- ?? Great advancements from Google! Their new Video-To-Audio (V2A) technology combines video pixels with text prompts to create rich, synchronized soundtracks for silent video clips. Imagine bringing your silent films to life with realistic sounds for horror scenes, drumming performances, or even a howling wolf.
- ?? Exciting news from ElevenLabs! They've launched a text-to-sound effects API, including the first-ever video-to-sound effects app, available for free and fully open source. Just upload your video, and their AI generates the perfect sound effects in about 15 seconds, making it easy to enhance everything from action scenes to memes. Whether you’re a filmmaker or just having fun, you can try it out at videosoundeffects.com.
- ?? TikTok has introduced a new AI feature called Symphony avatars. This allows users to create custom avatars or use pre-built ones that can speak on video using generative AI technology. Users can upload text, which the avatars will then speak, offering a new way to engage with content. This development highlights TikTok's ongoing innovation in AI-driven features.
- ?? This week saw a surge of new open-source AI models released by companies like Apple, Microsoft, Meta, and Nvidia. Apple introduced 20 machine learning models on Hugging Face, including depth estimation, semantic segmentation, and Transformer-based language models. Microsoft launched Florence 2 for vision tasks, Meta unveiled Chameleon for multimodal inputs, and Nvidia introduced Neotron 4 with a massive 340 billion parameters, all accessible for experimentation and development.
- ?? Anthropic has introduced Claude 3.5 Sonnet, a significant advancement in AI technology that surpasses its predecessor, Claude 3 Opus, in intelligence, cost-effectiveness, and operational speed. The new model, available for free on Cloud, showcases notable improvements in vision models and benchmark performance, surpassing both CLA 3 Opus and GPT 4omni across various metrics. What sets Claude Sonnet 3.5 apart is its enhanced capabilities in graduate-level reasoning, undergraduate-level knowledge, and coding proficiency. This model excels in understanding nuances, humor, and complex instructions, delivering content with a natural and relatable tone.
- ?? Ilya, who played a role in changes at OpenAI, has reemerged with a new venture called Safe Superintelligence Inc. (SSI). They're focusing on a significant challenge: creating safe superintelligence. It's based in Palo Alto and Tel Aviv, prioritizing safety and progress over immediate commercial gains. It'll be interesting to see how they navigate funding and their approach to building towards superintelligence.
- ?? Apple has announced a new feature through Apple Intelligence that will tag AI-generated images in metadata, responding to concerns about accurately labeling AI content on social media platforms. This step is part of a broader industry trend to ensure transparency in distinguishing between AI-generated and non-AI content, especially on platforms like Instagram, where even minor edits can trigger AI-generated tags. However, how this metadata will be integrated and communicated on social media platforms is yet to be determined.
- ?? Perplexity.ai introduced streamlined updates this week, offering direct results for weather, currency conversion, and simple math queries without external search engines.
- ?? The CMR M1 world's first AI Cinema Camera that seamlessly integrates AI technology into the video capture process. This experimental prototype, created in collaboration between special guest X and First Avenue Machines. With features like AI stylization, interchangeable lenses, and a rotary knob for adjusting AI levels, this camera sets the stage for a new era of creative filmmaking.
Senior Software Architect
8 个月this is a very interesting and informative article. i used some of these tools they are cool by the way. prompt engineering is the game for using this generative technology . now after you generate the content , you will ask yourself is it pretty? is it a good content? is it a high quality content . This is the main issue in using these it needs a creative person with some sense of quality and beauty.