Midjourney V6.1, OpenAI Voice, Flux & more

Enjoying this newsletter? Give a shoutout to our sponsor ❤️

Journalist AI is an advanced AI-driven platform designed for generating high-quality, SEO-optimized articles swiftly. Users can create, publish, and manage content tailored to their business needs without requiring prior writing expertise. The platform allows users to input specific keywords, topics, or titles, and generates unique articles in under 30 seconds.

Get Started

✨ Midjourney V6.1: Major Upgrades

• Midjourney has officially launched version 6.1 of its AI image generator after a seven-month wait since the initial V6 release, with upgrades driven by community feedback.

• The new version features significant enhancements, including improved coherence in image rendering (especially for limbs and animals), reduced pixel artifacts, and better textures, particularly for skin and retro styles.

• Instead of a typical incremental update, Midjourney is releasing updates in two phases (V6.1 and V6.2) to gather user data and feedback on V6.1 before introducing the next version.

🔊 OpenAI’s Advanced Voice Mode Now Live

• OpenAI's new Advanced Voice Mode for ChatGPT is now rolling out to a limited number of ChatGPT Plus subscribers, allowing for more natural, real-time conversations with emotional and non-verbal cues.

• The feature was originally planned for release last month but was delayed for safety reasons. It is currently in alpha and accessible to only 10,000 users who receive an email invitation.

• All ChatGPT Plus subscribers will gain access to Advanced Voice Mode in the fall. To use it, users open the ChatGPT app and click a tooltip in the bottom-right corner to initiate voice mode.

⚡ Stability AI’s Stable Fast 3D: Assets in 0.5 Seconds

• Stability AI's new Stable Fast 3D (SF3D) can generate textured UV-unwrapped 3D assets from a single image in just 0.5 seconds, significantly faster than previous models that took 50-60 seconds.

• SF3D is specifically trained for mesh generation and utilizes a fast UV unwrapping technique, which allows for quick texture generation without depending on vertex colors. It also predicts material parameters and normal maps to enhance visual quality.

• The system includes a delighting step that effectively removes low-frequency illumination effects, making the reconstructed meshes suitable for various lighting conditions.

🏆 Meet Flux: New AI Image Generator Beats the Best

• Flux is introduced as the largest state-of-the-art open-source text-to-image model, developed by Black Forest Labs, featuring 12 billion parameters and delivering high-quality images akin to those produced by Midjourney.

• The model is available in three variations: FLUX.1 [dev] (open-sourced for community use), FLUX.1 [schnell] (a faster distilled version), and FLUX.1 [pro] (a closed-source API version), each designed to cater to different user needs.

• Key features of Flux include enhanced image quality, improved human anatomy and photorealism, better adherence to prompts, and exceptional processing speed, particularly with the FLUX.1 [schnell] variant, which operates up to ten times faster than the base model.

💡 EarlyAI: Revolutionizing Software Unit Testing

• Early AI, a startup founded by former Google and Anthropic researchers, has developed EarlyAI, a tool that aims to revolutionize software unit testing by automatically generating test cases.

• EarlyAI uses large language models and reinforcement learning to generate test cases that maximize code coverage and catch bugs early in the development process.

• The tool integrates with popular IDEs and CI/CD pipelines, making it easy for developers to adopt and incorporate into their existing workflows.