AI image features: PixelVault vs Cloudinary vs ImageKit
There are three jobs AI can do in an image pipeline: create an image from a prompt, edit an existing one, and understand it by tagging or moderating it. Cloudinary and ImageKit cover all three. PixelVault does the first, plus background removal, and nothing else. Here is what each offers and what it costs.
Full disclosure: PixelVault publishes this page. Cloudinary figures were read from Cloudinary's own pages on 30 October 2026 and ImageKit's from ImageKit's on 30 October 2026. AI features and prices change often, so check the vendor before you decide. Sources are listed at the bottom.
At a glance
What one generated image costs
The three bill AI differently, so a fair comparison starts with what each plan includes. This is the plan price spread over its allowance, which is not the same as the marginal cost of one more image, and the quality of the models is not the same either.
PixelVault
Free: 10 FLUX images a month. Starter ($9): 300. Pro ($29): 1,500
About $0.03 per FLUX image on Starter and $0.02 on Pro. A Leonardo image costs 20 credits, about $0.60 of Starter
ImageKit
Free and Lite: 650 units, about 10 generated images. Pro ($89): 4,000 units, about 61
About $0.90 per image on Lite if you spend the whole allowance. Past Pro's allowance, units cost $9 per 5,000, about $0.12 per image
Cloudinary
Generation add-on: a free monthly allotment, then a separate paid generation plan
No published per-image price. The docs say roughly 1 credit is about a cent, varying by model and output size
Two things make these numbers softer than they look. ImageKit's units and Cloudinary's credits are shared with the rest of each product, so spending them on generation uses them up for background removal, transformations or bandwidth. And ImageKit's units buy different things at different rates: a cheap background removal is 10 units, a premium one 130.
Where PixelVault is narrower
- No generative editing. No generative fill, remove, replace, recolor or background replace.
- No upscaling or retouching as a built-in effect.
- No auto-tagging or AI analysis. Nothing reads an image and returns tags, descriptions or moderation results.
- No frontier models. Generation uses open Cloudflare Workers AI models. They are good at illustrations, backgrounds and abstract art, and you should test your own prompt before relying on them for text inside an image or photoreal faces.
- No media-library UI or video pipeline, which both rivals have.
Where it is different
- One meter. A generated image is a credit count you can read off the response, with no transformation counts or unit table to work out.
- Included in every plan. Generation is part of Free, Starter and Pro, with no separate add-on to register for.
- Built for agents. The same call is an API endpoint, a CLI command, an SDK method and an MCP tool, and each returns a hosted URL. See how it works, with examples.
Which to choose
- Cloudinary if you need frontier models, generative editing, or a full asset manager and video pipeline, and can budget a credit pool.
- ImageKit if you want AI edits such as background removal or fill applied by URL inside an image CDN you already use, and the unit model suits you.
- PixelVault if scripts or agents need to create illustrations, backgrounds and hero images and get a hosted URL back, at a cost you can predict.
For plain hosting and transformation pricing, see Cloudinary pricing explained, ImageKit pricing explained and our own plans.
Questions
Does Cloudinary generate images from a text prompt?
Yes. Its Image Generation API has text-to-image and image-to-image endpoints and saves results as managed assets you can transform and deliver. It needs a separate add-on, with a free monthly allotment and then a separate paid generation plan, priced in credits of roughly a cent each that vary by model and output size.
Does ImageKit generate images from a text prompt?
Yes. It is a URL transformation (ik-genimg) and a feature in its media library. Each generated image costs 65 extension units. The Free plan includes 650 units a month, about ten images, and ImageKit labels most of its AI transformations beta.
How does Cloudinary's auto-tagging work?
It is done by add-ons you register in the console: Amazon Rekognition, Google and Imagga. On an upload or update call you set categorization (for example aws_rek_tagging) and, optionally, auto_tagging to a minimum confidence such as 0.7. Cloudinary stores the detected categories with confidence scores on the asset and, when you set a threshold, adds the ones above it as ordinary tags. The Rekognition add-on needs images of at least 80 pixels on each side. Cloudinary's newer AI Vision add-on, in public beta, lets you describe your own tags and moderation questions in natural language through an analyze API. We did not find published prices for the tagging add-ons.
Can PixelVault edit images with generative AI or tag them automatically?
No. PixelVault generates images from a prompt and removes backgrounds with a URL parameter. It does not do generative fill, remove or replace, upscaling, or automatic tagging and analysis. If you need those, Cloudinary or ImageKit are the better fit.
Sources
Read on 30 October 2026: Cloudinary pricing, Image Generation, generative AI transformations, transformation counts, Rekognition auto-tagging, AI Vision and its MCP server.
Read on 30 October 2026: ImageKit plans, how pricing works, AI transformations and its MCP server.
Corrections are welcome: hello@pixelvault.dev.