Video Generation Processing Time and Factors affecting it

simonwilbert Member Posts: 5
edited September 2023 in Developer APIs

Hey there,

I know the video generation is still in Beta, but I'd love to understand how long you feel videos should take to get generated and what factors affect this in the following 2 scenarios:

  1. A user uploads a zipped video manifest file via API and then takes the value of the URL property in the response of this upload and replaces the extension with .mp4 to get a preview of the uploaded manifest
  2. A user does an eager fetch of an existing video/manifest passing some vars property to customise the video (e.g. a different background colour, etc.) and waits for the eager webhook notification to arrive

Maybe you have some documentation on this already?

I presume the duration of the video, width/height, and FPS greatly affect the generation time?

I also presume the generation has to go into some kind of queue before getting processed. If so, are we able to affect the prioritisation of the queue in any way? for example, do customers who pay more get more dedicated resources and priority queues?

Also, are there ways of short-cutting video generation? for example, if we only change the background colour, does that somehow reduce the generation time?

Any help you can offer would be great. I'd love to understand all of this area in much more detail so we can fully utilise Cloudinary.

Sorry, this is quite a big question!



  • Cloudinary_John_H
    Cloudinary_John_H Cloudinary Staff Posts: 37

    Hi Simon,

    Thanks for reaching out!

    As you mentioned - yes, this product is in Beta. As for all of our Beta products, they do not come with SLAs or guaranteed performance expectations.

    We do not have any expected performance timelines documented for video generation, as performance is subject to change as the product evolves.

    Yes, the size of the output - which is reflected in factors like duration, height/width, and FPS are all factors in how long the video generation will take.

    There is no way to impact queue prioritization at this time.

    I'm not aware of any shortcuts at this time, but feel free to test the small iteration use-case. It is possible that this could save some time if the system is smart enough to only render the portion that is changing, but being that this is in beta, I am inclined to believe those sorts of performance optimizations have yet to be implemented.

    Apologies for not having better answers on these topics. Our features in Beta are more aimed at finding the right fit for our customers, and then performance upgrades normally come afterward.