Create Images From Nothing
Midjourney, Ideogram, and ChatGPT compared -- what each does best and what they cost.
Core reading
These days, AI image generation has reached the point where the results are genuinely useful for social media, presentations, personal projects, and creative work. The quality varies significantly between tools.
The main image generators
| Midjourney | |
|---|---|
| What it is | The highest-quality AI image generator, accessed through Discord or the web |
| Best at | Artistic, photorealistic, and stylised images with exceptional quality and coherence |
| Free tier | No free tier (previously offered a trial, now requires a subscription) |
| First paid tier | Basic Plan - US$10/month (~A$16/month) |
| Ken's take | Midjourney produces the best-looking images of any AI tool I have used. The quality gap between Midjourney and free alternatives is noticeable. The catch is that it used to require Discord, which confused many people. They now have a web app that is much easier to use. The latest model, V8 Alpha (released March 2026), runs about five times faster than V7, supports native 2K output via the --hd parameter, and is markedly better at complex compositions. Worth the US$10 if you create images regularly. |
| Sign up | https://midjourney.com |
| Ideogram | |
|---|---|
| What it is | AI image generator that excels at putting readable text into images |
| Best at | Logos, social media graphics, posters, anything that needs legible text in the image |
| Free tier | Free tier with limited daily generations |
| First paid tier | Basic Plan - US$8/month (~A$13/month) |
| Ken's take | Most AI image tools are terrible at text. Letters come out garbled and unreadable. Ideogram solves this. If you need a poster, a social graphic with a quote, or a logo concept, this is the tool. The free tier is enough to test it. |
| Sign up | https://ideogram.ai |
| ChatGPT Image Generation (GPT Image 1.5) | |
|---|---|
| What it is | Image generation built directly into ChatGPT. The current model is GPT Image 1.5, OpenAI's anointed successor to DALL-E (DALL-E 3 was retired from ChatGPT in late 2025 and is being deprecated from the API on 12 May 2026). |
| Best at | Quick image generation without leaving your chat, iterative editing through conversation, text rendering inside images (which Midjourney still struggles with), and generating images as part of a broader task. Currently sits at the top of LM Arena's image-generation leaderboard. |
| Free tier | Included in ChatGPT free tier (limited generations) |
| First paid tier | Included with ChatGPT Plus - US$20/month (~A$32/month) |
| Ken's take | If you already pay for ChatGPT Plus, you have this included. No need for a separate subscription. The quality has caught up to a point where, for most non-artistic uses, you no longer need Midjourney as well. Where Midjourney still wins is on stylised and editorial work; where GPT Image 1.5 wins is text inside the image and conversational iteration. |
| Sign up | https://chat.openai.com |
Brief mention: Adobe Firefly is worth knowing about if you already use Photoshop or Illustrator. It integrates AI generation directly into those tools. Free tier gives 25 credits/month. Standalone plans start at around US$5/month. The image quality is behind Midjourney, but the commercial safety (trained on licensed content) is a selling point for business use.
Other image tools worth knowing about
Beyond the three above, four more deserve a mention. They sit in different niches: open-source for full control, fast-and-cheap via developer APIs, all-in-one creative suites, and game-art specialists. None replace Midjourney for headline image quality, but each is the right pick for a particular kind of work.
| Leonardo AI | |
|---|---|
| What it is | An AI image generator with a strong free tier and a focus on game art, fantasy illustration and stylised work. Owned by Canva (acquired 2024) but operates as a separate product |
| Best at | Stylised and game-art generation. Fine-tuning on your own characters or assets. The free tier of 150 daily tokens is the most generous in the category |
| Free tier | Free tier with 150 daily tokens (resets each day) |
| First paid tier | Apprentice - US$10/month (~A$16/month) |
| Ken's take | If you make game art, fantasy illustration or anything where you want to lock in a particular character or style across many images, Leonardo is the right tool. The fine-tuning workflow is friendlier than Stable Diffusion. The free tier is genuinely usable, with daily refresh meaning you can experiment without committing. For straight headline-quality realism, Midjourney still wins. For ongoing creative work with consistent characters, Leonardo earns its place. |
| Sign up | https://leonardo.ai |
| Krea AI | |
|---|---|
| What it is | A creative suite covering image generation, video and 3D work. Real-time generation means you see changes as you adjust the prompt |
| Best at | Iterative creative work. The Plus plan ($25 a month) gives unlimited image generation across multiple premium models, which is unusual at this price |
| Free tier | Free tier with limited daily generations |
| First paid tier | Basic - US$9/month (~A$14/month). Plus tier at US$25/month gets unlimited image generation |
| Ken's take | Real-time generation is the headline feature. You see the image change as you type, which makes iteration fast in a way the others are not. Krea bundles image, video and 3D in one tool, so if you do all three, the Plus plan ($25/month with unlimited image gen) is good value. Brand recognition lags Midjourney and Adobe Firefly. Quality is good but specialists beat it on specific tasks. Worth a try on the free tier first. |
| Sign up | https://krea.ai |
| Stable Diffusion | |
|---|---|
| What it is | Open-source image models from Stability AI that you can run on your own computer or via hosted services. Stable Diffusion 3.5 Large is the current flagship |
| Best at | Full creative control with no vendor lock-in. The massive ecosystem of community LoRAs (small fine-tunes) and extensions means you can adapt the model to almost any style |
| Free tier | Free to download and run on your own GPU (open weights) |
| First paid tier | Pay-per-use via hosted providers like Replicate or fal.ai (typically a few cents per image), or Stability AI's own API |
| Ken's take | Stable Diffusion is the open-source option. If you have a capable GPU and do not mind a bit of setup, you can run it locally and generate as many images as you want for free, no subscriptions, no content filters, no internet connection required. The community ecosystem is enormous. The trade-off is image quality below Midjourney V8 and Flux on most benchmarks, plus a steeper learning curve. Not for casual users. The right pick for power users, hobbyists, and anyone with privacy concerns about cloud generation. |
| Sign up | https://stability.ai |
| Flux | |
|---|---|
| What it is | Image models from Black Forest Labs (the original Stable Diffusion team). Mainly used by developers via API rather than a consumer web app |
| Best at | Image-to-image work and very cheap generation. The Schnell variant is fast and costs around $0.003 per image. Pro is competitive with Midjourney on quality. The Krea-tuned variants lead image-to-image work |
| Free tier | None directly. Free tiers exist on third-party providers |
| First paid tier | Pay-per-image. Schnell from US$0.003/image, Dev and Pro from US$0.04-0.05/image |
| Ken's take | Flux is mostly used inside other tools rather than directly. If you use Krea, fal.ai, Replicate or any of the open AI image platforms, you have probably already used Flux without knowing. Worth knowing about as a name because it shows up in vendor specs and pricing pages. For end users with no developer interest, just use Krea or Midjourney. The quality varies widely between Schnell, Dev and Pro tiers, so check which one a host is actually using. |
| Sign up | https://bfl.ai (developer-focused) or via Krea |
Try this right now (free)
Open ChatGPT (even the free version) and type: "Create an image of a cosy Australian country kitchen with morning light coming through the window, a cup of coffee on a wooden table, and a cat sleeping on a chair." See what comes back. Then try: "Make the light warmer and add a bookshelf in the background."
See also: AI-generated images pair well with Presentations and Video. For turning ideas into full slide decks with images, the presentations page shows how.