MC.15: KREA's Beta, Together's Speed, and Alibaba's AI Animation

Tech updates & tools

AI Models in Software UI
This post explores a variety of UI modalities that enable user interaction with AI, unveiling a world of possibilities. We are still in the exploratory phase and I think we might discover other ways. I'm particularly intrigued by a shift where the user becomes not the initiator, but rather the recipient of AI interactions. MORE

Together: Mixtral MoE available at 100 tokens/s
If you haven't yet explored this platform, there's no better time than now. You can try this mode which is better than GPT 3.5 on most benchmarks. Its speed is breathtaking. I'm eager to see how much fine-tuning it will undergo.. MORE

TLDRAW Lens: A Fun Artistic Experiment
Imagine a scrollable canvas that fosters collaboration and creativity where you can paint your imagination. Your drawing is interpreted and images are generated on-the-fly for an animated experience. TLDRAW has been one of the most exciting experiments in the past month. PLAY 

KREA: The Beta is Open
You can try real-time compositing with parts generated by AI. It feels magic and the response time is fast just like you can feel with the TLDRAW experiment. It's not quite ready to replace DALLE, particularly since it struggles to generate clear text, such as the issue numbers for this newsletter. MORE

Fal.ai: Build Real-Time AI Image App
Talking about KREA and TLDRAW, why not embed similar features in your app? With Fal.ai, touted as the market's fastest real-time inference engine, integrating these features seems effortlessly achievable. They provide a REST and websockets API at a fair price. MORE

Alibaba: Animate Anyone
Generative AI image and videos are going strong. Using pose-sequences, the model control how to animate a reference image. I'm curious about the transformative impact these technologies will have once they mature and become widely accessible. Will you create your own cartoon or anime serie? MORE

Lindy.ai: Agents at your Service
This service lets you build a team of AI employees working together. I haven’t been able to try the service yet but the demo looks promising. What I like is how agents can react to external events like emails, webhooks and so on. MORE | DEMO

Subclippy: Edit Videos with Python and AI
A Python script that uses AssemblyAI and GPT-4 to auto-edit long videos into sub-clips. At a high-level, it download video and audio from YouTube, transcribe the audio, select words/phrases for extraction, turn phrases into timestamp and clip videos with FFmpeg. Numerous content creators are already leveraging this kind of service, an easy business would be to make a minimal and cheap SaaS service using open-source tech to displace the current leaders. MORE | GITHUB