Nano Banana 2.1 — Keep Characters Consistent and Combine Up to 14 Images
Nano Banana 2.1 is Google's latest image generation and editing model. It keeps up to 4 characters consistent, combines up to 14 reference images into one scene, and outputs at 1K, 2K or 4K.
Google's Gemini Nano Banana 2.1, released Oct 6, 2026 · Live on Topview
Nano Banana 2.1
Available model
Nano Banana 2.1
Upload Reference Image (Max 6 images)
Image size
Aspect ratio
What Is Gemini Nano Banana 2.1?
Gemini Nano Banana 2.1 is the newest version of Google's Nano Banana image model, built on Gemini 3.6 Flash. Google describes it as an update focused on visual quality, prompt adherence, character consistency across edits, and text rendering. Besides Topview, Google uses it across its own products, including the Gemini app, AI Mode in Search and Flow.
Model ID:
gemini-nano-banana-2.1
Released:
October 6, 2026
Reference images:
up to 14 (10 object + 4 character)
Output:
1K, 2K or 4K
Other inputs:
text, video or a YouTube link
Consistent Character AI: Keep Up to 4 Characters the Same
Keeping the same face from one image to the next is one of the most common asks from creators. Nano Banana 2.1 takes up to 4 character references and holds their features steady as the scene, outfit or pose changes. In Google's own human-preference tests, multi-character consistency is one of its biggest gains over earlier versions.
Same Face, New Scene
Upload one clear portrait and place that person in a café, on a stage or in a product shot. Face shape, hairstyle and skin tone carry over, so a series of images reads as one character, not a set of lookalikes.
Group Shots with Up to 4 Characters
Add a reference for each person and describe who stands where and what they're doing. Naming each character in the prompt helps the model keep faces from blending together in busy scenes.
Series, Mascots and Storyboards
Brand mascots, comic panels and storyboard frames all need the same character again and again. Build the frames here, then turn them into video with Topview's .
Combine Images AI: Up to 14 References in One Scene
Nano Banana 2.1 doesn't paste pictures side by side. It reads up to 14 references, 10 objects and 4 characters, and draws one new image where they share the same light, perspective and style. Think of it as an AI image combiner that builds a scene, not a collage.
People and Products in One Shot
Put a model, a bag and a pair of shoes into one lifestyle photo without a shoot. Each item keeps its shape, color and label. For catalog work, see AI product photography.
Outfits, Props and Locations
Combine photos of a jacket, a hat and a street corner, then describe the moment you want. The model fits the clothing to the character and places everyone in the location.
Your Brand Assets in a New Setting
Drop in a logo, packaging and a color reference, and Nano Banana 2.1 places them in a fresh scene that still looks like your brand.
Change One Thing, Keep Everything Else the Same
To change one part of an image, describe it in plain words: what to change, and what should stay as it is. For example, "Change her hair to a short platinum bob; keep the face, the clothes, and the room." Google lists mask-based editing among the main improvements in 2.1, and on Topview you make those changes by describing them. If you'd rather brush over an area, try AI Inpaint.
Multi-Round Edits Without Drift: What Testers Found and How to Avoid It
Many rounds of edits can still wear an image down. Public tests shared after launch found that Nano Banana 2.1 redraws the whole frame on each edit, so parts you didn't ask to change can slowly shift. One independent test saw colors in untouched areas drift toward pink after eight rounds. These are findings from user testing, not Google's claims. Five habits help:
Go Back to the Original Every Few Rounds
Instead of editing an edit of an edit, return to your first image and apply the changes you've settled on from there.
Put the Changes in One Instruction
List everything you want changed this round in a single clear prompt, rather than nudging the image over many small rounds.
Say What Should Stay the Same
Name what should stay fixed, such as the face, background or lighting, every time you ask for a change.
Re-Attach Your Character References
Add the character images again on each round so the model checks the face against the source, not just the last output.
Use 2K or 4K for Small Text
Google's model card notes that small text can blur at 1K. If your image has labels, captions or fine print, generate at 2K or 4K.
Turn a Video or YouTube Link into Thumbnails and Posters
Nano Banana 2.1 can use a video or a YouTube link as a reference, not just still images. For creators who already have footage, that's a quick path from a clip to the artwork around it.
YouTube Thumbnail
Paste the video link and describe the thumbnail: the key moment, the expression, the title text. The design stays tied to what's actually in the video.
Movie-Style Poster
Use a trailer or a short clip as reference to build a poster with the same characters, mood and color palette.
Summary Infographic
Turn a tutorial or explainer video into a one-page infographic of its main steps. Check facts and figures before you publish.
How to Use Nano Banana 2.1 on Topview
Step 1: Add Your References or Start with a Prompt
Upload up to 14 reference images, a video or a YouTube link, or skip references and start from text.
Step 2: Describe the Scene or the Change
Write what you want to see or what to change, then choose 1K, 2K or 4K.
Step 3: Generate, Refine and Download
Generate, adjust in a few rounds if needed, and download the version you like.
What You Can Make with Nano Banana 2.1
Product Shots
Place one product in several settings for ads, listings and social posts. More in AI product photography.
Posters and Infographics
Google improved text rendering and infographic layout in 2.1, so headlines and labels come out cleaner. Start with the AI poster generator.
Character Series
Keep the same character across a comic, a campaign or a storyboard, then bring it to video with the .
Video Thumbnails
Use your own video as reference to make thumbnails and cover images that match the content.
Nano Banana 2.1 vs Nano Banana 2 vs Nano Banana Pro
| Nano Banana 2.1 | Nano Banana 2 | Nano Banana Pro | |
|---|---|---|---|
| Overall preference (Elo) | 1015 | 990 | 935 |
| Multi-character consistency (Elo) | 1068 | 978 | 1011 |
| Multi-reference editing (Elo) | 1041 | 988 | 989 |
| Mask/ink-based editing (Elo) | 1042 | 965 | 927 |
| Infographic factuality (score) | 0.328 | 0.179 | 0.265 |
| Reference images | 10 object + 4 character | 10 object + 4 character | 6 object + 5 character + 3 style |
Source: Google DeepMind Nano Banana 2.1 model card (Oct 6, 2026), Google's own human-preference evaluation; the lower of the two reported Nano Banana 2.1 configurations is shown. Reference limits: Gemini API image generation guide.
Which One Should You Use?
Choose Nano Banana 2.1
For scenes with several characters, many references, or edits where the rest of the image matters.
Choose Nano Banana Pro
When you need a fifth character reference or dedicated style references. See Nano Banana Pro.
Choose Nano Banana 2
If your prompts are already tuned for it, is still on Topview. Need the fastest drafts? Try .
More Image Models on Topview
Want a different look for the same brief? Topview also offers other image models, such as GPT Image 2 and Seedream 5.0 Pro.
Create with Nano Banana 2.1 on Topview
Bring your characters, products and references, and build your next image with Nano Banana 2.1.