Verification Post - It’s so over 🤯

A few hours ago, a Reddit user posted this picture as a proof of concept for creating verification selfies using StableDiffusion. The original post(and the second) has been removed ever since, but his 3rd post is still up and it includes a workflow to recreate similar images.

Workflow for consistent faces:

• I combined 2 loras. (e.g. <lora1:x><lora2:y><lora3:z> x+y+z=1 experiment till consistency and aesthetic) Others have pointed out face swapper is quicker. I haven’t tried.

• Generated a batch without Loras and selected an image I liked (prompt ethnicity, hairstyle, and length for consistency). Recycled seed and ADetailered face with loras and controller.

• Workflow for text on card: SD will generate blank paper. Physically write what you want on a paper, the more textured, the better. Layer as hard light in Photoshop and clean up.

• Workflow for text on skin: Write using a brush in Photoshop. Warp to match body contour. Run img2img inpaint with paint on the skin in prompt, high denoising, with Controlnet Canny. May have to crop to 512x512 and enlarge the text area before img2img for good results. Again layer the image in Photoshop.

Get Rabbit’s First AI Device – r1 –Your Pocket Companion

On January 9th, prepare to witness a revolutionary shift in human-machine communication with the unveiling of the Rabbit r1:

• Technology that adapts to you for a seamless digital experience.

• Speak naturally, and r1 will understand and act for you.

• Rabbit’s AI and Large Action Model (LAM) revolutionizes your daily digital interactions.

The r1 is not just another gadget; it's a window into a futuristic way of living.

Sign up for the waitlist now for the January 9th launch.

AI-improved Siri to launch at WWDC 2024, claims leaker

A leaker with a mixed track record, yeux1122, suggests that Apple will announce new generative AI features for Siri at this year's WWDC. The update is expected to enhance Siri with more personalized and natural conversations, better device management across various Apple products, and new Apple-specific creation services. It may also include integration with various external services and distinct features for subscription services.

Bloomberg reported last year that Apple has been developing its own ChatGPT-like AI, dubbed Apple GPT, since 2022. This project is a collaborative effort between Apple's AI, software engineering, and cloud services teams. While there's no concrete plan yet, insiders believe Apple aims to make a significant AI-related announcement in 2024, with WWDC being a likely venue.

The possibility of subscription-based services is also hinted at, suggesting Apple's interest in generating revenue through generative AI.

9to5Mac comments on yeux1122's track record, noting their mixed accuracy in the past. The article suggests that it's likely Apple will introduce significant generative AI improvements to Siri this year, considering the industry trends and Apple's capabilities. The article also speculates that Apple might offer both free and subscription versions of its AI services, similar to ChatGPT, to cover the high computational costs of such services.

Google is preparing a paid version of Bard

A user on Twitter, Dylan Roussel, has shared some insights on upcoming features for Google's Bard, based on his first look at some unreleased features. Here are the key points he mentioned:

1. Motoko: A codename for a new feature related to creating 'Bots'. Details are limited, but it suggests a function for users to create their own bots within Bard.

2. Bard Advanced with Google One: Google plans to offer 3 months of Bard Advanced for free through Google One. This advanced version will utilize Gemini Ultra.

3. Gallery: This feature, still in development, aims to help users explore different topics and see potential uses of Bard. It will likely showcase example prompts on various subjects.

4. Tasks: This appears to be a feature for managing long-running tasks, possibly including image generation, but specific details are scant.

5. Sharing Background/Foreground: While details are unclear, it involves visual aspects of sharing, with Roussel's screenshots only showing the "Foreground" part.

6. Power up: A feature in the works, intended to enhance user prompts by running them through an AI to expand or improve them. No working examples are available yet.

Roussel notes that these features are unreleased and subject to change, and there's a possibility they might not be implemented at all.

Other stuff

• Generative AI startup Rabbit to release revolutionary pocket AI assistant on Jan 9

• Using chatGPT in research and academic applications

• 26 principles to improve the quality of LLM responses by 50%

• OpenAI’s news publisher deals reportedly top out at $5 million a year

• A simulation of me: fine-tuning an LLM on 240k text messages

• GPT Builder: A GPT to help you build GPTs.

• From Audio to Photoreal Embodiment: Avatars that are only animated based on your voice

• HackerGPT (aka whiterabbitneo) is a LLaMA 2 model fine-tuned to be a cybersecurity expert.

• This AI avatar was created using 30 seconds of video plus 30 seconds of audio 🤯

ChatGPT Tips and Tricks - Part 3: Timestamps and counters

Hope you enjoyed today's newsletter