AI Video Companies are going Bankrupt

In today's Newsletter, we are going to see some exciting updates from Runway, Apple, and Meta, so let's dive into it.In February 2024, OpenAI introduced their impressive text-to-video model called SORA. This model can generate realistic and imaginative videos, with complex movements, accurate details, and expressive characters. A few months later, a company called Kuaishou Technology released its own text-to-video model, Kling, to compete with SORA.

None of these models were available to public, and were limited access only.

You’ll soon get to know why.

Runway recently joined the race by unveiling its AI video model called Gen 3 Alpha. Gen 3 Alpha excels at creating highly realistic, consistent videos with impressive fidelity, motion, and control. Its fine-grained temporal abilities and slow-motion generation give creators more flexibility. This rapid progress underscores the immense potential of these AI technologies to revolutionize visual storytelling.

What’s the catch?

First, not available to free users.

Second, costs $2.5 / video. Yes.

We tried it out.

You can only generate upto 6 videos in their premium plan which costs $15 / month.

Also explains why SORA is not open to public.

Video Generation is expensive.

Why?

A typical video is 25fps. 25 Frames / images per second.

Each image generated by a model like SDXL is already expensive.

A 10 second video has 250 frames = 250 images = now imagine consistent generation and other complexities.

No wonder it costs $2.5 / video. Very certain they’re still not very profitable with this, which is, if they are.

More news this week, Apple's former marketing executive Phil Schiller has joined OpenAI's board as a non-voting observer. This strategic partnership signals Apple's intent to deeply integrate OpenAI's ChatGPT technology into its platforms, including Siri.

GEN-3 Alpha is Public!

Runway has made its latest text-to-video generation model, Gen-3 Alpha, widely available to users. Gen-3 Alpha can produce hyper-realistic 10-second video clips from text prompts, showcasing improved speed, fidelity, and motion compared to Runway's previous models. However, unlike Runway's free Gen-1 and Gen-2 models, Gen-3 Alpha requires a paid subscription starting at $12/month per editor. While the model has limitations compared to some competitors like OpenAI's Sora, Runway sees Gen-3 Alpha as a step towards building more advanced "General World Models" that can simulate a wide range of real-world situations. The creative community is now exploring the potential use cases for this latest AI video generation technology from Runway.

Phil Schiller from Apple Joins OpenAI Board

According to a recent report by Bloomberg, Phil Schiller, Apple's App Store chief and former marketing executive, is set to join OpenAI's board as a non-voting observer. This move is part of a broader agreement between Apple and OpenAI to integrate ChatGPT into Apple's operating systems, including iOS, iPadOS, and macOS. Schiller's new role will provide him with insights into OpenAI's operations as Apple enhances Siri's capabilities with ChatGPT's AI technology. While the current arrangement does not involve financial transactions, Apple may potentially benefit from future ChatGPT subscriptions generated through its platforms. Microsoft had previously secured a similar non-voting observer position on OpenAI's board, alongside Apple. This dual representation by tech giants could potentially complicate discussions involving both companies. It is noted that it is uncommon for Apple executives to hold board positions in companies they collaborate with.

Meta Shows labels for AI Generated Content

Meta has changed the label it uses to indicate the use of AI in photos on its platforms from "Made with AI" to "AI Info". This change comes in response to feedback from photographers who were unhappy with the previous label being applied to their photos even when they had only used basic editing tools. Meta acknowledged that the "Made with AI" label failed to provide adequate context and clarity around the extent of AI involvement in photo creation.

The new "AI Info" label aims to better align with user expectations by encompassing content modified with AI, such as photos edited with tools like Adobe's Generative AI Fill. However, Meta is not altering the underlying technology it uses to identify AI usage in photos, which relies on metadata standards like C2PA IPTC. This means that photos edited with AI-powered tools may still receive the "AI Info" label, even if the edits were minor. The change does not address the challenges of identifying fully AI-generated images or quantifying the level of AI editing in a photo.

Hand Picked Video

OpenAI’s SORA an innovative AI model capable of creating videos from text prompts. This model can generate realistic and imaginative scenes, complete with complex motions, accurate details, and expressive characters.

Top AI Products from this week 

• BrainyAI - Enjoy seamless AI chat, search, summarization, and browsing with the best AI models, including GPT-4, GPT-4o, Claude, Gemini, Moonshot, & LLaMA3, all from your browser sidebar.

• WebSim -AI Creator - Make CSS and JavaScript To Create any and every web page! Share the results!

• QR Code AI 2.0 - QR Code AI V2 is here! Next-gen AI allows you to create custom QR codes, including artistic designs. Enjoy smart link previews and advanced analytics.

• Ariglad - Powered by AI, Ariglad automatically updates your knowledge base articles, and creates new articles by analyzing support tickets and product release notes.

• Wanderboat AI - Wanderboat is a travel + outing AI companion to find and sort the best point of interest with videos, images, and insights. From signature dishes to photo spots,

• Rapport Self Service - Rapport Self-Service Rapport makes it easy to combine all of the building blocks necessary to create, animate, and deploy your own Virtual Interactive Personality (VIP).

• BuilderKit - Builderkit helps you ship an AI SaaS product super fast with Pre-Built AI tools, Pre-Built Landing Pages and a comprehensive NextJS boilerplate

This week in AI

• OpenAI's AI Critic for Improved Code Generation - OpenAI has developed CriticGPT, a novel AI model based on GPT-4, designed to identify errors in code generated by ChatGPT. This innovative approach aims to enhance the reliability and accuracy of AI-generated outputs.CriticGPT outperformed human reviewers by catching approximately 85% of bugs compared to the 25% identified by humans. The model's feedback was preferred over human critiques in 63% of cases involving natural LLM errors.

• Pinterest Embraces Video Sharing for Gen Z Users - Pinterest now lets users turn boards into shareable videos for Instagram & TikTok, responding to Gen Z's video-centric content preferences to showcase creativity, seek inspiration & engage across platforms.

• YouTube Tackles AI Impersonation - YouTube now allows users to request removal of AI-generated content impersonating their voice or face, addressing privacy concerns amid the rise of synthetic media and potential misuse, especially during elections.

• Webpage2Code: AI-Driven Web Design Translation - MBZUAI's Webpage2Code uses AI to convert webpage designs into functional code, automating web development and helping designers, devs, and non-technical users quickly bring visual concepts to life.