Spext

Hey Everyone 👋 🔥Open Source Alert🔥 Some of you might already know, but for those who don't, we are the same team that built Spext (https://spext.co). With Spext, we focused on creating an amazing user experience for interacting with video recordings, scaling it to 10,000 hours of video before pivoting to VideoDB (the story behind our pivot is one for another time 📖). Over the last three years, we've developed beautiful interfaces for: ✂️ AI-driven video editing: Edit videos by editing text directly in the browser. Link to video editor 🔗 https://lnkd.in/gmveVBFm 🔍 Search across videos: Implement RAG over a collection of videos. Try it here 👉🏼 https://lnkd.in/gN8_C3r9 💬 Chat-based editing and clipping workflows for creators. Here’s an example workflow ⚡️https://lnkd.in/gUd_X_rQ 🎛️ Beautifully designed video player: A cross-platform video player that supports search and chapters. Try it here 👉🏼https://lnkd.in/gwSUQRTd We believe these UI tools offer great value to anyone working in this space, so we’ve decided to 🔔open source all of them🔔 in the coming weeks 🚀 We're excited to kickstart our open-source project on video agents, designed to help you build tools and agents that can interact with natural language and run complex video pipelines for you 🎥🤖 and these UI interfaces are going to be integral part of it. If you're interested in collaborating on any part of this project, feel free to reachout to me ✉️ #opensource #video #editor #AI #LLM #RAG #agent #infra #ML #multimodal

Prompts are crucial in extracting information from large vision models. Strategic prompt engineering can significantly enhance the ability to extract detailed information from video frames, opening up new possibilities for advanced video search and analysis. 🔦 👉🏼 Here’s a comprehensive guide on mastering video scene indexing through prompt engineering. Check it out and let us know your thoughts - https://lnkd.in/gaUT8YY2 #video #indexing #RAG #search #multimodal #AI #LLM

🌟 Exciting News! 🌟 We're kicking off a new series on unlocking 😎 multimodal search😎 with agents and assistants. 📹 🔊Videos are inherently multimodal, combining visual and audio content. Our brains naturally process this information, and with VideoDB, we can now develop assistants that replicate these cognitive processes externally. For example: ✄Locate the segment where the reporter discusses the election results and show the crowd reactions 🔦 Find the footage of the fire at the factory and the firefighter's explanation of the rescue efforts. 🔎Show the meeting segment where the CEO discusses the quarterly financial results and display the corresponding charts. ... and many more! Multimodal search enables more natural, human-like information retrieval from videos. Checkout our detailed notebooks on ▶️ Instantly access video footage with Multimodal Search Query: https://lnkd.in/gkrsZT5v 📘 Conference Slide Scraper with VideoDB - https://lnkd.in/ganeg4ky 🔔 Stay tuned as we add more content on multimodal search queries and how it can revolutionize video content accessibility! #VideoDB #MultimodalSearch #AI #TechInnovation #Developers #VideoRetrieval #Video #ContentRetrieval #LLMs #multimodal

While using video search APIs, developers face two significant problems: 1. Lack of Configurability: Developers can't configure which frames of the videos to select for optimal indexing. 2. Limited Output: The APIs return only timestamps, leaving the burden on developers to figure out clipping, compilation, & streaming of the actual video content themselves. Today, VideoDB is launching "Scene Index" to take these challenges head on 🎉 Check out the documentation 📄 https://lnkd.in/g_WktbQ5 🔎 Scene-Level Indexing: Define scenes, generate descriptions, capture activities 📝 Custom Annotations: Add your data and context ⚡️ Visual Indexing Pipelines: Integrate custom vision models. 💻 Experimentation Playground: Test scene extraction and prompts. #video #genAI #LLM #multimodal #openAI #gemini #search #RAG

Finding visual information in videos opens up a world of possibilities. Imagine searching your large video library for queries like: "show me where birds are flying near a castle" "Show me scene with explosion" "Show me when the person took the gun out" "Show where people are running towards the sea" We have been engaged in expanding our visual indexing pipelines by working with a few early customers. This helped in making the interface flexible and powerful. VideoDB now supports: 🔎 Scene level indexing options: You have ways to define and extract scenes and generate it’s description according your prompt. It also capture activities like running. 📝 Custom Annotations to bring your own data, application context and annotations to help search information. ⚡️Advance visual indexing pipelines that integrate well with custom vision models of your choice. 🤔 A playground for experimenting with scene extraction and prompts. We are very excited to work on innovative usecases 🙌🏼. Check out our Scene Index Documentation- https://lnkd.in/gCCssexp We’re eager to hear your feedback and suggestions for improvement 🫡 Thanks for being a part of our journey towards revolutionizing video search! #video #genAI #LLM #multimodal #openAI #gemeni #search #RAG

On November 30th, 2023, I announced to my team that we were pivoting our startup to focus on VideoDB. Just one month later, on December 31st, I was reviewing the first set of tutorials we planned to share for the launch of VideoDB. The pivot from our previous product, Spext, was a challenging one. We opened up our infrastructure to everyone for the first time, and the whole team worked tirelessly for the past few weeks to make the first launch of VideoDB a reality. The energy in the team was electric as we approached the big day. ⚡️ The best part was the early response we received, not just from new developers, but also from seasoned professionals and companies who had been in the infrastructure technology space for over 15+ years. Our unique, first-principles approach of removing the need for direct interaction with mp4 files for developers, and providing a simple, intuitive interface that abstracts away the complexities, resonated strongly with the community. Innovation is hard and building simple solutions is never easy! VideoDB is our life's work. We applied the insights and learnings we had accumulated over the past 5+ years, fuelled by hard work and many sleepless nights. As founders, we had faced multiple cash-out moments, gone without salaries for years, and even supported the team from our personal funds to keep the dream alive. But nothing beats the daily joy of seeing the growth with search queries on VideoDB skyrocketing from tens to hundreds to thousands in just 3 months. 🚀 This makes me immensely proud of the commitment of our team and deeply grateful for the love and support we're receiving from our customers 🙏🏼 500+ devs on our platform, 150+ on our discord community and 500+ combined stars on our github repos and 1000+ search/day 🙌🏼 These developers are building solutions across a wide range of industries, including OTT platforms, e-learning platforms, real-time analytics and visualization (RAG) applications, chat companies, video surveillance systems, AI consulting firms, and even developers from companies like ByteDance. It's been an incredible journey, and we're excited to see what the future holds for VideoDB 🫡

🔥 Add styled captions to your videos in just 2 lines of code in real-time!! Automated captions on platforms like YouTube often lack precision, especially with specialized terminology or context-specific information 🧐 👌🏻 VideoDB offers the best interface to developers by enabling more accurate identification and spelling. The captions can be styled according to your video's theme or your brand's guidelines, making them visually appealing and consistent 👌🏻 The styling options include: * Font: Choose the font name from a list of supported fonts, and set the font size in pixels. * Colors: Customize the color of the primary text, secondary effects, outline, and background using hex color codes. * Text Styles: Enable bold, italic, underline, or strikethrough formatting. * Scaling: Adjust the horizontal and vertical scaling of the text, as well as the spacing between characters. * Rotation: Specify the angle of text rotation in degrees. * Border: Set the style and width of the text outline. * Shadow: Adjust the depth of the text shadow. * Alignment: Positioning the text on the screen (e.g., bottom-left, bottom-center, bottom-right). * Margins: Customize the left, right, and vertical margins around the text. These options allow you to create captions that seamlessly integrate with the visual style of your video. You can explore these styling options and see examples in this guide - https://lnkd.in/gPkVsQv4 #AI #VideoDB #LLM #video #content #creators #devtool #subtitles #captions

Thrilled to share how ElevenLabs and VideoDB are revolutionizing storytelling with AI-generated voiceovers for silent footage🥷🏻 The future of video storytelling is going to be ⚡️ efficient, expressive, and effortlessly yours 🤗 Use VideoDB's programmable video streams with Eleven Labs' voice gen technology to create captivating narratives in just few lines of code👩💻 without any manual effort! Here’s the link to tutorial 👉🏼 https://lnkd.in/g4NKuMgq Link to the video ▶️ https://lnkd.in/gkDZmzja #ai #genai #elevenlabs #videogeneration #VideoDB #llm #chagpt #sora #openai

Exciting News 🎉 Now, instantly search and stream videos with LlamaIndex & VideoDB and make your RAG #multimodal. Our integration with LlamaIndex unlocks a completely new dimensions in RAG systems! With VideoDB, you can not just analyse but instantly watch video streams of your search results 📺 without worrying about the hassle of building & deploying complex video infrastructure. ⏭️ VideoDB provides an easy to use database abstraction enabling the power of LLMs on your video data in just few lines of code. Checkout our llamaIndex retriever 👉🏼https://lnkd.in/grHR2rsC Our team loved working with llamaIndex codebase. Such a beautifully written framework 💛 Thanks for democratising RAGs Jerry Liu 🙌🏼 Ravi Theja Desetty and the whole team! #rag #VideoDB #ai #llm #video #search #genai #multimodal

If you are editing & producing videos and haven't used AI, it's time to quickly get on board. I created all these clips with just a "text prompt" ⏬ in less than a minute ⏱️ "Find every moment where a deal was offered" Shark Tank ABC "Show me the feedback from every judge" MasterChef Watch more examples here 👉🏼 https://lnkd.in/gfYTddyC Link to the opensouce repo ➡️ https://lnkd.in/gBuD-9kg The future is here ⚡️ Intelligent automation is reshaping the whole media industry! #ai #videodb #llms #multimodalai #multimodal #videoeditors #videoclip