⚡️ Realtime GenAI

It’s Friday. The realm of generative AI art just hurdled into an exhilarating future with the introduction of LCM-LoRA—a game-changing ML technique that slashes the time it takes to bring AI-generated visuals to life. Let’s dive in.

Today’s Highlights:

• DeepMind's Lyria AI ushers in new era of music

• Menlo Ventures’ whopping $1.35B investment fund

• YouTube's experiments with AI-generated music

DEEP DIVE

Realtime Generative AI Art Revolutionizes Creation

Latent diffusion models (LDMs) are already a game-changer in the world of GenAI, producing realistic images from multimodal inputs. However, they are notoriously slow and memory intensive, requiring hundreds of inference steps and extensive GPU resources.

Enter, LCM-LoRA. LCM-LoRA is a universal stable-diffusion acceleration module that can speed up LDMs by up to 10 times—while maintaining (or even improving) the image quality. Furthermore, it could drastically cut production costs and times, revolutionizing industries beyond the art world, from gaming to filmmaking.

What's turning heads isn't just the fact that LCM-LoRA revs up the efficiency. It's the model's plug-and-play nature. Got a fine-tuned LDM for, say, generating anime characters from descriptions? Plug in LCM-LoRA, and voilà—up to ten times faster without added training.

At its core, LCM-LoRA is about training select 'LoRA layers' instead of the entire behemoth LDM. Nestled within the LDM's convolutional blocks, these layers echo the full model's outputs but with far less computational fuss.

Thanks to the LCM-LoRA technique, you can paint simple, almost stick-figure-like drawings alongside descriptive text and apps like Krea.AI and Fal.AI will automatically render different, new, generated art instantaneously—even swapping out the imagery in fractions of a second as the user moves their shapes or paints simple lines on their digital canvas.

Keep Reading Here.

PUNCHLINES

Artificial Trickery: Stable Diffusion and DALL-E 2 can be tricked into generating disturbing images showing violence and nudity.

Byte-Sized Brains: Qualcomm announces new Snapdragon 7 Gen 3 chipset with AI acceleration.

Technical Difficulties: Google's competition with OpenAI reportedly hits a speed bump, delaying the release of its Gemini AI.

Gold Rush: Menlo Ventures raises $1.35B in funds for AI Investments focused on nascent startups and early-stage investments.

GPT-Lover: China leads the world in ‘ChatGPT’ searches with a peak popularity score of 100.

TLDR

YouTube's experiments with AI-generated music: YouTube introduces AI experiments letting users generate music in celebrity voices or from hummed tunes. The "Dream Track" feature allows the creation of stylized music tracks for YouTube Shorts, while "Music AI Tools" assist artists' creativity, using Deepmind's Lyria system and SynthID for audio watermarking to secure AI-generated works.

DeepMind's Lyria AI ushers in new era of music: Google DeepMind's Lyria, a groundbreaking GenAI model, can craft vocals, lyrics, and music mimicking famous artists. With YouTube Shorts collaborations, it pushes for creative expansion and fresh industry dynamics. Lyria incorporates ethical AI practices with content watermarking to track origins and encourages responsible AI use in creative arts.

Aussie breakthrough in perovskite solar cells: A team of Australian researchers has used AI to speed up the creation of perovskite solar cells, cutting down development time from years to weeks. Their ML model accurately predicts chemical compositions for new cells and has achieved a record 16.9% power conversion efficiency.

Meta unveils Emu AI for creative content: Meta introduces Emu Edit and Emu Video, cutting-edge AI tools for text-to-video generation and image editing. Training on 10M samples, Emu Edit can edit images with text prompts, while Emu Video simplifies creating videos from text or images. These tools could revolutionize user interactions with visual content on platforms like Facebook and Instagram.

TRENDING TOOLS

🤖 Chatling: Craft AI chatbots effortlessly without coding

🎨 Chat2Design: Instantly convert text into sleek UI designs

✉️ Auto Summarize by Superhuman: Auto-summarize your emails

💡 OpenOpenAI: Run OpenAI’s Assistants API on your own servers

🌐 Netmind Power: Dive into decentralized machine learning and AI platform operations

That’s all for today—if you have any questions or something interesting to share, please reply to this email. We’d love to hear from you!

P.S. If you want to sign up for the Supercharged newsletter or share it with a friend, you can find us here.