✨ChatGPT Gets GPT-4o-Powered Image Generation

Today we will discuss: 

• Don’t miss this week’s Workflow Wednesday—AI innovation unlocked with Grammarly, CoWriter, and an AI-powered travel planner using Tally, Zapier & ChatGPT;now available at only $20/month! 

• Images in ChatGPT—OpenAI’s latest GPT-4o-powered feature

• Google unveils Gemini 2.5—Upgraded AI model promises better coding, reasoning, and understanding

• In other AI news today—Microsoft adds ‘deep reasoning’ AI to Copilot, Apple will use Apple Maps imagery to train AI, Amazon tests AI-powered shopping and health assistants, and Waymo plans a 2026 robotaxi launch in Washington, DC

• Saved the best for last—10 must-try AI tools

We just launched the 12th edition of Workflow Wednesday for AI-minded professionals like you—delivering actionable AI workflows straight to your inbox.

This week’s topic: AI Innovation

Supercharge your workflow with Grammarly, CoWriter, and a fully automated AI-powered travel planner using Tally, Zapier & ChatGPT.

Join here

OpenAI rolls out new GPT-4o-powered image generation feature for all users

Key Points 

• The new model improves accuracy, maintaining relationships between 15-20 objects and rendering readable text more reliably.

• Unlike diffusion models, this system builds images step by step, improving text placement and object consistency while taking slightly longer.

🤩News - OpenAI is rolling out a new feature called “Images in ChatGPT,” allowing users to generate images directly within ChatGPT using GPT-4o. This feature is available across all subscription tiers. 

For free-tier users, limits will match those of DALL-E, though OpenAI hasn’t provided exact numbers. Previously, free users could generate up to three images daily with DALL-E 3. As for DALL-E itself, OpenAI says it will remain available through a custom GPT.

🌟What's new & improved? One of the biggest improvements in this new model is its ability to maintain accuracy when handling multiple elements in an image. AI-generated images often mix up attributes—like switching colors or shapes when asked to generate multiple objects. OpenAI says this new model can correctly handle 15 to 20 objects at once, a significant jump from the typical 5 to 8 in other systems.

Text rendering has also improved. AI image generators often struggle with readable text, producing letters that appear distorted or jumbled. OpenAI spent months refining this capability, and while small text can still be tricky, overall readability is much better than before.

🤓What's driving these changes? Unlike DALL-E and most other AI image generators, which use a diffusion model to create images all at once, this system builds images sequentially—left to right, top to bottom, similar to how text is written. OpenAI suggests this method may be why it performs better at text rendering and object accuracy. The tradeoff is slightly longer generation times, but OpenAI believes the quality improvements justify the wait.

🥸Safeguards & transparency - To prevent misuse, OpenAI has included safeguards to block watermark removal, deepfake creation, and CSAM requests. While these images won’t have visible AI watermarks, OpenAI is embedding metadata using the C2PA standard to mark them as AI-generated and will have internal tools to track image origins.

Read the full story here.

Google introduces Gemini 2.5 with major performance upgrades

Key Points 

• Google claims Gemini 2.5 outperforms OpenAI, Anthropic, and others on AI benchmarks for reasoning, coding, and problem-solving.

• Google says Gemini 2.5’s step-by-step reasoning improves accuracy, making it better at handling complex, real-world tasks.

👨‍💻News - Google has introduced Gemini 2.5, an upgraded AI model designed to deliver better reasoning, coding, and problem-solving.

Gemini 2.5 Pro is now available in Google AI Studio and for Gemini Advanced subscribers, who can select it from the app’s model dropdown menu. The company claims its latest model outperforms rivals from OpenAI, Anthropic, xAI, and DeepSeek on key benchmarks, measuring everything from language understanding to mathematics.

🕵️‍♂️What's more? One of Gemini’s biggest strengths, according to Google, is its native multimodality—the ability to work with not just text but also images, video, audio, and code. DeepMind CEO Demis Hassabis called Gemini 2.5 Pro “an awesome state-of-the-art model,” highlighting that it now ranks first on LMArena’s leaderboard with a significant +39 ELO point lead.

The company has also announced that a 2 million-token context window is “coming soon,” allowing the model to process much larger amounts of information at once.

Google says the model’s improved performance comes from a shift toward “reasoning” AI. We’re building these thinking capabilities directly into all of our models,” the company wrote, explaining that this will lead to more capable and context-aware AI agents.

Here's a demo that highlights this in action, showing Gemini 2.5 Pro generating a fully programmed video game from a single prompt.

Read the full story here.