- Published on
Gemini 2.5 in March: Deep Research with Your Files and Lyria 3 Music Generation
Have you ever wished you could hand your own research materials to an AI and say, "Now dig deeper, using these as your foundation"?
Existing AI research tools have been good at scraping the web. But using your own internal resources — the papers you've collected, meeting transcripts, interview notes — as the basis for additional AI-powered research was still difficult. In March 2026, Gemini changed that.
On top of that, Gemini introduced Lyria 3, a music generation model that turns text and images into audio. In a single month, Gemini significantly expanded its role as both a research tool and a creative partner. Here's a breakdown of what mattered.
Table of Contents
- Deep Research Evolved: Your Files Are Now the Source
- Lyria 3: An AI That Generates Music from Text and Images
- Gemini 2.5 Flash Improvements: Free Users Got Stronger Too
- Personal Intelligence and API Updates
- Practical Scenarios: A Guide for Researchers and Creators
1. Deep Research Evolved: Your Files Are Now the Source
The AI research tool that only looked outward can now look inward — at your materials.
In March 2026, Gemini's Deep Research gained the ability to accept file and image uploads as source material. Users can now include their own documents (PDFs, Word files, images) directly in the research process. AI combines your internal resources with external web searches to produce more contextually grounded reports.
Why does this matter? Trust. Traditional AI research often relied entirely on external search results — leading to context mismatches or redundant coverage of what you already knew. By anchoring research in your own materials, AI can more accurately fill the gaps in your thinking.
Deep Research Is Now Free
Even better news: Gemini 2.5 Flash-powered Deep Research is now open to free-tier users. Previously a paid-only feature, anyone can now experience AI-driven research grounded in their own sources. The quality differs from the premium tier, but the core workflow is now accessible to everyone.
2. Lyria 3: An AI That Generates Music from Text and Images
Describe a mood, get a song. No instrument skills required.
Launched in March, Lyria 3 is Google's music generation AI. It comes in two versions:
| Model | Optimized For |
|---|---|
| lyria-3-clip-preview | 30-second clip generation |
| lyria-3-pro-preview | Full-length song generation |
Both output 48kHz stereo audio and accept text and image inputs simultaneously. You can upload a reference image and ask Lyria 3 to generate music that matches its atmosphere.
How You Can Use It
- Educational content creators: Generate original background music for videos
- YouTubers and creators: Produce copyright-free original tracks for content
- Presentation makers: Score background music to match the mood of slides
- EdTech: Set the tone at the start of a class with generated ambient sound
"Music creation is being democratized. Whether or not you can read sheet music or play an instrument no longer matters — if you can describe a feeling in words, that's enough."
Currently available to developers via the Gemini API, broader consumer-facing access is expected to follow.
3. Gemini 2.5 Flash Improvements: Free Users Got Stronger Too
Better formatting and image understanding for everyone.
An improved version of Gemini 2.5 Flash has been rolled out to all Gemini app users. Two core improvements:
Enhanced Formatting: Complex outputs now make more active use of headers, lists, and tables. Instead of walls of text, Gemini produces structured, readable information.
Improved Image Understanding: Gemini more accurately interprets uploaded images and responds to them. Chart reading, graph analysis, and in-photo text recognition have all improved.
4. Personal Intelligence and API Updates
Gemini is beginning to understand your digital life as a whole.
Personal Intelligence Now Free (U.S.)
Personal Intelligence is now free for all Gemini users in the United States. By connecting Gmail, Google Photos, and YouTube, Gemini becomes a personalized assistant with real context about your life — helping plan vacations, manage projects, and organize personal records.
International rollout timing remains uncertain, but based on Google's typical expansion patterns, this feature is likely to reach global markets in the coming months.
Key API Updates
Developers also received meaningful API updates:
- Cloud Storage Support: Use Google Cloud Storage buckets directly as data input sources
- File Size Limit Increased: From 20MB to 100MB (a 5x increase)
- Built-in Tools + Function Calling: Use Gemini's native tools alongside custom function calls in a single API request
The 100MB file limit expansion makes a real difference for workflows involving high-resolution images, longer video clips, or large PDFs.
5. Practical Scenarios: A Guide for Researchers and Creators
Two usage patterns that maximize these updates.
Research Workflow:
- Prepare your primary source materials (papers, reports, interview notes) as files
- Upload to Gemini Deep Research and set a research goal
- AI combines internal sources with external search to generate a comprehensive report
- AI automatically surfaces gaps in the argument and potential counterarguments
Creator Workflow:
- Describe a video concept in text (e.g., "warm spring classroom, gentle piano melody")
- Optionally upload a reference image for mood
- Generate a 30-second background track with Lyria 3
- Insert into your video — no copyright concerns
Closing Thoughts
March's Gemini updates can be summarized in two themes: integration of your own materials and expansion into creative territory. Deep Research now bridges the external and the internal; Lyria 3 dissolves the boundary between text and music.
AI tools are accelerating their expansion beyond their original domains — converging and connecting in ways that weren't possible a year ago. In this landscape, what matters most isn't knowing which tool to use. It's the clarity to express what you actually want.
Related Posts
- NotebookLM Completely Transformed: Cinematic Videos and the 1M Token Era
- Claude Code Now Controls Your Mouse: The Era of Computer Use Has Arrived
Between Lyria 3 and Deep Research, which feature would you try first? Let us know in the comments!
Sources: