ByteDance Introduces OmniHuman — A New AI Video Model With Incredibly Realistic Video AND Audio Outputs

⚡️ Headlines

🤖 AI

Google Ends Its Promise Not to Use AI for Weapons, Surveillance – Google has revised its AI principles, removing prior commitments to not using AI for surveillance, weapons, or technologies that could harm people. [The Information].

Google Revenue Growth Slows Slightly, Shares Fall 7% as Investors Await AI Boost – Alphabet reported earnings that slightly beat consensus estimates, but revenues fell short due to slower growth in cloud computing revenue. [The Information].

Responsible AI: Our 2024 report and ongoing work – Google has released its 2024 Responsible AI report, detailing ongoing efforts and updates in AI development and ethics. [Google Blog].

Adobe Acrobat's New AI Assistant Enhances Contract Features – Adobe introduces an AI assistant in Acrobat to streamline contract creation and management. [The Verge].

Deepfake Videos Are Getting Shockingly Good – Advances in deepfake technology are making it increasingly difficult to distinguish fake videos from real ones. [TechCrunch].

OpenAI's Altman Meets with India IT Minister to Discuss Country's AI Plans – OpenAI CEO Sam Altman met with India's IT Minister Ashwini Vaishnaw to discuss collaboration on developing a low-cost AI ecosystem. [Reuters].

China's DeepSeek Challenges U.S. AI Dominance – Chinese AI firm DeepSeek's rapid advancements highlight China's growing capabilities in the AI sector. [Rest of World].

🦾 Emerging Tech

Figure Drops OpenAI in Favor of In-House Models – Robotics company Figure has decided to develop its own AI models, moving away from reliance on OpenAI's technology. [TechCrunch].

🤳 Social Media

Instagram Tests Profile Grid Editing – Instagram is experimenting with a feature that allows users to rearrange their profile grid posts. [Social Media Today].

YouTube Reports Record Ad Revenue in Q4 2024 – YouTube achieved its highest-ever advertising revenue in the fourth quarter of 2024, contributing significantly to Google's earnings. [Variety].

Threads Now Allows for Public Custom Feeds – Social media platform Threads introduces a feature enabling users to create and share public custom feeds. [TechCrunch].

🔬 Research

OmniHuman: Advancing Human Video Generation – The OmniHuman project presents a framework for generating realistic human videos based on minimal input signals. [OmniHuman Lab].

⚖ Legal

Judge Says Elon Musk's Claims of Harm from OpenAI Are a 'Stretch' but Welcomes Possible Trial – A U.S. District Judge expressed skepticism over Elon Musk's lawsuit against OpenAI but allowed the case to proceed to trial. [Associated Press].

🎱 Random

Spotify Reports First Year of Profitability – Spotify announces its first profitable year, marking a significant milestone for the streaming service. [The Information].

Apple Launches New 'Invites' App – Apple introduces 'Invites,' an app designed to help users plan and manage events seamlessly. [MacRumors].

🔌 Plug-Into-This

OmniHuman: ByteDance's new AI creates realistic videos from a single photo

ByteDance has introduced OmniHuman, a new AI video model making waves for it’s incredibly realistic outputs. It’s capable of transforming a single photograph into a realistic video of a person speaking, singing, and moving naturally.

• OmniHuman was trained on over 18,700 hours of human video data, utilizing a novel approach that combines text, audio, and body movements to enhance learning efficiency.

• The system generates full-body videos, enabling natural gestures and movements that align with speech, expanding applications in digital entertainment and communications.

• In testing, OmniHuman outperformed existing systems across multiple quality benchmarks, demonstrating its effectiveness in creating realistic human animations.

• The development comes amid intensifying competition in AI video generation, with companies like Google, Meta, and Microsoft pursuing similar technologies.

🎥 This newest model rising in a crowded AI video field reflects a broader trend in AI research focusing on multimodal learning, where systems are trained to process and generate content across various data types, leading to more versatile and powerful applications.

OpenAI unveils its first-ever rebrand, enhancing ChatGPT interactions

OpenAI has introduced a comprehensive rebranding initiative, featuring a new typeface, wordmark, symbol, and color palette, aiming to create a more cohesive and “human-centered” visual identity across its products, including ChatGPT.

• The rebranding effort was led by OpenAI's in-house design team, collaborating with Berlin-based type foundry ABC Dinamo and Rotterdam's Studio Dumbar for motion design.

• A key component of the rebrand is "OpenAI Sans," a bespoke typeface that blends geometric precision with a rounded, approachable character, used prominently in the new wordmark.

• The redesigned "blossom" logo features a slightly larger central space and cleaner lines, reflecting the company's evolution and commitment to “clarity”.

• This visual overhaul addresses previous inconsistencies in OpenAI's branding, unifying the aesthetic across various platforms and products.

• The updated design is now visible on OpenAI's official website and has been integrated into the ChatGPT interface, enhancing user experience.

🎨 This rebrand signifies OpenAI's transition from a research-focused entity to a product-centric tech leader, emphasizing the importance of a unified and approachable visual identity in engaging a broad user base.

EU establishes guidelines to prevent AI misuse by employers, websites, and law enforcement

The European Commission has issued guidelines to prevent the misuse of artificial intelligence (AI) by employers, websites, and law enforcement agencies. Effective from February 2, 2025, these guidelines clarify prohibited practices under the Artificial Intelligence Act, which will be fully applicable by August 2, 2026.

• Employers are prohibited from using AI to monitor employees' emotions through tools like webcams and voice recognition systems.

• Websites are banned from deploying AI-enabled "dark patterns" designed to manipulate users into making significant financial commitments.

• The use of AI to exploit individuals based on age, disability, or socio-economic status is forbidden.

• Social welfare agencies and other entities are prohibited from employing AI-enabled social scoring using unrelated personal data, such as origin and race.

• Law enforcement agencies are restricted from using AI to predict individuals' criminal behavior solely based on biometric data without verification.

🧑‍⚖️ These measures aim to provide legal certainty for AI providers and users in the European market, ensuring ethical deployment and preventing potential abuses.

 🆕 Updates

📽️ Daily Demo

🗣️ Discourse