7 examples of Geminis multimodal capabilities in action – Google Developers Blog

Gemini image analysis with text prompts

Define the action, context, source, audience, output format, constraints, and success criteria. Availability and limits may depend on the account, plan, device, and file type. Researching, planning, writing, designing, coding, and evaluating in one prompt may produce shallow results. Gemini cannot analyze a document, image, spreadsheet, or video that has not been attached or made available through a connected source. Define the question, region, timeframe, customer, criteria, and source priorities. It is which workflow best matches the task and available source Computer vision material.

  • You can provide one or more images alongside your text prompts for editing, composition, and style transfer.
  • Create a image type for brand/concept with the text „text to render” in a font style.
  • Using the provided image of subject, please add/remove/modify element to/from the scene.
  • A caption box at the top reads „The city was a tough place to keep secrets.” The lighting is harsh, creating a dramatic, somber mood.
  • Provide an image and simply describe the change you want.

This page shows you how to add images to your requests to Gemini in Gemini Robotic process automation Enterprise Agent Platform by using the Google Cloud console and the Agent Platform API. For OCR tasks, use high-resolution images when possible. Start with the Subject-Context-Style-Mood-Format formula, use these prompts as your baseline, and build from there through iteration.

  • Using the provided image of a living room, change only the blue sofa to be a vintage, brown leather chesterfield sofa.
  • Generate a realistic, full-body shot of the woman wearing the dress, with the lighting and shadows adjusted to match an outdoor environment.
  • Using the provided image, change only the specific element to new element/description.
  • Create a new image by combining the elements from the provided images.
  • Keep the rest of the room, including the pillows on the sofa and the lighting, unchanged.

For image generation, always open with „Generate an image of” or „Create an image of” — Gemini Omni defaults to text output unless you make the format explicit. This guide gives you the exact formula and ready-to-use prompts across every output type. Refine unclear instructions, verify important information, and save successful prompts as reusable templates.

AI Task Guides

Gemini image analysis with text prompts

Pull every figure, date and named entity from the attached text into a table with columns for value, unit, date, and the source sentence. Read the attached document and https://aicurated.com/ answer only from it. Gemini follows long instructions closely, which is an advantage only if the instructions are worth following. Before generating images, the service verifies storage.objects.create permission on the target Cloud Storage bucket using the caller’s credentials. You can set media_resolution in generationConfig to apply low, medium, or high resolution to all media in the request, or set resolution for individual media parts, which will override the top-level setting. You can add a single image or multiple images in your request to Gemini.

Gemini image analysis with text prompts

World Context & Reasoning

Deep Research is useful when a task requires broader source collection, comparison, evidence evaluation, and a structured report. Review the attached project brief and create a https://aicurated.com/computer-vision-and-perception/google-gemini-puts-text-prompts-to-work-in-computer-vision/ prioritized execution plan with milestones, dependencies, risks, and immediate next actions. A stronger prompt tells Gemini what to do, what information to use, who the output is for, how the response should be structured, and what a useful result should achieve.