AI Image Generation Tool Comparison
The below is a quick comparison of various AI Generation tools. Will continue to expand this as access is gained / releases are made :).
Image Generation[edit | edit source]
General Descriptions[edit | edit source]
MidJourney[edit | edit source]
This tool is built by an independent research lab, and currently has its main interaction via Discord. Strives for more of an "art" vibe, built with community and collaboration first in mind, and appears to be quite open to community involvement. Currently in Open Beta with v3 algorithm. You can learn more at:
- http://midjourney.com/
- Quick Start Guide - Midjourney Documentation (gitbook.io)
- Midjourney - Wikipedia
DALL-E 2[edit | edit source]
This tool was built by the artificial intelligence research and deployment company, OpenAI, currently available in Beta.
CrAIon[edit | edit source]
Formerly DALL-E mini (https://www.craiyon.com/), a free alternative that uses a smaller / less detailed model. It's not listed in the comparison below, but worth trying if you don't need the same detail as DALL-E 2!
Stable Diffusion / Dream Studio[edit | edit source]
An open-source approach to image generation - eventually available for anyone to use, currently available via Dream Studio in preview. You can also run it directly on your computer as Stable Diffusion is Open source!
- Stable Diffusion launch announcement — Stability.Ai
- https://beta.dreamstudio.ai/
- https://stability.ai/blog/stable-diffusion-public-release
- https://github.com/AUTOMATIC1111/stable-diffusion-webui
Cost Comparison[edit | edit source]
Each of the three tools has different cost models, which may change over the course of time - so check with the tools themselves for the latest status!
Summary Comparison[edit | edit source]
Note the below comparison chart is for reference only - please see the individual sections and tools for an understanding of how their pricing works! Since they use different models, the below attempts to compare a few "normal" usage scenarios with various plans / setups. Also note all tools have some type of free / trial tier, which is only included in this comparison if it refills monthly. For example - it's hard to compare fast/slow queue, unlimited vs. pay per image, etc.
Scenario | MJ Basic | MJ Standard | Dall-E 2 | Dream Studio |
---|---|---|---|---|
I only generate ~5 images per month | $10.00 | $30.00 | Free Tier | $2.37 [£2] (single images)
$2.37 (£2) (4x image options) |
I generate ~200 images per month | $10.00 | $30.00 | $24.13 | $2.37 [£2] (single images)
$9.48 [£10] (4x image options) |
I generate ~1000 images per month | $63.34 | $30.00 (allow slower queue)
$36.67 (all fastest) |
$128.49 | $11.83 [£10] (single images)
$47.32 [£40] (4x image options) |
I generate ~5000 images per month | $330.00 | $30.00 (allow slower queue)
$273.33 (all fastest) |
$650.22 | $59.15 [£50] (single images)
$236.60 [£200] (4x image options) |
I generate ~10,000 images per month | $663.33 | $30.00 (allow slower queue)
$606.67 (all fastest) |
$1302.39 | $118.29 [£100] (single images)
$473.16 [£400] (4x image options) |
Above are all total cost / month, prices in USD (prices converted 8/20 when not in USD), and are approximate. Note for Dream Studio I've included both a 4x grid as well as single images, as the other tools all generate a 4x image grid of variations to start. Please check the respective tools for more details!
Image Comparison[edit | edit source]
This is likely the hardest section, as crafting the best prompts for each tool is difficult. Dall-e 2 and Stable Diffusion (as of this writing) seem to understand prompts better, while MidJourney has more tools for image refinement, and MidJourney and Stable Diffusion result in the most "art-like" results by default. A perfectly crafted prompt for each of them likely looks best... in that tool.
That said, the below images are to help understand the "default" results without much prompt crafting involved, to give you an idea of which tool is best suited to what art styles. This comparison also ignores additional features (such as Dall-e 2's infill capability), which may have specific use cases - you likely will use some combination of these tools to help you develop your artwork.
"A beautiful sunrise"[edit | edit source]
A man with dark hair and a goatee, deep in thought, a brush in his hand, surrounded by ornate architecture[edit | edit source]
A woman with pink hair piloting a mech, flying in the sky above a mountainous scene in the style of anime[edit | edit source]
A cluster of clergy in a cathedral, expressionistic painting[edit | edit source]
Video Generation[edit | edit source]
General Descriptions[edit | edit source]
Make-A-Video[edit | edit source]
This tool is built by Meta, and is an upcoming video generation tool: