The Google Play Store's malware scanner could add this useful new button Summary Meta AI offers ChatGPT-like capabilities within Meta apps for easy access without leaving the platform. New features include image editing and personalized image generation, with more countries gaining access. The AI can create images based on simple prompts, but refinement may require repeating the process with additional input.
AI has been the buzzword of the consumer tech space for a couple of years now and every major tech company has a finger in the pie. Meta is no different, building the rather unimaginatively-named Meta AI into a Google Gemini-like chatbot with typical capabilities such as immediate search results, contextual replies, and image generation, but all within the confines of Meta's social media platforms, including the world's most widely used chat app, WhatsApp. Well, the company just announced that Meta AI is now multilingual, smarter and better at image generation.
Related Meta AI: What is it, who can use it, and how?
Learn about Meta AI and how you can use this free, open source artificial intelligence to chat, create images, and more
Meta AI is designed to offer ChatGPT and Gemini-rivaling capabilities directly within Meta apps, so users don't need to exit the social media platforms to use AI for an answer or other assistance. In WhatsApp, users can simply append @MetaAI before their query or prompt, and the AI will respond in the chat, like a participant tagged in a conversation. You can also hop into a chat with Meta AI on WhatsApp, like you would in the ChatGPT app, for more detailed conversations where you could benefit from longer context windows.
Now, WhatsApp just shared that Meta AI is receiving an update, unlocking new features like image editing and AI-powered restyled shots of yourself, and bringing the service to a few more regions worldwide.
Features and geographic availability continue to expand Source: Meta
Currently, image creation works exactly like you would expect it to -- You provide a simple prompt and Meta AI spits out one result -- but unlike some purpose-built image generators that offer choices. However, you'll need to run the entire prompt again with additional input for any sort of refinement. Now, you can just start a prompt with "imagine..." in the chat with Meta AI and go on to describe the image you want. You can then ask the AI to add, remove, or animate elements of the generated image by replying to it.
In my limited testing, replying with a request for one edit made the AI generate a new image altogether, even when I used specific keywords WhatsApp used in the announcement, such as "add, subtract" etc. It retained context, and you'll see in the screenshots below that both images are of happy dogs at weddings, but the AI failed to edit the original image. WhatsApp says the service is available in English only, for now.
Close
Meta AI cannot refine images perfectly
The other new feature is called Imagine Yourself, which we caught wind of in July. Here, you need to first type "Imagine me" in the chat with Meta AI, go through a quick onboarding of sorts to help the uAI understand what you look like, so you can then follow up with creative prompts like "imagine me in Aviator shades and a white hat" to see what AI thinks you will look like in those settings.
Source: Meta
For now, WhatsApp has rolled out Imagine Yourself as a beta test in the US only. Attempts to imagine me in other countries created random images of people. Both the above-mentioned features are also available on other Meta properties such as Messenger and Facebook. Speaking of countries, though, Meta AI is now available in 22 countries in toto, now including Argentina, Chile, Colombia, Ecuador, Mexico, Peru and Cameroon. You can also interact in a language you prefer, because the supported ones are French, German, Hindi, Hindi-Romanized Script, Italian, Portuguese, and Spanish.
While you should temper your expectations of early-stage text-to-image models, Meta AI might just make it easier to decide your next outfit or convert imagination into pictures with minimal effort.
Thanks: Moshe
