Public Release: Kling AI Video Generator

Author

AI Breakfast
July 26, 2024

Good morning. It’s Friday, July 26th.

Did you know: On this day in 2016, Verizon announced that it would acquire Yahoo's operating business for $4.83 billion.

In today’s email:

  • SearchGPT

  • Kling AI Video Generator is Live

  • Udio AI Music Upgrade

  • 5 New AI Tools

  • Latest AI Research Papers

You read. We listen. Let us know what you think by replying to this email.

Today’s trending AI news stories

OpenAI announces SearchGPT, its AI-powered search engine that offers real-time access to internet information. Unlike its keyword-centric predecessors, SearchGPT employs the sophisticated GPT-4 architecture to coalesce disparate information into coherent, attribution-rich summaries. Currently undergoing closed beta with a select cadre of 10,000 early adopters, SearchGPT integrates features like "visual answers," which may include AI-generated videos and multimedia content. Read more.

Kling AI Video Generator Now Available Globally to Take on OpenAI's Sora. The platform, previously limited to China, now allows users to generate videos using both text-to-video and image-plus-text-to-video capabilities, providing competition for OpenAI's Sora.

Users receive 66 free daily credits and can create videos in three resolutions: 960x960 square, 720p landscape, and 720p portrait. Kling AI supports English and Mandarin prompts, maintaining high quality across languages. Users can adjust creativity levels to influence the video's alignment with prompts and use negative prompts for enhanced accuracy. Notably, Kling AI can generate realistic videos, as demonstrated by its accurate portrayal of wolf pups on a gravel road.

However, it sometimes struggles with consistency in minimal detail animations, requiring post-production corrections. The platform plans future updates for HD video production and additional features, with content restrictions in place to prevent inappropriate or unauthorized content creation. Read more.

Udio v1.5 music generator narrows the gap between human-created and AI-generated music enhancing the quality of AI-produced music to closely resemble human-created tracks. The update introduces improvements in audio fidelity, including clearer vocals and better instrument separation in 48 kHz stereo tracks. The new version offers a warmer sound with reduced mechanical qualities, making AI-generated vocals less distinguishable from human performances. Read more.

Microsoft unveils serverless fine-tuning for its Phi-3 small language model streamlining the adaptation of AI models without managing servers. This update allows developers to fine-tune Phi-3 models, including Phi-3-small and Phi-3-vision, directly via Azure AI’s serverless endpoints. Pricing for serverless fine-tuning starts at $0.004 per 1,000 tokens. Read more.

DeepMind hits milestone in solving maths problems — AI’s next grand challenge as it has cracked open four of the six problems presented by the 2024 International Mathematical Olympiad. The system, which integrates reinforcement learning and formal proof techniques, demonstrated high proficiency in algebra, number theory, and geometry, earning a score of 28/42, just short of gold-medal performance. Despite these advances, AlphaProof struggled with combinatorics, indicating areas for further development. Read more.

Google's free Gemini chatbot gets 1.5 Flash update, making responses faster and smarter with a 32,000-token capacity—four times more than its predecessor. Users will experience improvements in response speed, reasoning, and image understanding across the free versions of the Gemini mobile app and web interface. New functionalities include file uploads for contextual prompts and data file analysis for visualizations, features previously limited to paid users. Google is also tackling AI hallucinations by incorporating source citations for all responses, including those within Gmail integration. Read more.

Nvidia's latest AI offering could spark a custom model gold rush with its new AI Foundry service, offering businesses a turnkey solution to forge custom language models tailored to their idiosyncratic needs. This service integrates Nvidia’s hardware, software, and expertise, allowing companies to customize popular open-source models such as Meta's Llama 3.1. Alongside this, Nvidia has launched Nvidia Inference Microservices (NIM), which packages customized models into containerized formats with API access for easier deployment. Read more.

Video game performers to strike over AI concerns starting Friday morning after contract talks with major game studios failed. The dispute centers on AI protections, with the union arguing that current proposals inadequately cover performers’ rights. SAG-AFTRA contends that the studios’ definitions of who qualifies as a “performer” and the potential for AI to replicate voices and likenesses without proper compensation remain unresolved. Read more.

Etcetera: Stories you may have missed

5 new AI-powered tools from around the web

MrScraper simplifies data scraping with an AI-powered, no-code tool and API, bypassing captcha blockers and scaling data collection effortlessly.

HackerPulse consolidates GitHub, Stack Overflow, and LinkedIn profiles into a single developer profile, enhancing job applications and networking.

Arcane uses AI to integrate trend research, audience insights, and content repurposing tools into a single content marketing platform for startups.

DepsHub simplifies dependency updates with AI, prioritizing security and noise reduction. It supports monorepos, multi-language projects, and intelligent update scheduling.

Mureka integrates AI music generation, editing, and copyright trading, enabling creators to convert inspirations into high-quality musical works efficiently.

Dough is an open-source tool for creating AI-generated animations, offering precise motion control through Steerable Motion and Motion Director features.

arXiv is a free online library where researchers share pre-publication papers.

Thank you for reading today’s edition.

Your feedback is valuable. Respond to this email and tell us how you think we could add more value to this newsletter.

Interested in reaching smart readers like you? To become an AI Breakfast sponsor, apply here.