News
Magenta RealTime makes music creation with AI more accessible
Google released Magenta RT — an open AI model for real-time music creation with easy access for everyone
MiniMax-M1 processes a million tokens and approaches the level of Gemini 2.5 Pro
MiniMax-M1 is available for free and suitable for large-scale text analytics and multimedia projects
Mistral AI introduced an improved open model Small 3.2
Mistral Small 3.2 received targeted improvements, became available on Hugging Face, and supports popular frameworks
Voice Search in Google Allows Dialogue with AI
Google is testing Search Live — a new voice communication feature with AI in its search for mobile users.
ChatGPT Changes the Speech Style of YouTube Users and Chats
Studies show that ChatGPT and similar AIs influence the vocabulary, structure, and emotional tone of digital communication.
Firefly by Adobe is now available for iOS and Android
Firefly is now available on iOS and Android, bringing all image and video generation capabilities to mobile devices.
Gemini on Android can now identify music at the user’s request
From now on, Gemini on Android allows identifying nearby music through a voice command using Google's Song Search function
ChatGPT Boosts Creativity but Reduces Diversity of Ideas in Groups
New studies have shown that ChatGPT helps generate interesting ideas but limits diversity in group creativity
OpenAI allows recording voice sessions in ChatGPT on Mac
The desktop ChatGPT app for macOS now features voice session recording and transcription for quick summaries.
Veo 3 AI Model Coming to YouTube Shorts This Summer
YouTube integrates the new Veo 3 model into Shorts for video generation, expanding opportunities for creators and audiences.
Resemble AI Releases Free Chatterbox Model for Voice Cloning
Chatterbox allows for quick voice cloning with emotion control and operates locally on Windows, Mac, and Linux
The release of GPT-5 from OpenAI is expected this summer
OpenAI is preparing the new GPT-5 model, but the exact release date and launch format remain unknown.
