Real-time Text-to-Speech Model for lifelike conversations

In today’s email:

📣 A Text-to-Speech Model that is built for carrying out conversations like a real person in real-time
😻 The famed Stanford Smallville is officially open-source
📚 How to implement a paper without crying
🛠 Various AI-related tools and platforms, including
- A tool to write SQL without knowing SQ
- A video editor that is just like a text editor
- Your autonomous podcast production intern
- An AI tool for conditional music generation
- An AI job interviewer you can self-host
- and more.

PlayHT 2.0 Conversational – Text-to-Voice AI model built for conversational speech in real-time. [Link]

The world’s first Large Language Text-to-Speech Model that is built for carrying out conversations like a real person in real-time.

This model also introduces for the first time the concept of Emotions to Generative Voice AI, allowing you to control and direct the generation of speech with a particular emotion. The model is available in closed beta and will be made accessible through our API and Studio.

The famed Stanford Smallville is officially open-source! [Link]

The famed Stanford Smallville is officially open-source!

25 AI agents inhabit a digital Westworld, unaware that they are living in a simulation. They go to work, gossip, organize socials, make new friends, and even fall in love. Each has unique personality and backstory.… twitter.com/i/web/status/1…

— Jim Fan (@DrJimFan)
Aug 9, 2023

Try Divvy. Get $100 [Link]

Ever felt trapped in the relentless cycle of expense reports? It’s time to breathe. Introducing Divvy – where your financial management finds its rhythm. It’s not just about tracking and controlling; it’s about freedom from the mundane and regaining precious hours.

Experience Divvy. Let us show you, in just 20 minutes, a world where books close faster than a chapter in a thrilling novel. And for your time? A $100 Amazon gift card awaits.*

Say goodbye to manual expense reports.
Embrace budgets that adapt and enforce.
Discover business credit that scales with your ambition.

Less about the technical, more about the feeling. What do you think?

Llama from scratch (or how to implement a paper without crying) [Link]

My approach for implementing papers is:

Make all of the helper functions required to test your model quantitatively (data splits, training, plotting the loss).
Before you even look at the paper, pick a small, simple, and fast model that you’ve done in the past. Then make a helper function to evaluate the model qualitatively.
Start by picking apart different components of the paper, and then implementing them one-by-one, training and evaluating as you go.

Other stuff

Custom instructions are now available to ChatGPT users on the free plan! [Link] (And so is the Custom instruction profile from Superpower ChatGPT) [A custom instruction for more accurate responses]
AI Botched Their Headshots [Link] HIGHLY RECOMMENDED
Stability AI Is Losing Executives, Engineers, and Its Edge [Link]