v1.3.4API credit, right in the headerChangelog →
MergeRef
Notify Me
AI models

Google — Gemini & Veo

Google provides four Gemini image models (the Nano Banana family) and two video models, Gemini Omni Flash and Veo 3.1, in MergeRef. Veo 3.1 works only when you connect Gemini Enterprise Agent Platform (formerly Vertex AI); with a Gemini API key, use Gemini Omni Flash. Gemini is the strongest choice for free-form edits that keep the original image intact, and it is especially good at photorealism. Nano Banana Pro is the default for identity-safe editing; Nano Banana 2.1 is the fast, low-cost option.

Models in MergeRef

Image models
2.1Nano Banana 2.1Latest, recommended by Google
LiteNano Banana 2 LiteCheapest, 1K only
2Nano Banana 2Fast generation, low latency
ProNano Banana ProHigh quality, best for production
Video models
Gemini Omni FlashFast video generation + conversational editing (720p)
Veo 3.1Start/end frame interpolation · Agent Platform only

API key — set it up first

MergeRef runs Gemini with your own Google account. There are two ways in: an AI Studio API key, the quickest for one person, or Agent Platform (Vertex AI), where requests run inside your own Google Cloud project with no key at all (next section).

Get an AI Studio key

1
Sign in
Open Google AI Studio and sign in with your Google account. New users get a default Google Cloud project and key when they accept the terms.
2
Create a key
Go to Dashboard → API keys and press Create API key.
3
Billing
You can start on the Free Tier. For the Paid Tier, press Set up billing on the project row (see Billing). Watch spending on Spend.
4
Copy the key
Treat it like a password and never share it. See Using Gemini API keys.
5
Add it in MergeRef
Settings › Account & API › Providers › AI Studio: press Add Key, give it a name, paste the key (AIza…) and press Validate Key. Keep Enable Google AI Studio on.

In MergeRef

  • Several keys: add as many keys as you like and pick one in the list; the selected key is the one MergeRef uses and validates.
  • One Gemini route at a time: AI Studio and Agent Platform (Vertex AI) drive the same Gemini models, and only one can be on. Turning one on turns the other off.
  • Shortcuts: the AI Studio card links straight to Get API key from Google AI Studio, View Google AI spending and Set monthly usage limit.

How your key is kept and billed

  • Where it is kept: on this computer only. MergeRef encrypts saved keys with the operating system's own data protection, so the saved file is useless on another PC or user account.
  • Where it is sent: only to Google, with each request. There is no MergeRef server in between, and MergeRef never sees your usage.
  • Who bills you: Google, under your own account and limits. MergeRef takes no cut. The Model pricing card at the bottom of the same settings page lists each model's price.

Connect Agent Platform (Vertex AI)

For studios whose requests must run inside their own Google Cloud project, under its quotas and billing account. You sign in with Google in MergeRef; no API key or service-account file is needed.

Official name for company/IT requests: Google Cloud — Gemini Enterprise Agent Platform (formerly Vertex AI), API aiplatform.googleapis.com

Google renamed Vertex AI to Gemini Enterprise Agent Platform. It is the same Google Cloud service and works with an ordinary billing project. MergeRef’s settings still show it as Vertex AI.

1
Create a Google Cloud project
Create an account and a project in the Google Cloud console.
2
Link billing
Link a billing account to the project in Billing. Agent Platform is billed by usage.
3
Enable the Agent Platform API
4
Connect in MergeRef
Settings › Account & API › Providers › Vertex AI: turn on Enable Vertex AI, press Connect with Google, choose your account and allow access in the browser, then pick your project in the list (search appears with more than five projects) or type its project ID.
  • Change Project, Switch Account and Disconnect are on the same card. Disconnecting returns Gemini to the API key route.
  • The credits line reads "billed by usage": Google Cloud bills the project directly.

Nano Banana Pro

Google's high-end image model. It renders legible text, supports multi-turn editing and keeps a character consistent across up to 14 reference images. It reasons about the instruction before drawing, which makes it reliable with long, multi-part prompts.

Pricing

OutputPrice per image
1K and 2K$0.134
4K$0.24
Each input image560 input tokens (about $0.001)

Billing is token based: image output is $120 per million tokens, so a 1K/2K image uses 1,120 tokens and a 4K image 2,000.

Use it in MergeRef

  • Outfit swaps: ORIG + costume role. Ask it to keep the face and hairstyle.
  • Pose changes: ORIG + pose role, or an extracted OpenPose skeleton as pose.
  • Auto Painting and Part Compositing, where staying faithful to the original matters most.
  • Draft at 1K, re-run the winner at 4K only if you need it for print.
  • Inpainting & Object Insert: Gemini-only generation tools in the prompt panel. Paint a mask to regenerate an area or composite a new object into it. See Prompt panel.

Nano Banana 2.1

An update to Nano Banana 2 (Gemini 3.1 Flash Image) with better visual quality, multi-turn character consistency and text rendering, at 1K, 2K and 4K. It is the lowest-cost Gemini image model at each resolution.

OutputPrice per image
1K$0.0336
2K$0.0504
4K$0.113

Gemini 3.1 Flash Image (Nano Banana 2)

The Flash image model built for speed and throughput, with outputs from 0.5K drafts up to 4K.

OutputPrice per image
0.5K$0.045
1K$0.067
2K$0.101
4K$0.151

Gemini 3.1 Flash Lite Image (Nano Banana 2 Lite)

The lowest-latency, lowest-cost option at $0.0336 per 1K image. Use it for single edits and quick checks; for edits with many references or several turns, Nano Banana 2.1 or Pro is a better fit.

Gemini 2.5 Flash Image (Nano Banana) was shut down by Google on October 2, 2026 and is no longer in the MergeRef model picker.

Gemini Omni Flash

Google's fast video model for generation and conversational editing. It turns text and images into 720p clips, and you can keep refining the clip with follow-up instructions instead of starting over. Billed at $0.10 per second.

Veo 3.1

Google's video model. It produces 4, 6 or 8 second clips with native audio, supports first-and-last-frame control and up to three reference images, and outputs up to 4K.

Agent Platform (Vertex AI) only. Veo 3.1 is available only when Google is connected through Agent Platform (Vertex AI). Google ended Veo 3.1 for Gemini API keys from AI Studio on 22 October 2026, so with an API key MergeRef dims Veo and shows why; use Gemini Omni Flash instead. See Veo 3.1 in the Agent Platform documentation.
TierPrice per second
Standard (720p / 1080p)$0.40
Standard (4K)$0.60
Fastfrom $0.10
Litefrom $0.05

Agent Platform (Vertex AI) prices for video with audio, checked 6 November 2026. Video without audio costs less. For exact prices, see Google’s Agent Platform pricing.

You are only billed for successful generations. The MergeRef model picker offers Veo 3.1 only; Fast and Lite prices are for reference.