Google Bard Enhances Image Generation and Rolls Out Advanced Gemini Pro to Compete with ChatGPT

Google is enhancing its Bard AI chatbot to strengthen its competition against OpenAI’s ChatGPT. Under the guidance of Sundar Pichai, Google announced the addition of image generation capabilities via its own Imagen 2 AI model, along with a more advanced version of Gemini Pro.

These updates give users broader access to Bard's AI functionalities, including a new free tool for creating AI-generated images.

“These updates position Bard as a more efficient and globally accessible AI partner for tasks ranging from large creative projects to everyday activities,” said Jack Krawczyk, product lead for Bard, in a blog post.

In addition, Google is testing another image generator called ImageFX, starting today.

Gemini Pro with Multilingual Support

Over a month ago, Google introduced the Gemini AI model in three versions: Nano for mobile use, Pro for intermediate applications, and Ultra, which is expected to be the most powerful language model ever created—more advanced than GPT-4—though the Ultra version is not set for release until later this year.

Initial comparisons between Gemini Pro and other models indicated that it may lag behind OpenAI’s older GPT-3.5 Turbo. This presents a challenge for Google as it aims to showcase its capabilities in the competitive landscape of generative AI. A fine-tuned version of Gemini Pro was released on Bard last month, but it was only available in English.

Today’s series of new AI features aims to help Google bridge this gap. The latest version of Bard will support over 40 languages—including Korean, Spanish, Tamil, Italian, and Russian—across more than 230 countries and territories. This expansion provides more users with access to Gemini Pro's advanced capabilities in understanding, summarization, reasoning, and coding, alongside Bard’s feature that verifies responses by searching the web.

Imagen 2 on Bard: Competing with ChatGPT Plus and DALL-E 3

Perhaps the most exciting development is the introduction of AI image generation using the Imagen 2 model, which is designed to create high-quality, photorealistic images from text prompts. This positions Bard as a direct competitor to OpenAI’s ChatGPT Plus, which incorporates the DALL-E 3 image generator.

“Simply describe what you want—like ‘create an image of a dog riding a surfboard’—and Bard will generate a variety of visuals to bring your concept to life,” Krawczyk explained.

During testing, Bard produced images in approximately 30-40 seconds, demonstrating good consistency. However, there were instances where it failed to generate images altogether, even when adhering to guidelines that filter out images involving well-known individuals to avoid potential scandals.

Currently, there is no support for altering the aspect ratio or using non-English prompts, based on our initial tests of the tool.

Addressing copyright concerns around AI-generated media, Google Bard allows users to report legal issues related to data protection and copyright for all generated content. The platform also enforces limits on violent, offensive, or sexually explicit content. Furthermore, Google has embedded digitally identifiable watermarks into the pixels of generated images using DeepMind-developed SynthID, helping differentiate AI-generated visuals from those created by human artists.

New Iteration Features with ImageFX

In addition to Bard, Google is exploring ImageFX, which is powered by Imagen 2. Available now in AI Test Kitchen, Google’s experimental app, ImageFX encourages creative exploration through “expressive chips” that provide users with suggestions and adjacent dimensions to enhance their prompts. This feature is similar to offerings found in other creative tools, such as Ideogram.

The AI Test Kitchen also hosts innovative projects like MusicFX, which can create tunes of up to 70 seconds with text prompts and expressive chips, along with TextFX, aimed at lyricists and creative writers.