Google acaba de DESTRUIR la IA: 8 Funciones GRATIS que lo demuestran
The video showcases eight new free features from Google's Gemini AI platform, including interactive 3D visualizations, real-time screen-sharing video calls, a highly realistic text-to-speech model, on-device AI, and Chrome extensions. The presenter argues these updates collectively represent a major leap that competitors cannot easily match. Most features are demonstrated live through practical use cases.
Summary
The video opens by framing Google as an accelerating force in AI, arguing that despite competitors like Anthropic appearing to lead at times, Google's infrastructure gives it a structural advantage that allows it to launch capabilities others cannot replicate at scale — and for free.
The first feature demonstrated is Gemini's new interactive visualization capability. When a user asks Gemini to explain how a car engine works, it generates a fully animated, step-by-step interactive diagram showing intake, compression, power, and exhaust cycles with adjustable RPM speeds. The presenter also shows this working for gravity (producing a 3D space-time well simulation with adjustable star mass) and for visualizing how Gemini's own AI pipeline processes an input into an output. Notebook LM sources can also be connected to generate custom interactive visualizations from personal knowledge bases.
The second feature is Gemini 3.1 Flash's live video call and screen-sharing capability, accessed via Google AI Studio. The presenter conducts a live demo where Gemini watches a spreadsheet in real time and provides step-by-step guidance on sorting data by margin, adapting its suggestions when the first approach wasn't satisfactory. The response latency is highlighted as near-instantaneous, comparable to talking to a human.
The third feature is the Gemini 3.1 Flash TTS (text-to-speech) model, described as among the best in class. The presenter tests it with multiple Spanish regional accents (Sevillana, Canarian, Valencian, Mexican), complex emotional sequences including hesitation, coughing, laughter, whispering, shouting, and pauses — all produced convincingly on the first attempt.
A sponsored segment follows, promoting Hostinger's Horizon platform, which generates fully functional websites and apps from a text prompt, demonstrated by creating a professional voiceover company website.
The fourth feature shows how the Gemini TTS model can be connected to custom-built tools via Google AI Studio's 'Build' section, allowing users to create their own voiceover platforms similar to ElevenLabs. A secondary feature highlighted here is the new prepaid API billing option, which lets developers set a fixed credit balance without auto-renewal, eliminating the risk of runaway API costs.
The fifth feature is Gemma 4, a fully open-source on-device model paired with Google's AI Edge Gallery app (available on iOS and Android). The presenter demonstrates running the model entirely offline in airplane mode, with fast response times and no censorship limitations.
The sixth feature is the integration of Notebook LM directly into Gemini's chat interface, enabling users to organize conversations into notebooks, import files from existing notebooks, and create new notebooks from within Gemini — though this feature is noted as only rolling out to some accounts.
A promotional segment for the presenter's small-group intensive AI course is included, featuring student testimonials and a closing enrollment pitch.
The seventh feature covers two Google Chrome extensions for Gemini: 'Chat Architect,' which adds folder and subfolder organization to Gemini chats (replicating Notebook LM-style organization), and a second extension that expands Gemini's text display to fill the full screen width by removing unnecessary margins.
The eighth and final feature is the official Gemini desktop application for Mac, which mirrors the web interface but adds a persistent screen-capture shortcut for fast, always-available access. The presenter notes that Google's own CEO revealed this app was built in just a few days by a small team using Google's own AI coding tools. A Chrome extension alternative is offered for Windows users in the interim.
Key Insights
- The presenter argues that Gemini's new visualization feature goes beyond simplifying explanations — it literally generates interactive 3D simulators tailored to the concept being learned, such as a manipulable space-time gravity well with adjustable star mass, not just static diagrams.
- The presenter claims that Gemini 3.1 Flash's live screen-sharing mode responds with near-zero latency, describing the experience as indistinguishable from talking to a human in real time — and demonstrating that it adapts its recommendations mid-conversation when the first suggestion is rejected.
- The presenter asserts that the Gemini 3.1 Flash TTS model correctly executed a complex sequence of emotions — whispering, pausing, soft laughing, escalating intensity, loud laughing, and shouting — that even a human voice actor would find challenging to produce consistently.
- The presenter highlights that Google introduced a prepaid API billing option with no auto-renewal, which he frames as a major quality-of-life improvement for developers who previously faced the risk of uncontrolled API spending when automating workflows.
- The presenter states that Google's CEO publicly revealed the Gemini desktop application was built in just a few days by a small team using Google's own AI coding tools (referred to as 'Google Antigravity'), citing this as evidence that AI is enabling companies to accelerate their own development cycles dramatically.
Topics
Full transcript available for MurmurCast members
Sign Up to Access