Video Magic

Hi there, this week we had the announcement that Google Gemini, a more capable model, benchmarks that outperform GPT-4, and a video demonstrating what Gemini is. It turns out that Google wasn’t being honest with us, the video was edited and the benchmarks are misleading, Gemini isn’t ready yet. I’m sure it will be there, let’s see what the new year brings. This week is all about AI video generation and chatbots, there have been cool advancements in video generation, Pika labs (on the waiting list), RunwayML and HeyGen are some of the tools I am exploring. I’m setting up a Chatbot for Goblin in the Attic Facebook page using Manychat, and this week I’m redesigning the website and I’m adding an FAQ bot too. I will be reviewing some of the different chatbot tools available, and I will post a tutorial on how I build my chatbot.

Full disclosure: Some of the links on this blog are affiliate links, which means that I may receive a small commission if you purchase after clicking on one of these links. This commission does not affect the price of the product or service for you, and it helps me to keep this blog up and running.

News

Google's Gemini AI model, developed for versatility and high capability, is optimized in three versions: Ultra, Pro, and Nano. Highlighted by Sundar Pichai and Demis Hassabis, Gemini excels in processing diverse information types like text, code, audio, images, and video. It achieves top performance in language understanding, reasoning, and image interpretation, functioning efficiently on various devices, from data centres to mobile phones. Gemini's applications extend to coding, mathematics, and more, supported by Google's investment in specialized hardware like TPUs. Safety and responsibility are key in its development, with Gemini 1.0 now being integrated into Google products for various uses.

Meta's new project, Purple Llama, is a comprehensive initiative aimed at responsible generative AI development. This project adopts a dual approach of offensive and defensive evaluations, inspired by cybersecurity's purple teaming, to address the unique challenges posed by Large Language Models (LLMs). Purple Llama will initially focus on cybersecurity tools and input/output safeguards. For cybersecurity, Meta is sharing valuable safety evaluations and tools across the industry to assess and mitigate LLM-related risks, such as insecure or malicious code generation. A key component of the project is Llama Guard, a foundational model designed to filter risky outputs from LLMs, which developers can customize for specific needs. Emphasizing open and transparent science, Meta is collaborating with major partners like AI Alliance, AWS, Google Cloud, IBM, and Microsoft, to establish standardized trust and safety tools in the field of generative AI.

Azure AI Studio is a comprehensive platform enabling the development and deployment of AI-powered copilot applications using GPT and other generative models. Integrated into Microsoft's ecosystem, it offers copilots in various applications, including Bing and Microsoft 365, for tasks like code generation and security response. The platform provides access to a wide range of AI models and supports integration with multiple data sets, enabling users to create sophisticated, multimodal AI applications. Azure AI Studio facilitates full lifecycle development from prompt engineering to deployment, including advanced features for data integration, evaluation, and responsible AI practices. This makes it a versatile tool for creating AI applications that can understand and respond using natural language, vision, and speech.

Rask AI's new Multi-Speaker Lip-Sync feature revolutionizes video and audio localization by enabling fluent translations into over 130 languages with realistic lip synchronization. This addresses the long-standing issue of mismatched lip movements in dubbed content, particularly in English-speaking countries. The technology, based on generative adversarial network (GAN) learning, allows users to upload videos, translate, and synchronize lip movements for more natural-looking dubbed videos. This advancement, available in beta to Rask's subscription customers, is expected to significantly enhance the viewing experience and accessibility of translated content

Tools & GPTs

  • HeyGen - AI video generation tool, this tool focuses on creating AI avatars, turning a still image into a talking moving avatar. It could be used for sales videos, staff training, tutorials, video bios and many other use cases. Review coming soon.

  • RunwayML - Image and video generation platform, with lots of great tools for generation and editing. some recent updates to video generation. create video from images or text prompts. The video below was created using text to video tool.

RunwayML

  • MagicAnimate - Animate images, turn a still image into a moving video, you make them dance for example. It’s only a demo currently, but it looks like a very promising tool.

  • Prompt master - I created this to help with the prompt generation, if you use it feed feedback would be great, it will help me refine it.

  • Pika Labs - Looks like the top player in video generation, you will have to get on a waiting list, when I get access I will do a review.

  • Mindstudio - Chatbot building tool, easy no code builder, create chatbots with many different integrations.

Let AI become your web developer! Try Pineapple builder

Projects

  • Goblin Academy - A collection of tutorials on AI, What AI is, different AIs available, AI explained the tech behind it, and how you can put it to work - use case examples, free beginner tutorials and GPT primers. more updates after the website redesign.

  • FAQ chatbot for Goblin in the Attic, trying some different tools like Make.com

  • MunchbyteTV - I starting a YouTube channel, that will showcase AI tools,

Upcoming blogs -

  • Chatbot building tools - an overview of the tools available for building

  • HeyGen - review of AI video avatar

Check out the video on Facebook -

Video generation is coming on leaps and bounds, with a variety of tools for business and content creation, watch out for some reviews coming soon, I will be demonstrating some case uses, and I will also publish some of these on YouTube. And some more detailed posts on chatbots.

All the best

Munchbyte

If you found this newsletter valuable, don’t keep it to yourself. Spread the word. Your support means the world to me. Thanks for being a part of my community!