OpenAI's image tool gets way faster, smarter: OpenAI has launched a major upgrade to ChatGPT Images, delivering faster generation, sharper results, and far more precise control. The new image system is powered by GPT Image 1.5, which OpenAI says can generate images up to 4× faster while preserving fine details like lighting, facial structure, and composition. One of the biggest improvements is instruction following. Instead of reworking the entire image when you ask for a small change, the model now edits only the specific elements you request. Users can add, remove, blend, combine, or transpose objects while keeping the rest of the image intact. This makes multi-step creative workflows far more practical than before. OpenAI also highlighted stronger consistency across iterative edits, meaning artists, marketers, and designers can refine visuals without starting from scratch each time. The model handles complex prompts more reliably and produces clearer text and layouts inside images, a long-standing pain point for AI image generators. To tease the release, OpenAI shared an AI-generated yearbook-style photo of Sam Altman, showcasing the model’s improved realism and stylistic control. Users can access the new system directly by selecting “Images” inside ChatGPT, where preset styles and prompts are now available for faster experimentation. Overall, this update pushes ChatGPT Images closer to a true creative production tool rather than just a novelty generator.

Source: OpenAI

🤖 Robi: “Can it also generate my regrets from 2007? Asking for a friend.”

Google’s Gemini agent now reads your Gmail: Google Labs has unveiled CC, an experimental productivity agent powered by Gemini that connects directly to Gmail, Calendar, and Google Drive. The goal is simple: reduce daily mental load by turning your inbox into an organized action center. CC’s standout feature is a daily email called “Your Day Ahead.” Each morning, the agent scans your connected services and sends a concise briefing that highlights meetings, appointments, bills, deadlines, and other time-sensitive tasks. Instead of manually checking multiple apps, users get a single summary of what matters most that day. Beyond summaries, CC can also take action. Users can email CC directly and ask it to draft replies, schedule meetings, or send calendar links. This conversational control turns Gmail into a command interface rather than just a messaging platform. Because CC runs inside Google’s ecosystem, it can cross-reference emails, calendar events, and documents to provide context-aware assistance. For example, it might flag an upcoming meeting and surface the relevant Google Doc automatically. The agent is currently in testing through Google Labs, with early access available to select users. While still experimental, CC signals Google’s push toward agentic AI that works quietly in the background, organizing, prioritizing, and acting without constant user input.
Source: The Verge

🤖 Robi: “Finally, a coworker who replies to emails faster than... anyone in marketing.”

Chatterbox Turbo emotes better than your manager: Resemble AI has released Chatterbox Turbo, claiming it is the fastest open-source text-to-voice model available today, and with a key twist: emotional control. According to benchmark tests conducted by Podonos, Chatterbox Turbo outperforms several leading voice models in both speed and expressiveness. What sets Chatterbox Turbo apart is paralinguistic prompting. Instead of relying solely on tone sliders or presets, users can type emotional cues directly into the script, such as gasps, laughter, sighs, or pauses, and the model will vocalize them naturally. This allows creators to generate speech that feels far more human and dynamic. Because the model is open-source, developers can fine-tune it for specific use cases like podcasts, audiobooks, games, virtual assistants, or character-driven content. The speed improvement also makes it practical for real-time or near-real-time applications, where latency has traditionally been a blocker. Resemble AI positions Chatterbox Turbo as a flexible foundation rather than a locked-down product, encouraging experimentation and community-driven improvements. For creators who want expressive voice generation without relying on closed APIs, this release could become a major alternative in the rapidly evolving voice AI landscape.
Source: X Post

🤖 Robi: “Finally, a voice model that understands the art of the dramatic pause.”

AI detects dementia via EEG with 97% accuracy: Researchers at Örebro University in Sweden have developed AI models that can detect dementia by analyzing EEG brain signals. The system distinguishes healthy individuals from patients with Alzheimer’s disease and frontotemporal dementia with over 80% accuracy. A second, privacy-focused version uses federated learning, allowing models to train across institutions without sharing sensitive patient data. This approach achieved accuracy above 97%, while maintaining data privacy. The researchers say these models could make early dementia detection faster, cheaper, and more accessible, potentially enabling routine screenings in clinics or even at-home testing in the future..
Source: Fau Edu

🤖 Robi: “Great news, now your brainwaves can ghost you before your memory does.’’

Disney's OpenAI exclusivity deal has a time limit: Disney has clarified that its partnership with OpenAI includes only a one-year exclusivity period, despite the overall licensing agreement lasting three years. Disney CEO Bob Iger confirmed that after the first year, Disney is free to pursue similar deals with other AI companies. Under the agreement, OpenAI’s Sora video generator can legally use over 200 characters from Disney, Marvel, Pixar, and Star Wars. The limited exclusivity suggests Disney is keeping its options open as AI-generated entertainment rapidly evolves.

Source: Tech crunch

🤖Robi: “So Sora gets the lightsaber for a year. After that, who knows—Barbie meets Blade?”