cloudinaryapipricing

Most Cloudinary alternatives for video are hosts, not processors

··10 min read
Most Cloudinary alternatives for video are hosts, not processors

Most Cloudinary alternatives for video are hosting platforms, not processors. Gumlet, ImageKit, FastPix and Hyperserve each accept your file, transform it, and serve it back to viewers, but what you buy is ingest plus storage plus delivery, so the per-GB and per-minute meters keep running for months after the transform finished. If you already have S3 and a CDN and you need one resize, one watermark or one burned-in caption track, none of the four is answering the question you asked.

That mismatch is baked into the lists. Search for a replacement and you get directory grids and vendor-written roundups that score every candidate on storage, delivery and digital-asset-management features. G2 files Cloudinary's competitors under asset management and video hosting. The category that would hold a processing API doesn't exist in the taxonomy, so it never appears.

The test that sorts hosts from processors

One test separates the two kinds of product: submit a file, get the processed file back, stop paying. Every name on page one passes the first two steps and fails the third, because the asset now lives in their storage and gets served from their CDN, and both are metered.

Here's the same job priced as each kind of product. You have a 10 minute 1080p master in your own bucket. You want a 9:16 crop at 1080x1920 for Reels. You will store the output yourself and serve it from the CDN you already pay for.

What you're buyingMeters that startMeters that stop when the job ends
Video host (Gumlet, Hyperserve)Storage minutes or per-GB storage, monthly streaming bandwidthNone while the asset exists
Media CDN with video transforms (ImageKit)Processing units on first transform, then bandwidth on every viewProcessing units only
Upload-to-playback platform (FastPix)Encoding minutes, storage minutes per month, delivery minutes streamedEncoding only
Processing API (FFmpeg Micro)Input seconds of the jobAll of them

The right-hand column is the whole argument. On a host, finishing the transform does not end the bill, it starts the recurring part of it.

What each of the four recommendations actually sells

Gumlet is the pick every editorial list hands to "video-first" readers, and it's a video hosting product: storage minutes plus monthly streaming bandwidth, with adaptive HLS and DASH and Widevine or FairPlay DRM on top. Two details from Gumlet's own documentation matter if you planned to use it as a processor. The original upload is retained in their storage unless you pass keep_original=false when you create the asset, and a plain MP4 output exists only if MP4 is enabled in the processing profile. Those are Gumlet's descriptions of its own product, not an independent source, but they're the two settings that decide whether you rent a second home for a file you already store.

The same list also repeats that Gumlet ships affordable enterprise DRM "out of the box." Gumlet's own material does not read that way: DRM is free for a handful of videos and then a monthly add-on with a per-view rate above its included allowance. If DRM is why you're shortlisting a host, price that line instead of trusting a roundup.

ImageKit is the usual developer-grade like-for-like, and its video work is metered in processing units charged on the first run of a given transformation, after which every repeat delivery bills bandwidth. Capterra's pricing listing puts the free plan at 25 GB of delivery bandwidth, 5 GB of storage and 500 video processing units, with video units around $9 per 10,000 after the included allowance. That structure means four separate meters: bandwidth, video units, extension units and seats. The transform is the cheapest of them. ImageKit's adaptive streaming output is an .m3u8 or .mpd manifest generated from a URL parameter on first request. That is a delivery feature. There is no returned file to put in your bucket.

FastPix frames itself as the unified upload-to-playback pipeline, with encoding, storage and delivery each metered per minute. G2's pricing page lists the free entry as $25 in credits, roughly 800 encoding minutes or 26,000 delivery minutes, with the trial framed as 10 videos plus 100,000 streaming minutes. Getting an MP4 out means enabling a static rendition and fetching it from their playback domain, which is the same shape as Mux: the file comes back through the player's front door, not as a job output. We walked that exact mechanic and its cost in getting an MP4 back out of Mux.

Hyperserve argues per-GB video pricing against Cloudinary's credit pool, usually with a worked month in the teens of dollars against a $249 plan. The arithmetic is fair and the conclusion is still a hosting bill: that worked month prices storage and bandwidth. A reader who needs one transform on a file already in S3 is being quoted for a second copy plus traffic.

Even the way this category gets benchmarked gives the game away. A developer walkthrough on dev.to measures time-to-upload and time-to-playable across five providers, including FastPix, Mux, api.video, Cloudinary and Gumlet. Time-to-playable is a meaningful number for a host. For a processor it's the wrong stopwatch, because nothing is ever played.

Why the delivery meter dominates the bill

Delivery, not encoding, is where video money goes. One cost breakdown of three-cloud video infrastructure puts CDN egress at 50% to 70% of a streaming bill against 10% to 20% for encode (Spendark). Every pricing page in this category is built on that ratio: the encode is the hook and the egress is the business.

If you already run a CDN, buying a platform whose economics assume you don't means paying twice for the same gigabytes. The honest comparison here is not the headline price, it's which meters you're already covering. We did the same read on Cloudinary's pooled currency in what a credit actually buys for video.

Five questions to run against your own setup

Five questions about your own pipeline narrow the field before you shortlist. Each one eliminates a class of product, not a brand.

  1. Does anything play this video from the vendor's domain? If no viewer ever streams it from them, every delivery meter on the price sheet is overhead you'll pay anyway.
  2. Where does the output need to land? A bucket path means you need a job output. A playback URL means you need a host.
  3. Do you need adaptive bitrate, a player, or DRM? If yes, a host is genuinely the right answer and a processing API is not.
  4. How long does the asset live with the vendor after the job? Storage minutes and per-GB storage bill monthly for as long as the file sits there, whether or not anyone touches it.
  5. Can your automation tool reach it? If the video step lives in n8n, Make or Zapier, a URL-transform CDN was never callable from there.

Answer those and the shortlist usually collapses to one of two products, not ten.

The processing-only version of the same job

A processing API takes the file, runs the transform, hands back an output URL and stops. The FFmpeg command you'd run yourself for that 9:16 crop:

ffmpeg -i master.mp4 -vf "crop=ih*9/16:ih,scale=1080:1920" \
  -c:v libx264 -crf 23 -c:a aac out.mp4

Same job as one call, with your own bucket on both ends:

curl -X POST https://api.ffmpeg-micro.com/v1/transcodes \
  -H "Authorization: Bearer $FFM_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "inputs": [{"url": "gs://your-bucket/master.mp4"}],
    "outputFormat": "mp4",
    "preset": {"quality": "medium", "resolution": "1080p"}
  }'

You get a job id, poll GET /v1/transcodes/{id} until completed, then call /download for a signed URL. Status checks, job listings and downloads are free. Billing runs on the duration of the video processed, per second, with no rounding up to the minute, and failed jobs are never billed. That 10 minute master is 60 tokens. A 90 second UGC clip is 9 tokens, and a $5 top-up of 500 tokens covers about 55 of them. There is no asset to store, so there is no bill next month. FFmpeg Micro exists for exactly this shape of job.

Self-hosting is the other honest option on these lists. Openinary, the open-source pick, is AGPL-3.0, Dockerized, wants an S3-compatible bucket and uses Cloudinary-compatible URL transformation syntax (GitHub). "Infrastructure costs only" leaves out who patches FFmpeg and who gets paged, which we priced in your PHP Docker image has an FFmpeg layer.

When a video host is the right answer

A hosting platform wins whenever playback is part of the product. If you need adaptive bitrate ladders so a phone on 3G gets 480p and a desktop gets 1080p, a player SDK with analytics, signed playback tokens, or Widevine and FairPlay DRM on a paid catalog, buy a host. FFmpeg Micro does none of that: it has no player, no CDN, no DAM, and no asset library. It hands back a file.

The split also works. Keep images and delivery wherever they are today, and move only the video transform to a job API. That is the version most readers of these lists actually need, and the one no roundup recommends because every roundup is written by somebody with storage to fill.

FAQ about choosing a Cloudinary alternative for video

Do Gumlet, ImageKit or FastPix hand back an MP4 file?

Gumlet and FastPix can produce a downloadable MP4, but only as a feature of the hosted asset: Gumlet needs MP4 enabled in the processing profile, and FastPix needs a static rendition that you fetch from its playback domain. ImageKit's video output is a URL, including its HLS and DASH manifests, so there is no job output file to move into your own bucket.

Does downloading the output bill my delivery meter?

Downloading the output from a hosting platform bills your delivery meter, because the file leaves through the same CDN that serves viewers. On FFmpeg Micro, downloads, status checks and job listings are free, and the only metered thing is the input duration of the job itself.

What happens to the bill after the transform finishes and nobody watches the video?

On a per-GB or storage-minute plan, the storage meter keeps charging every month the asset exists, whether or not it is ever streamed. A processing API has no asset to hold, so the cost of a finished job stops at the job.

Can I process a file that stays in my own S3 bucket?

A processing API reads from a URL you control and writes the output back, so the file never becomes a hosted asset on anybody's platform. Hosting platforms require ingest: the file is copied into their storage, which is where the recurring meters attach.

Can I call any of these from n8n, Make or Zapier?

URL-transform CDNs and player platforms were never designed for no-code workflow tools, which is why video steps in n8n and Make usually end at an HTTP request node. FFmpeg Micro has an official n8n community node with more than 1,800 downloads and a Make app, plus Zapier through a plain HTTP call, and more than 300 builders run it from Make or n8n today.

Take the 10 minute master from the comparison above: 60 tokens on an input meter, nothing afterward, against a storage line that renews every month on any of the four hosting plans. A new account starts with 200 tokens, about 33 minutes of video, no card required. Sign up free and run one real file through it: no servers to run and no FFmpeg to install, called from your code, from n8n, Make and Zapier, or from your AI agents.

About Javid Jamae

Founder & CEO at FFmpeg Micro

Javid is a software engineer, author, and entrepreneur with over 25 years of professional software development experience across enterprise, startup, and consulting environments. He founded FFmpeg Micro to make video processing accessible to developers through a simple, automation-first REST API.

Software EngineeringVideo ProcessingFFmpegCloud ArchitectureAPI DesignAutomation

Ready to process videos at scale?

Start using FFmpeg Micro's simple API today. No infrastructure required.

Get Started Free