VideoPoet – A Revolutionary Tool by Google for Zero-Shot Video Generation

Google Quietly Expands Imagen 3 Availability to U.S. Users

Google unveiled Veo, an AI generator, image creator, and music-making toolbox, at Google I/O in May. The video generator can transform text prompts into 1080p videos longer than 60 seconds. "The creative potential for generative AI is immense, and we can't wait to see how people around the world will bring their ideas to life with our new models and tools," Google said in May when it announced the tool. Back in May Google made the tool available to "select creators in private preview," now, after some updates and a name change over the past few months, it's making the tool available to everyone through its ImageFX platform, VentureBeat reports. Now called Imagen 3, Google describes the tool as " a latent diffusion model that generates high-quality images from text prompts." According to the company, "Imagen 3 is preferred over other state-of-the-art models at the time of evaluation." Users of the tool have remarked that the tool has improved word recognition and texture over previous models; however, some users have expressed frustration over the tool's censors. The tool is designed to block users from creating some videos as part of Google's focus on responsible AI creation; however, a number of users have reported that their prompts, which shouldn't be considered problematic, are also being blocked. In contrast, xAI also launched Grok-2 this week. The feature is available exclusively through Xs Premium subscription tiers and offers image generation with relatively no restrictions.

PC Magazine

Sat, 17 Aug, 8:00 PM UTC

People are generating entire short films using new AI app | Digital Trends

Alibaba- and Tencent-backed startup Minimax, one of China's "AI tigers," has released its Video-01 text-to-video model, which can generate highly accurate depictions of humans, down to their hand motions. Minimax unveiled the new tool Saturday at its inaugural developer conference in Shanghai. Recommended Videos Video-01 enables users to create 6-second 1280 x 720 resolution videos at 25 frames per second (fps) using text prompts, reportedly in about 2 minutes. At the conference, Minimax CEO Yan Junjie pointed out that this is only the first iteration of the model and that future updates will include the ability to generate video from images and edit those generated videos, per the South China Morning Post "We have indeed made significant progress in video model generation, and based on internal evaluations and scores, our performance is better than that of Runway in generating videos," he told reporters. To try Video-01 for yourself, head to the Video-01 website. While writing of this article, the site was inundated with users, so the video-generation process took closer to 5 minutes but did not require me to enter a mobile number or otherwise log in to use it. As with other leading video AIs like Kling and Runway, you can dictate cinematic camera movements in your text prompt. The resulting video quality was pretty good, strikingly lifelike, and did not present any significant visual hallucinations commonly found in similar free AI tools. The video is also downloadable as an .MP4 file. Video-01 joins an increasingly crowded video-generation market, even without the highly anticipated arrival of OpenAI's Sora. Two other "Tigers," Shengshu AI and Zhipu AI (both of which are valued at more than $1 billion, released their own text-to-video tools in July, as did Kaishou Industries with Kling.AI. Byte Dance also recently released its Jimeng text-to-video app on the Apple and Android app stores, though it is only available to users in China. Stateside, Minimax will have to contend with the likes of Luma Labs' Dream Machine and Runway's Gen-3.

Digital Trends

Thu, 5 Sept, 6:07 PM UTC

How few-shot learning with Google's Prompt Poet can supercharge your LLMs

Join our daily and weekly newsletters for the latest updates and exclusive content on industry-leading AI coverage. Learn More Prompt engineering, the discipline of crafting just the right input to a large language model (LLM) to get the desired response, is a critical new skill for the age of AI. It's helpful for even casual users of conversational AI, but essential for builders of the next generation of AI-powered applications. Enter Prompt Poet, the brainchild of Character.ai, a conversational LLM startup recently acquired by Google. Prompt Poet simplifies advanced prompt engineering by offering a user-friendly, low-code template system that manages context effectively and seamlessly integrates external data. This allows you to ground LLM-generated responses to a real-world data context, opening up a new horizon of AI interactions. Prompt Poet shines for its seamless integration of "few-shot learning," a powerful technique for rapid customization of LLMs without requiring complex and expensive model fine-tuning. This article explores how few-shot learning with Prompt Poet can be leveraged to deliver bespoke AI-driven interactions with ease and efficiency. Could Prompt Poet be a glimpse into Google's future approach to prompt engineering across Gemini and other AI products? This exciting potential is worth a closer look. The Power of Few-Shot Learning In few-shot learning, we give the AI a handful of examples that illustrate the kind of responses we want for different possible prompts. In addition to a few 'shots' of how it should behave in similar scenarios. The beauty of few-shot learning is its efficiency. Model fine-tuning involves retraining a model on a new dataset, which can be computationally intensive, time-consuming, and costly, especially when working with large models. Few-shot learning, on the other hand, provides a small set of examples with the prompt to adjust the model's behavior to a specific context. Even models that have been fine-tuned can benefit from few-shot learning to tailor their behavior to a more specific context. How Prompt Poet Makes Few-Shot Learning Accessible Prompt Poet shines in its ability to simplify the implementation of few-shot learning. By using YAML and Jinja2 templates, Prompt Poet allows you to create complex, dynamic prompts that incorporate few-shot examples directly into the prompt structure. To explore an example, suppose you want to develop a customer service chatbot for a retail business. Using Prompt Poet, you can easily include customer information such as order history and the status of any current orders, as well as information about current promotions and sales. But what about tone? Should it be more friendly and funny, or formal? More concise or informative? By including a "few shots" of successful examples, you can fine-tune the chatbot's responses to match the distinct voice of each brand. We can also provide the model with our real customer data, so that it can answer questions accurately about the user's past and current orders. If appropriate we can also include current sales and promotions. But first, we need to provide the examples of style, tone and voice that we promised the model in the base instruction. Setting the tone: friendly or formal? Some brands have a more casual, friendly and jocular tone, whereas others prefer a more refined elegance. For a brand that has a friendly and casual tone, the few-shot learning examples could look like this: For a brand that emphasizes a more formal and professional tone, different answers can be provided to the very same prompts, guiding the model toward a different style: Bringing it together The code below uses Prompt Poet's 'Prompt' class to form a single, coherent prompt from the separate elements of the base instructions, the few-shot learning examples, and the actual user data. This allows us to invoke AI responses that are accurately informed and highly crafted in purpose and style. Elevating AI with Prompt Poet Prompt Poet is more than just a tool for managing context in AI prompts -- it's a gateway to advanced prompt engineering techniques like few-shot learning. By making it easy to compose complex prompts with real data and the voice-customizing power of few-shot examples, Prompt Poet empowers you to create sophisticated AI applications that are informative as well as customized to your brand. As AI continues to evolve, mastering techniques like few-shot learning will be crucial for staying ahead of the curve. Prompt Poet can help you harness the full potential of LLMs, creating solutions that are powerful and practical.

VentureBeat

Sat, 7 Sept, 2:00 AM UTC

Adobe Firefly Expands into AI-Generated Video, Challenging OpenAI's Sora

Adobe announces the addition of AI-generated video capabilities to its Firefly platform, positioning itself as a competitor to OpenAI's Sora in the rapidly evolving field of artificial intelligence-driven content creation.

22 Sources

Wed, 11 Sept, 4:07 PM UTC

Google Veo: The ultrarealistic AI video generation tool explained

Generative AI models, like Bing Image Creator and Google Images, made headlines in 2023. These AI models were integrated into Google Pixel phones with features like Reimagine. But the tech world never stands still, and the focus has shifted to artificial intelligence video generation. OpenAI jumped ahead, introducing its text-to-video tool, Sora, in February 2024. Now, Google is stepping up the competition. Let's look at Google Veo, its video generation tool. Related What is Bing Image Creator? Discover Bing Image Creator's AI power, turning your words into beautiful visuals Google Veo: Turning text into video Google Veo, introduced at Google I/O 2024, is an AI video generator that turns text prompts and visual references into videos. It creates 1080p videos that last more than a minute, using language processing and visual rendering to produce realistic scenes. It also interprets detailed instructions to add effects like time-lapse, slow motion, or other cinematic touches. For example, a prompt like a time-lapse of a flower opening produces a polished, professional video. You can also fine-tune the video results by adding more prompts. Source: Google Image-based video creation Alongside text prompts, Veo combines images with text to generate videos that carry the visual style of your image. This feature is ideal for creators looking to maintain a consistent visual style across their projects or add motion to static visuals. Close Masked editing capabilities Google Veo includes a masked editing feature that modifies specific areas of a video without affecting the entire scene. For example, in a beautiful drone shot of a coastline, you might want to add kayaks to the water. Veo does this without changing the rest of the video. Close Consistent frame generation Thanks to latent diffusion transformers, Google Veo keeps visual elements consistent across video frames. For example, when working on a video with complex movements, like a car driving through a cityscape, Veo keeps the car consistent from one shot to the next. This feature leads to a fluid, professional video, allowing creators to produce high-quality content with fewer technical glitches. Extended video length Google Veo's ability to generate videos over sixty seconds in length opens new possibilities for storytelling. Combining prompts allows you to create a continuous narrative that adds depth to your video projects. For example, you can make a video that tells a story in chapters, such as a day in the life of a city, from sunrise to sunset, within one continuous video. It's an excellent tool for adding structure and flow to longer projects. Advanced filmmaking potential Google Veo understands cinematic terms and seamlessly integrates these elements into your video. You can prompt filmmaking instructions that let you add professional effects like closeup shots or panning sequences. In partnership with Google, Donald Glover and his studio Gilga explored Google Veo's filmmaking capabilities. They experimented with creative techniques like dynamic tracking shots, where precise camera movements and consistent framing are fundamental. The partnership highlights Veo's potential in pushing creative boundaries within the film industry. The technology behind Google Veo Google Veo is built on years of research from earlier projects like Generative Query Networks (GQN), Imagen-Video, Transformer models, and Gemini. These projects helped Veo interpret prompts with impressive accuracy, generate realistic motion (even with physics simulations), and ensure frame consistency to avoid flickering or object morphing. Here's how the architecture works. Related What is a Transformer model? Decoding Transformer models: The tech shaping future communication Input and encoding When you provide a text prompt (optionally, an image prompt), Google Veo uses separate encoders for each. The text is processed using a UL2 encoder, while the image (if provided) goes through a dedicated image encoder. The outputs from these encoders combine into a single embedded prompt, which acts as the input for video generation. Latent diffusion The embedded prompt passes through a latent diffusion model, which generates compressed video frames from the embedded representation. This allows Veo to create videos quickly and efficiently without sacrificing visual detail. Decoding and output After generating the latent video, a decoding step turns the compressed representation into a full 1080p video. This process ensures the final output is clear, visually stable, and coherent throughout the video sequence. Source: Google How to access Google Veo and join the waitlist Google Veo is accessible to a small group of creators, with public access pending. It's being rolled out through VideoFX, a new experimental tool in Google Labs. Creators who want to get their hands on it early can join a waitlist to test and explore the platform. While Veo is in limited preview for select creators, Google hinted that its text-to-video generation and editing tools might be added to YouTube Shorts and other products. Google Veo prevents misuse with built-in safety features Google Veo is responsible by design and includes multiple safety features to ensure responsible use. One key tool is SynthID, which embeds a unique watermark into AI-generated videos. This allows the content to be identified as AI-created, preventing misuse and offering transparency regarding where the media comes from. On top of that, Veo has built-in protection against bias, copyright, and privacy through memorization checks. This ensures the model doesn't accidentally reproduce copyrighted material or sensitive data, keeping generated videos original and compliant with legal standards. Filters are in place to block inappropriate or harmful content. Exploring the future: Potential uses for Google Veo With the anticipated introduction of Google Veo, several industries could be on the brink of transformation. Let's explore the sectors where Google Veo could make an impact. Marketing and advertising Google Veo could be a game-changer in marketing by automating the creation of dynamic, personalized video ads tailored to viewer preferences. Brands could use Veo to tell stories that resonate emotionally, strengthening their brand identity by building a deeper connection with their audience. Education Google Veo could make educational content more interactive and illustrative. This tool can break down complex subjects into understandable segments through engaging visuals and animations, making learning more accessible and enjoyable for students of all ages. Entertainment Google Veo could transform how content is produced for social media and streaming services by tackling complex scene production head-on. Typically, expensive and challenging scenes could be accessible and cost-effective to film with Veo. Its potential in developing interactive media could provide novel experiences in video games and virtual reality. A new era in AI video creation is coming Google Veo marks a step forward in generative AI technology. The anticipation builds as we eagerly await its release. Who will emerge as the leader in the generative video model space? Will it be OpenAI, or will Google take the mantle? The competition is heating up, and we can't wait to explore each platform firsthand.

Android Police

Fri, 13 Sept, 6:06 PM UTC

Creative and design

Your one-stop AI hub

The Outpost

News

About

VideoPoet by Google

Contact for Pricing

About VideoPoet by Google

Related fields

Related News

Similar products

Video GPT

Text to Video AI

Gen-2 by Runway

VideoMaker - by Invideo AI

Make a Video

Subscribe to our newsletter