Skip to content
Albastar Corp.

Voice models

Access: Keys with the media:generate scope, on an account with media access granted by Albastar Corp.

Text-to-speech models by name, with the capabilities, languages and a sample of each.

Each text-to-speech model carries one or more labels. Built-in voices: the model speaks in its own named voices, such as Kokoro's af_heart. Voice cloning: the model speaks in the voice of a clip, given as a saved voice vo_…, an upload up_… or an https URL. Voice design: the model makes a voice from a text description, without a clip. Every cloning sample clones the same reference clip, Narrator A (vo_tBeDCRb2byGFjB6HGyQq6a, a synthetic voice), and every model page shows the request that made its sample.

  • BreezeTTS 2breeze22 languagesVoice cloningClones a voice in Chinese and English; instruct describes the delivery.
  • Chatterboxchatterbox19 languagesVoice cloningClones a voice in 19 languages, using up to 20 s of the clip.
  • CosyVoice 3cosyvoice39 languagesVoice cloningClones a voice in 9 languages; the clip and the text may be in different languages.
  • FireRedTTS 3fireredtts324 languagesVoice cloningClones a voice in 24 languages, including Arabic, Thai and Vietnamese.
  • Fish Audio S2 Profish-s2pro80+ languagesVoice cloningClones a voice in 80+ languages, using up to 20 s of the clip.
  • Higgs Audio v3higgs3100+ languagesVoice cloningClones a voice in 100+ languages, using up to 20 s of the clip.
  • IndexTTS 2.5indextts2.52 languagesVoice cloningClones a voice in English or Chinese; language on a media job selects Chinese.
  • Kokorokokoro8 languagesBuilt-in voicesSpeaks text in 54 built-in voices in 8 languages; speed sets the rate from 0.5 to 2.0.
  • MioTTSmiotts2 languagesVoice cloningClones a voice in English and Japanese, using up to 20 s of the clip.
  • MOSS-TTS Local Transformermoss-local31 languagesVoice cloningClones a voice in 31 languages, including Cantonese, Persian and Tagalog.
  • OmniVoiceomnivoice600+ languagesVoice cloningVoice designClones a voice, or designs one from a description in instruct, in 600+ languages.
  • Qwen3-TTSqwen3-tts10 languagesVoice cloningClones a voice in 10 languages, using up to 22 s of the clip.
  • VoxCPM 2voxcpm230 languagesVoice cloningClones a voice in 30 languages, detected from the text, using up to 11 s of the clip.