Welcome to Get The Gist, where every weekday we share an easy-to-read summary of the latest and greatest developments in AI—news, innovations, and trends—all delivered in under 5 minutes! ⏱

In today’s edition:

• xAI Launches New API Service for Grok Models

• Haiper 2.0: A Free AI Video Generator That Could Outshine OpenAI

• Meta to Fight Celebrity Deepfake Scams with Facial Recognition Tech

• ChatGPT Advanced Voice Feature Now Available in EU

• And more AI news….

Top Developments

xAI

xAI Launches New API Service for Grok Models

The Gist: Elon Musk’s AI company, xAI, has rolled out an API for its Grok models, allowing developers to integrate these AI tools into their apps. While it offers powerful features, the pricing is higher than competitors like OpenAI.

Key Details:

• The xAI API, now in public beta, supports integrations with OpenAI and Anthropic SDKs and offers models like Grok-2 and Grok-2 mini.

• Despite its competitive features, the Grok models are more expensive than OpenAI, charging $5 for input tokens where OpenAI only charges $2.50.

• The xAI Console offers tools for model comparison, billing, and usage tracking, though it's currently limited to one region (us-east).

• Developers can integrate the API using REST, gRPC, or xAI’s Python SDK, but Grok’s token capacity is notably lower than GPT-4o.

Haiper

Haiper 2.0: A Free AI Video Generator That Could Outshine OpenAI

The Gist: Haiper just launched its upgraded AI video generator, Haiper 2.0, which offers stunning, hyper-realistic videos for free, posing a direct challenge to OpenAI’s video tools.

Key Details:

• Haiper 2.0 creates lifelike videos faster than ever, using advanced transformer models and diffusion techniques.

• It’s free to use, unlike OpenAI’s Sora, and allows users to turn still images into high-quality videos with new Video Templates.

• The platform now supports 1080p video generation with future upgrades aiming for 4K resolution.

• Haiper 2.0 includes advanced tools like an HD upscaler and keyframe conditioning for precise video control, with plans for even more features on the way.

Meta

Meta to Fight Celebrity Deepfake Scams with Facial Recognition Tech

The Gist: Meta will trial facial recognition technology to combat deepfake ads featuring celebrities on Facebook and Instagram, aiming to quickly identify and remove misleading content.

Key Details:

• Starting in December, Meta will use facial recognition to compare suspected deepfake ads against celebrity profiles for quick detection and deletion.

• The trial involves 50,000 celebrities and public figures worldwide, with an opt-out option available.

• Meta emphasizes privacy by ensuring facial data is deleted immediately after each match and won’t be used for other purposes.

• While promising results have been seen, Meta expects scammers to adapt, requiring ongoing updates to detection tools.

Quick Gist

Gemini is set to introduce new features allowing users to generate images in various aspect ratios and manage calls and texts while the phone is locked, along with a redesigned user interface (Read More).

At ROSCon, NVIDIA unveiled generative AI tools and simulation workflows for ROS developers, enhancing robotics capabilities through improved perception and interaction using AI (Read More).

Midjourney announced an upcoming image editor that will allow users to upload and edit any internet image using generative AI, while planning initial access restrictions to prevent misuse (Read More).

OpenAI expanded its ChatGPT Advanced Voice feature to Plus and Team users in the EU, Switzerland, Iceland, Norway, and Liechtenstein, enhancing accessibility with new voices and improved performance (Read More).

OpenAI introduced TrigFlow, a novel trigonometric approach that enhances the stability and scalability of continuous-time consistency models, significantly improving training efficiency and sample quality while reducing computational costs (Read More).

NVIDIA and Deloitte are collaborating to develop AI-powered virtual teammates for hospitals to enhance patient communication and reduce pre-surgery anxiety, with a pilot launching at Ottawa Hospital by year-end (Read More).

ETH Zurich researchers developed COMPL-AI, a benchmarking suite that translates the EU AI Act's legal requirements into measurable technical standards for General Purpose AI models (Read More).

Honeywell and Google Cloud launched an AI agent for the industrial sector, set to debut in 2025, to streamline operations and address labor shortages by connecting assets, people, and processes (Read More).

News Corp is suing Perplexity AI, alleging the startup illegally copies its content, diverts revenue, and harms journalists by creating AI-generated summaries that misrepresent original news articles (Read More).

Creators have added copyright infringement claims to their federal lawsuit against OpenAI, accusing the company of using millions of YouTube videos for AI training without consent (Read More).

Anthropic is proactively assessing potential misuse of its Claude AI models by evaluating their susceptibility to sabotage and manipulation, highlighting the need for stronger safety regulations as AI capabilities evolve (Read More).

Google Chat is rolling out Gemini conversation summaries for unread groups, threads, and spaces on Android, iOS, and web, enhancing user experience with AI-generated synopses (Read More).

Meta launched several new AI models, including the Self-Taught Evaluator and SAM 2.1, to enhance cross-domain AI capabilities, but faces criticism over limitations and its use of the term "open-source" (Read More).

THAT’S IT FOR TODAY!