I need to find a good text-to-speech model for a project.

$1k manalink if I use a TTS you recommended for my audiobook project!

This market resolves positively if I find a TTS that satisfies me (and most of my desiderata).

It needs to have most of these characteristics:

• Be relatively cheap and preferably free, absolute maximum $200 per 500k words (or find me funding idc). (I need more than 2M words total so high-volume deals can be made)

• Be high enough quality that it is nearly indistinguishable from human speech (see ultra-realistic voices here). If I can hear weird artifacts every few sentences, this fails.

• Have implicit/context-dependent intonation.

• Have a usable API.

• Reply to my emails (this is where play.ht fails).

• Have at least 20 distinct voices or voice cloning technology.

• It must give the ability to test in demos/cheap subscriptions.

• Pay-per-use.

• Unrelated plus: have good voices in French (independent manalink of $1k for this, doesn't need to have most of the above characteristics).

• (I might add other characteristics over time if I think of them)

The resolution criterion does not depend on the TTS fitting all these desiderata perfectly or completely; so long as I am satisfied with the TTS, this market resolves positively. However, I guarantee that if all of these characteristics are respected, I will resolve it positively.