Stop 16 · AI APIs & Cloud AI PlatformsTopics · 3/9

Google Gemini API

In short

Gemini models take text, images, audio and video in the same request. If your feature starts with a photo of a bill or a lecture recording, this API is worth knowing.

Google's LLM API, notable for native multimodal support and tight integration with Google Cloud's AI ecosystem.

The blob holds up a phone showing a photo of a paper slip while a friendly chemist behind a counter nods.

Think of it like this

Sending a photo to the chemist

You WhatsApp a photo of the doctor's prescription to the local chemist and ask, 'Do you have these?' He reads the image, understands your question, and replies in text. Gemini is that chemist: you send a picture and a question together, and get an answer back.

Watch it happen

A photo and a question in one call

Image

Question

Gemini

JSON out

1/4A photo of the canteen bill is loaded and sent as one part of the contents list.

The big ideas

What you need to know

Multimodal contents

contents can mix strings, images and files. Order matters, so put the image before the question about it.

The google-genai SDK

The current Python SDK is google-genai, imported as from google import genai. Older tutorials use a different package.

Big files go via upload

Long videos or PDFs are uploaded with the Files API first, then referenced in the request.

Structured output

Set a response schema in the config when you need JSON that always matches a fixed shape.

See it in code

Walk through the code

python

1from google import genai2from PIL import Image3 4client = genai.Client()  # reads GEMINI_API_KEY5 6bill = Image.open("canteen_bill.jpg")7response = client.models.generate_content(8    model="gemini-2.5-flash",9    contents=[bill, "List each item and price from this bill as JSON."],10)11print(response.text)

1/3The client reads your key from the GEMINI_API_KEY environment variable.

Image plus question in one call with the google-genai SDK. Get a key from Google AI Studio and check the current model names there.

Save yourself the pain

Common mistakes

Oops

Following an old tutorial that uses the google-generativeai package.

Do this instead

Use the newer google-genai SDK. Google's docs use it now and older code may break.

Oops

Sending a 12 MB photo when a 1000 px wide one reads just as well.

Do this instead

Resize images before sending. Smaller images are faster and use fewer tokens.

Your mission

Read a real bill

~25 min

Quick check

Why put the image before the question in contents?

Go deeper

1Work through the Gemini API quickstart, including a multimodal (image + text) request.
2Compare Gemini's multimodal handling to a text-only workflow you've already built.
3Learn Gemini's pricing tiers and free quota for prototyping.