Skip to main content

Samples

One clip per engine, each made by one request to the same server. Under every player is the text that request sent, cue tags included, so you can hear what each tag did and whether the engine performed it.

The cloned voices are Sonny and Tiffani, two hosts from deadair.radio. Both were invented with a voice-design tool rather than recorded from anybody, and each engine that clones learned them from the same reference clip: 7 to 8 seconds of it, with its transcript, for Dia, Fish and Breeze, and the first 3.6 seconds alone for StyleTTS 2. Kokoro and Orpheus cannot clone, so they speak in one of their own voices.

Every clip came from rhapsode 0.1.26 in Docker on an RTX 4070 Ti SUPER, in October 2026. Each line was generated with three seeds and the best take kept. The takes left out misread "rhapsode", stopped early, or repeated words from the reference clip. The Fish and Breeze weights are licensed for research and not for commercial use, which Engines explains.

Cloned voices​

Chatterbox turbo, Tiffani
Okay so, like, this next record? It's totally the best song ever made. [laugh] I'm serious. My dad says I say that about every song, and he's, like, not wrong. This was generated by Chatterbox via Rhapsode.
Chatterbox turbo, Sonny
[sigh] It's late, baby. The city's asleep, and it's just you, me, and the radio. Lean a little closer. This was generated by Chatterbox via Rhapsode.
Dia, Sonny and Tiffani in one take
Sonny: Mm. Easy now, darling. This is the slow hour.
Tiffani: Oh my god, Sonny, it's, like, the slow hour every hour with you. [laugh]
Sonny: [chuckle] And you keep coming back, don't you.
Tiffani: This was generated by Dee-ah via Rhapsode.
Fish Audio S2 Pro, Tiffani in three languages. Built with Fish Audio.
This station is, like, totally the best. Esta estación es, o sea, totalmente la mejor. この放送局は、なんていうか、マジで最高なの。 This was generated by Fish Audio via Rhapsode.
StyleTTS 2, Sonny from 3.6 seconds and no transcript
Three seconds of my voice. That's all it took, baby. This was generated by Style T T S Two via Rhapsode.
Breeze, Sonny
[sigh] Another long night, darling. [chuckle] Stay with me for one more record. This was generated by Breeze via Rhapsode.

Stock voices​

Kokoro, af_heart
Good evening, and welcome back. It's eleven o'clock, the rain hasn't let up, and we've got three more hours of music before sunrise. This was generated by Kokoro via Rhapsode.
Orpheus q8, tara
[gasp] Oh, we're live? Sorry, sorry. [chuckle] Right. Coming up after the break: the weather, the traffic, and a song you definitely forgot you loved. This was generated by Orpheus via Rhapsode.

Two names are spelled the way they are said, "Dee-ah" and "Style T T S Two", so the engines read them as names.