Apple Academy Expands with AI Training for Students and Alumni
Apple is enhancing its educational offerings by incorporating AI training at the Apple Academy. The initiative, announced on Tuesday, will provide training in AI fundamentals to all students and mentors. Alumni of the Apple Developer Academy are also eligible to participate.
Commencing this fall, students will be equipped with skills to develop, train, and deploy machine learning models on Apple devices. The curriculum covers AI technologies, Core ML for optimized performance on Apple devices, and guidance on building AI models from scratch.
The latest Xcode 16 feature, Swift Assist, will support developers in their tasks and encourage experimentation. Swift 6 introduces enhancements for code clarity and simplifies concurrent programming. Susan Prescott, Apple’s vice president of Worldwide Developer Relations, expressed excitement about the potential creations students will develop and share with global communities.
Meta Unveils New AI Models for Audio, Text, and Watermarking
Meta’s FAIR team introduces innovative AI models and tools for researchers, focusing on audio generation, text-to-vision, and watermarking. The release includes JASCO, a model for text-to-music generation, allowing users to customize sound features through text inputs. FAIR will share JASCO’s inference code and pre-trained model under specific licenses.
Additionally, Meta launches AudioSeal, an AI-generated speech watermarking tool for localized detection within audio clips. This tool enhances detection speed significantly and will be available with a commercial license. The company encourages further research on image and text models.
Meta also unveils Chameleon, a multimodal text model available in two sizes for tasks involving visual and textual understanding. The release excludes the image generation model, focusing solely on text-related models. Researchers can access a multi-token prediction approach for training language models, emphasizing non-commercial and research-only usage.
ElevenLabs Introduces Open-Source Tool for Video Sound Effects
ElevenLabs unveils an open-source creator tool for adding sound effects to videos, following the launch of its Sound Effects text-to-sound AI offering. The tool allows creators to generate sound effect samples for their videos in just about 15 seconds. Developers can access the app’s code on GitHub, and a website is available for public use to try out the Sound Effects API.
The Video to Sound Effects app extracts frames from uploaded videos and uses OpenAI’s GPT-4o to create custom text-to-sound effects prompts. The generated sound effect is produced through ElevenLabs’s Sound Effects API, combining video and audio into a downloadable file up to 22 seconds long.
Ammaar Reshi, ElevenLabs’ design lead, sees the tool as a proof of concept for the SFX API’s potential in enhancing AI video creation workflows. The API enables developers to create custom AI sound effects based on descriptions, with pricing set at 100 characters per generation or 25 characters per second. The tool simplifies the process of adding sound effects to videos, offering possibilities for immersive experiences, like in video games.
Snap Unveils Real-Time Image Model for AR Experiences
Snap showcased its real-time, on-device image diffusion model at the Augmented World Expo, promising vibrant AR experiences. Co-founder and CTO Bobby Murphy highlighted the model’s efficiency in running on smartphones and re-rendering frames instantly based on text prompts.
The company aims to integrate this generative model into Snapchat Lenses soon for users and plans to release it to creators by year-end. Murphy emphasized the model’s role in reshaping AR experiences and the importance of accelerating machine learning models for impactful AR applications.
Lens Studio 5.0 launches with new generative AI tools, enabling developers to create AR effects swiftly, saving weeks of work. AR creators can design realistic ML face effects, custom stylization effects, 3D assets, characters like aliens or wizards, face masks, textures, and materials in minutes using Face Mesh technology. The update also features an AI assistant to assist AR creators with inquiries.
Butterflies: The Social Network Where Humans and AIs Coexist
Butterflies, a unique social network allowing interactions between humans and AIs through posts, comments, and DMs, officially launches to the public on iOS and Android after a five-month beta phase. Founded by former Snap engineering manager Vu Tran, Butterflies enables users to create AI personas, known as Butterflies, with distinct backstories, opinions, and emotions. Tran’s vision for Butterflies stemmed from a desire to enhance human-AI relationships beyond traditional text exchanges. Users can witness a diverse social media feed on Butterflies, featuring humans and AIs sharing updates and experiences. The platform’s innovative approach allows users to explore various scenarios, from creating multiple personas to forming connections with others based on shared interests and creations. Butterflies offers a refreshing take on AI integration in social networking, providing a space for users to interact with AI personas authentically and creatively. The app is currently free-to-use, with potential plans for a subscription model in the future.
