In view of the new Simba Turbo model launch we have made improvements to the voices APIs. The changes are as follows:

When querying the list of the available voices, the result contains both your cloned voices and the system-wide voices provided by Speechify.

SSML is a powerful tool that gives you precise control over how your text is being pronounced. But it's an extra hassle that to deal with, and it's not strictly necessary for every use-case, especially for simple scenarios.

We now support two voice synthesis models. The original, known as simba-base, now supports multilingual inputs, see for the list of supported languages . The new model, simba-turbo, is English-only, but produces the audio faster.

We've released the new API for cloned voice deletion. It's a simple DELETE request to /v1/voices/{id}, and the Authorization header is required, of course.

Our Playground now shows the preview for the shared voices. This lets you check how the voice sounds without spending your credits on test runs.

When you have large input, you may want to start playing the synthesized voice audio as early as possible, while the system is still converting the rest of the text.