- Everyday AI
- Posts
- Ep 766: ChatGPT Images 2: How Even Non-Creatives Can Unlock Growth With Images 2
Ep 766: ChatGPT Images 2: How Even Non-Creatives Can Unlock Growth With Images 2
OpenAI is shifting Codex to usage-based pricing, NVIDIA released a unified multimodal model, and Gemini finally gets in-app creation.
Everyday AI
April 29, 2026
👉 Subscribe Here | 🗣 Hire Us To Speak | 🤝 Partner with Us | 🤖 Grow with GenAI
Outsmart The Future
Today in Everyday AI
8 minute read
🎙 Daily Podcast Episode: OpenAI just dropped ChatGPT Images 2, generating production-ready visuals with accurate text and real-world context from a single prompt. Give today’s show a watch/read/listen to learn more.
🕵️♂️ Fresh Finds: DeepSeek rolled out image understanding, Musk vs. OpenAI trial gets testy, and Trump posts AI photo of himself wielding a gun and more. Read on for Fresh Finds.
🗞 Byte Sized Daily AI News: OpenAI is shifting Codex to usage-based pricing, NVIDIA released a unified multimodal model, and Gemini finally gets in-app creation. Read on for Byte Sized News.
💪 Leverage AI: OpenAI’s Images 2 models is beyond bonkers in terms of capabilities. But the new use cases unlocked in this update are just as impressive. Keep reading for that!
↩️ Don’t miss out: Miss our last newsletter? We covered: OpenAI reportedly missing growth targets is shaking AI stocks, Google just signed a classified AI deal with the Pentagon, and Microsoft is turning Outlook into an autonomous inbox manager, and more. Check it here!
Ep 766: ChatGPT Images 2: How Even Non-Creatives Can Unlock Growth With Images 2
"Not a creative"? 🧑🎨
That's no longer an excuse.
With ChatGPT's new Images 2 model, the gap between those that creatively can and those that can't has shrunk to basically nothing.
From creating striking editorial images without a camera or subject, to consultant-worthy slide decks and designer-inspired infographics, you all of a sudden have a creative superpower at your side in Images 2.
But how do you use it? And what are the tips to unlock its capabilities?
Also on the pod today:
• ChatGPT Images 2 “thinking mode” 🧠
• Live web search for visuals 🌐
• Flawless text rendering in images 🔤
It’ll be worth your 39 minutes:
Listen on our site:
Click to listen
Subscribe and listen on your favorite podcast platform
Listen on:
Here’s our favorite AI finds from across the web:
New AI Tool Spotlight – Plurai introduces vibe-training to build real-time, tailored evals and guardrails for your agent, Netlify Database is a fully managed Postgres database built into the Netlify platform, Karma AI is One brain. Every AI resource. Infinite scale.
Deepmind and Korea — Google DeepMind is partnering with South Korea’s MSIT to open an AI Campus in Seoul and bring frontier AI models to accelerate research in life sciences, climate and energy.
DeepSeek Vision — DeepSeek launched DeepSeek Vision beta, a new image-understanding mode for DeepSeek Chat.
Sigma AI Hub — Sigma now runs private OpenClaw agents locally in the browser, so you can control tabs and browse with Gemma 4, Qwen 3.5, or Nemotron 3 without sending data to the cloud.
NotebookLM Mind Map — Google’s NotebookLM may soon let you scope mind maps to a topic or subset of sources, so one notebook can host multiple focused diagrams.
AI Creativity — Base44 launched "Move to Base44," a one-click migration that imports apps, data, and workflows from Salesforce, HubSpot, Shopify, WordPress, Lovable, and Bolt
Trump AI — Trump posted an AI-generated image of himself holding a gun and warned Iran "no more Mr. nice guy," sparking new controversy and questions about misinformation and online escalation.
Raspberry Pi AI — Google DeepMind and Raspberry Pi’s Experience AI has trained 30,000+ teachers and reached 2.9 million students with free AI literacy resources.
PlurAI Vibe Training Platform — Plurai turns expensive, brittle LLM judges into cheap, fast, purpose-built SLM evaluators that run continuously for production-grade testing.
Musk vs. OpenAI — In court, things between Elon Musk and OpenAI’s lawyers got a bit testy.
Snapchat AI — Snapchat is testing AI Sponsored Snaps that put brand chatbots directly inside conversations, aiming to turn everyday chats into personalized ads.
SenseNova U1 Release — SenseTime’s new SenseNova U1 tosses out separate vision encoders and models images and text together in one unified token space, letting it generate interleaved text and images in a single pass.
Poolside Models — Poolside just dropped two agentic coding models: Laguna M.1 (225B, 23B active) and open-weight Laguna XS.2 (33B, 3B active)
1. OpenAI adds pay-as-you-go Codex seats to teams 💸
OpenAI today rolled out pay-as-you-go Codex seats for ChatGPT Business and Enterprise teams, letting organizations buy Codex access billed by token use instead of fixed seat fees, a move that aims to simplify pilots and scale-ups.
The company also cut ChatGPT Business annual pricing from $25 to $20 per seat and is offering limited-time credits — $100 per new Codex-only user, up to $500 per team — to encourage adoption. OpenAI says Codex usage is surging, with millions of builders and rapid growth inside business plans, framing the change as a response to rising demand and a way to make billing more predictable.
2. NVIDIA unveils Nemotron 3 Nano Omni, a unified multimodal AI for faster agentic workflows 🤖
Released today, NVIDIA announced Nemotron 3 Nano Omni, an open multimodal model that consolidates vision, audio and language into a single system to cut latency and preserve cross-modal context for real-time agents.
By combining perception encoders within a 30B hybrid mixture-of-experts architecture, NVIDIA claims large efficiency gains and higher throughput versus other open omni models while keeping accuracy for tasks like document intelligence and video-audio reasoning.
3. ElevenLabs launches ElevenMusic to let creators discover, remix, and earn 🎼
ElevenLabs today unveiled ElevenMusic, a new platform that pairs AI-driven music tools with a catalog of over 4,000 independent and emerging artists, aiming to shift discovery into active creation.
Users can remix tracks, generate new songs from lyrics or moods using the ElevenLabs music model, publish work on the platform, and earn revenue under a creator payout system modeled after ElevenLabs’ $11M+ payouts from its voice library. The launch also drops Eleven Album Vol. 2, giving early listeners exclusive new tracks while signaling the company’s push to expand its creator economy into music.
4. Gemini adds one-click export to Docs, Sheets, Slides and more 🗃️
Google’s Gemini app now lets users generate and directly download or export files like PDFs, Word, Excel, Google Workspace documents and common text formats straight from chat, rolling content creation and file export into one step.
The update, available globally, eliminates manual copying and reformatting by converting prompts into ready-to-share files such as .docx, .xlsx, .pdf, .csv, LaTeX, RTF and Markdown. This speeds the workflow from brainstorming to finished deliverables and makes it easier to move work between apps without losing formatting
5. OpenAI unveils5-point action plan to put AI defenses in more hands 🛡️
OpenAI today released a five-pillar Action Plan aimed at democratizing AI-powered cyber defense, stressing urgency as AI expands both defensive and offensive capabilities.
The plan calls for broader access to defensive tools, closer government-industry coordination, tighter safeguards around advanced cyber capabilities, and measures to keep deployment transparent and controllable while helping users protect themselves.
6. Mistral launches Medium 3.5 and cloud coding agents 🚀
Today’s announcement introduces Mistral Medium 3.5 as the new default for Le Chat and Vibe, plus cloud-run "Vibe" coding agents and a Work mode for multi-step tasks, marking a clear shift from laptop-bound assistants to persistent, parallel cloud agents.
The 128B dense model offers a 256k context window, open weights under a modified MIT license, and configurable reasoning to handle quick replies or long agentic runs while being small enough to self-host on as few as four GPUs.
Your creative team bottleneck just became optional. Most leaders don't know it yet.
Why?
OpenAI shipped ChatGPT Images 2.0, and it doesn't just generate visuals. It thinks first, plans composition, and searches the web for live data before rendering boardroom-ready output from a plain-English brief. You can generate up to eight cohesive images from a single prompt, with perfect text and character consistency across all of them.
You've prolly thought this yourself: "I'm not a creative person."
That sentence just stopped working. Biiiig missed opportunity if you keep believing it.
This might be the most practically significant visual AI release for non-creative leaders in years. Your competitors who figure this out first are going to move faster on everything from pitch decks to product specs.
What does it ACTUALLY unlock? The ability to go from written strategy to polished executive visual in under two minutes, no design team required.
Don't worry. We're breaking it all down.
On today's Everyday AI, we dig into the thinking mode that changes who gets to create, how live data grounding unlocks visuals at scale, and the 10 business use cases your team can run this week.
Let's get after it.
1. Brief It Like a Boss, Not a Prompt 🔥
Most AI image tools punished non-creatives. You had to describe bokeh, focal length, and depth of field to get anything useful back.
ChatGPT Images 2.0 flipped the whole thing.
What actually changed?
The thinking mode plans your image the way a senior designer reads a full brief before opening any file. Describe the goal, the audience, the tone. The model figures out the visual composition on its own.
This unlocks a lot.
A product manager can describe a new feature and get back a UI spec. A marketing director can describe a campaign theme and get polished creative. Neither one needs to know what "negative space" means.
That shift is going to change what non-creative leaders can ship, and how fast.
Try This
Open ChatGPT, hit the image icon, and write your request like a memo to a new hire. Tell it the audience, the goal, the feel. Don't describe what it should look like. Describe what it should accomplish. Give it real context, real stakes. Then iterate in plain language until it lands.
2. Ground Every Visual in Live Data ⚡
ChatGPT Images 2.0 searches the web before it renders. No other AI image tool on the market does this quite like it.
Ask for a competitor landscape infographic grounded in today's data, or a localized ad creative pulled from current market stats.
It doesn't hallucinate the numbers. It looks them up, then renders them into a polished, publication-ready image.
What's the business play?
Your marketing team can localize a master creative across any language or market without rebuilding assets. Upload your existing creative. Tell it to translate and adapt. DONE.
That collapses a workflow that used to require a designer, a translator, and two rounds of back-and-forth.
On the LM Arena image leaderboard, Images 2.0 set the biggest lead in platform history. It won 93% of blind comparisons.
That ain't a close race.
Try This
Pick one market you've never localized for because the lift felt too heavy. Upload your best-performing creative into ChatGPT Images 2.0 and ask it to translate and adapt for that audience. Ask it to pull live market data to back the visual up. Time it. The whole workflow should take under 10 minutes. Do it again next week.
3. Business Use Cases to Run This Week 🚀
Non-creative leaders always ask the same thing after a big AI launch: "But what do I actually DO with it?"
Product managers can describe a new feature and get a UI spec to hand straight to Codex for working front-end code. No designer in the loop.
Marketing teams can localize master creatives globally in minutes. No rebuilding assets, no translation vendor.
Sales can personalize one-pagers across demographics and regions at scale. Leaders prepping boardroom presentations can feed in a written strategy and get an executive visual back in under a minute.
Those are four of the 10 from today's episode.
Anyone on a free ChatGPT plan can run at least two of them today.
Try This
Scan that list and pick the one that would cost your team the most time to produce manually. Run it as a test today. Give the model your real context and goal. Iterate in plain language if the first output needs work. Then show the result to your team. One demo like this can do more for internal AI adoption than any training session you've run.
Reply