Canva goes all-in on AI. Perplexity introduces a blog product.

Happy Friday!

The race to go from 2D to 3D is heating up. Gaussian splatting, a term you might be hearing more often over the next few years, was introduced in a 2023 paper to describe a rasterization technique for rendering photorealistic 3D models from 2D images in real-time.

Gaussian splatting uses millions of tiny Gaussian particles to represent 3D scenes, achieving high quality and speed. The process involves structure-from-motion to estimate point clouds, converting points to Gaussians, and training with stochastic gradient descent. Take a deeper dive in an intro here.

However, while this technique shows a ton of promise, it currently requires high VRAM and is not yet integrated into mainstream pipelines. Check out a few recent Gaussian splatting demos for the future of street view and products for e-commerce videos.

Call for Strange Design Hackers! Thinking of building an application or exploring use-cases in the future of video, multimodal AI, or brand? Join us this summer as we bring together Strange Design Hackers and AI founders to leverage a design sprint method to uncover real needs and novel interfaces.

Apply by June 9: design.strangevc.com

Latest News

French startup AniML launches Doly: AniML has introduced an iPhone app named Doly that simplifies the creation of 3D product videos. By utilizing Gaussian splatting, an AI technique that estimates 3D shapes from 2D photos, Doly produces high-quality 3D models. Users can seamlessly integrate objects into 3D scenes from a comprehensive library, making 3D content creation more accessible than ever.

Researchers introduced SignLLM: In a significant leap for accessibility, SignLLM is the first multilingual Sign Language Production (SLP) AI model. It can generate avatar videos of sign language gestures from prompts in eight different languages, bridging communication gaps and fostering inclusivity.

Canva’s new AI features: Canva continues to innovate by integrating new AI features directly into their platform. Magic Media creates images and icons

from prompts, Magic Switch generates alternative media forms swiftly, Magic Write crafts personalized text from a short writing sample, and new AI video tools like 'Highlights' and 'Enhance Voice' improve video and audio quality.

Suno and YouTube Music’s music-to-music feature: Suno has introduced a fascinating feature allowing users to create songs from any sound. Meanwhile, YouTube Music's new hum-to-search update lets users find songs by simply humming or singing.

Perplexity Pages: Perplexity Pages is a new addition worth exploring. The next-gen research engine released its Pages feature, a way to create and compile publications and blogs and share them with the world.

ViViD, a impressive project from Alibaba researchers showcases how AI can create moving 3D representations of clothing from a flat lay photograph.