Discussions

We got complaints by our customers because D-ID video generation failed.

When looking at the root cause, we noticed the following changes:

• The voices endpoint now includes multiple languages as an array instead of returning the locale and the language as part of the main object. This is not even documented in your current API docs!
• After fixing this, we noticed that the talks endpoint does not accept Amazon voice ids anymore. Instead, it expects the name of the voice instead. Which is contrary to the Microsoft voices (here you have to provide the id). Previously, both used the id.
• So we changed this as well. Only to notice that not all voice names that are reported by the voices endpoint are accepted. For instance, if we try to select "Niamh", we get the error that this voice is not included in the supported voices.
• Also, the error message sent back on all of these errors claims that the voice provider can be "elevenlabs", "afflorithmics", "google", or "playHT". But it does not mention "amazon" or "microsoft". Which is strange, because we can use these voices (once we fixed the aforementioned breaking changes).

This is the second time for us that D-ID changes an API without giving any notice about this (and even without documenting the new API properly). Guys, this is completely unacceptable for enterprise use. If you have to make changes, then please version your APIs properly.