5,000 monthly search requests shared across Gemini 3.x, then USD 14 per 1,000. One prompt may trigger multiple billable queries. Search-retrieved context is excluded from input-token charges. The price-page footnote uses 3,780 output tokens for 4K, while the image guide’s table still shows 2,520. Use the current pricing page for budgeting, not a recalculation from the conflicting table. Recheck actual usage and tool charges; the displayed image amount is not a fixed full-request price.
Reference editing, local changes and multi-turn work
Google documents editing from images and conversational follow-ups, including previous_interaction_id in its Interactions API examples. It describes inpainting as semantic masking: name the region and desired change in language. We did not find an explicit 2.1 release claim establishing a new pixel-mask upload interface or a quantified improvement in local-mask accuracy. Do not treat a request to leave other pixels untouched as a preservation guarantee. Keep originals and compare every revision.
Text and character consistency still need review
Use clear references you have permission to provide. State the exact wording first, then request the image or edit. Check spelling, logos and identity across every turn rather than assuming continuity from a model-family claim. Google documents SynthID on generated images; generated output is not an editable design-layer file. For a poster that must remain exact, prefer its original editable source.
Migration: deprecation is not a shutdown date
Google’s October 6 changelog deprecates gemini-3.1-flash-image in favor of 2.1 but announces no shutdown date. That does not establish a same-day closure of Runware’s Nano Banana 2 endpoint. PromptToVisual’s current model names, providers and credit prices have not changed. For a direct Google integration, compare your existing requests, supported sizes, reference handling, safety behavior and total bills before switching production traffic.
Original prompt idea — UNTESTED EXAMPLE
Using my supplied fictional workshop poster, replace only “SATURDAY 12 SEPTEMBER 2026” with “SATURDAY 19 SEPTEMBER 2026”. Keep the remaining wording and layout as close as possible. Return the complete poster. This example has not been run on Nano Banana 2.1. Inspect every word, illustration and margin before publishing; the instruction does not guarantee unchanged pixels.
What you can do on PromptToVisual today
Nano Banana 2.1 is not available in our model selector. Our current Nano Banana 2 and 2 Lite pages describe different deployed models. For a concrete text-change task, open AI Image Editor and load its existing demo poster; it uses the current Qwen workflow. For two-image scenes or product photos, use our existing combiner and product-photography tools. These are alternatives, not a disguised 2.1 integration.
Sources and next steps
Sources checked October 7, 2026. External model guide; pricing-page accounting takes precedence where documentation differs. No integration or model test.
Google image generation and editing →
Flare generator and real samples → · Sunburst generator and real samples → · Images 2.5 and Sketch · LLaDA ComfyUI setup
Continue with the right tool
AI image combiner · Stitch images · Photo collage · Nano Banana 2 guide · Related composition guide