Comparison of AI Image Generator Apps

Understanding the Background

Since Stable Diffusion was released in 2022, many image generation tools have emerged, to meet a variety of needs. This article breaks down the core differences between different image tools.

Professional vs. Recreational use

Professional tools focus on complementing, augmenting, or replacing conventional image editing workflows done by professional photographers. Adobe Firefly is the leading example for static images, introducing AI image tools to Adobe’s Photoshop and Lightroom tools. Google’s Veo3 video generation tool was recently released and made a lot of impact on video generation, targeting professional clients who are already integrated with the Google Workspace.

Recreational tools unlock a new arena of creativity, allowing everyday users to quickly create photorealistic images of any scene, often through a consumer mobile app. While many image generation tools are available through web interfaces or other platforms (e.g. Midjourney started as a Discord server), the rest of this article focuses on mobile apps which are most appealing to most users.

Image Generation Mobile Apps

This section describes key dimensions which differentiate different image generation mobile apps focused on general users.

Types of apps

There are a variety of ways that image generation mobile apps have been developed, targeting different markets:

• Generalist Apps: As an image generation functionality within generalized AI tools like ChatGPT, Google Gemini, Meta AI, or xAI Grok. These generally offer text-to-image with text prompting, but no additional controls.

• Enthusiast-oriented tools: These are focused on providing a high level of control, typically offering a wide variety of models, filters/finetunes, and granular control over the generation process. The developers of these tools pride themselves on exposing many controls and offering high-quality models, but this can create a bewildering experience for new users, and often do not have quick or easy image generation.

• Specialized tools: These are focused on a particular style of image generation, e.g. anime or comic book art.

• Image Editing Workflow tools: These are intended to fit smoothly into another workflow, such as touching up photos or applying generative filters for photos and video for social media.

• Conversational tools: This emerging application allows you to create images at the same pace with which you think, and are intended as a replacement for image or gif search tools.

Fee Structure

Almost all image generations only offer limited free access, and protect further usage in one of several ways:

• ✅ A true freemium tier, with enough free daily images that users can explore the basic functionality: only a few providers (like Gigglebot and Meta AI) offer this.

• 😵‍💫 Ad-based freemium: A limited number of generations after the user watches a lengthy ad (examples: Davinci AI, Wonder AI Art Generator, AI Arta)

• 🪙 Pay-for-credits model: Some apps offer the ability to buy credits without subscribing. This can allow you to try out the tool without commiting to a subscription.

• ❌ Strict subscription paywall with no freemium: Many firms push you toward a recurring subscription with little or no opportunity to test the tool before you sign up for the subscription. While they may offer discounted offers, watch the recurring annual prices, which may run $40-100/year or more!

Many tools also require a user sign-in or account creation to use the basic services,

• ✅ Some apps do not require the user to create an additional account, creating a more seamless experience (e.g. Gigglebot)

• ⚠️ Some apps require an account when signing up for premium services

• ❌ Some tools require creating an account even to use the basic functionality. This is more common for the comprehensive AI tools like ChatGPT and Google Gemini, which offer other functionalities - but some image generators also require the user create an account for all their use!

Image Generation Speed

In text generation models, you can start reading the first words while the AI writes the rest of the paragraph. This isn’t the case in image generation models, which are an “all-or-nothing” generation: you don’t get to see the image until all the calculations are done. This means that slow generation speed can be more frustrating for the user for images than for text-based AI tools.

Sometimes, slower models can produce better-quality images: but just as often, the speed of the model depends on the quality of the engineering, the size of the output, and whether the model has been tuned for speed can all make a huge difference!

Often, users may want to try out different variations on the prompt to guide the model toward a particular image that they are imagining. In applications like this, speed is essential as it allows you to iterate quickly!

This can create a huge range in image generation speed between different models:

• 🐌 Slow: ChatGPT is a common go-to app for image generation, but it is one of the slowest image generation tools, typically taking 45 seconds to generate an image. This pace can be very frustrating if developing the right image takes multiple iterations.

• 🐢 Middle of the pack: Most image generation models will take 5-15 seconds.

• 🏃 Faster: Very few image models are able to produce images in less than five seconds; most of these have had significant engineering work to make them faster, and may sacrifice some quality. This pact can allow for easier iteration, but still may not be fast enough for quickly expressing an idea in conversation.

• 🏎️ Extremely fast: We are only aware of one consumer app (Gigglebot) that is able to reliably produce results in less than 1 second. This speed allows for

Summary Table