Cloudflare Calls Pricing

Developers Cloudflare Calls
1

I’m considering using Cloudflare Calls for video calling. I like the idea of using Cloudflare’s network. Now that it’s in open Beta, some pricing info is available.

“Starting May 15, 2024, customers with a Calls subscription will receive the first terabyte each month for free, with any usage beyond that charged at $0.05 per real-time gigabyte.”

I’d really like to better understand the pricing model, it’s quite different than video calling platforms like Twilio, Zoom, Dyte, etc., which is understandable, since it’s a different value proposition.

  1. What’s a real-time gigabyte?
  2. How many real-time gigabytes does a 1 hour, 2 participant call use? Using Orange Meets as a benchmark seems reasonable, since Cloudflare uses it as its internal video conferencing app :slight_smile:
  3. What, if any, other Cloudflare costs do I need to account for when running the end-to-end Orange Calls application?
3

Hi @matt47

We do not have this information yet, but the Calls documentation should be updated with the pricing structure soon, so please keep an eye on the documentation.

4

Hi,
I am also interested in this solution and I have the same question as @matt47 : what does ‘$0.05 per real-time gigabyte’ mean if I only use audio, for example ? There hasn’t been an update on the pricing.
For example how much Mo for 2 users audio call for 15minutes ?

Cc @louise2

5

I think it is unlikely that we will give you an estimate how to translate one gigabyte into minutes on a call, or estimate how many bytes a 15min will use. The reason is that these numbers depend on a lot of factors:
For example for video it depends on which video codec your system uses. When attendees face mute a lot it is obviously going to be a lot less.
The same for audio. The traffic generated/used depends a lot on the audio codec used. And even with the widely used Opus audio codec it depends on configurable things like DTX, FEC, RED etc.

I would recommend everyone to setup a test system your self and then measure the data usage for your specific test systems. This will give you way more realistic estimatest than anything we could guess here.