Voice & Music models
7 models for speech, voice and music, newest first. This list shows what exists and when it appeared. Side-by-side quality tests for media models are in preparation. Catalog data
| Model | Task | Price in / out | Weights | Added |
|---|---|---|---|---|
| StepAudio 2.5 TTSStepFun | speech and audio | per image, second or character | proprietary | Apr 2026 |
| Gemini 3.1 Flash TTS PreviewGoogle | speech and audio | $1 / $20 per 1M tokens | proprietary | Apr 2026 |
| Step TTS 2StepFun | speech and audio | per image, second or character | proprietary | Mar 2026 |
| Canopy Labs Orpheus V1 EnglishGroq (host) | speech and audio | per image, second or character | proprietary | Dec 2025 |
| Canopy Labs Orpheus Arabic SaudiGroq (host) | speech and audio | per image, second or character | proprietary | Dec 2025 |
| Gemini 2.5 Pro Preview TTSGoogle | speech and audio | $1 / $20 per 1M tokens | proprietary | May 2025 |
| Gemini 2.5 Flash Preview TTSGoogle | speech and audio | $0.5 / $10 per 1M tokens | proprietary | May 2025 |
Source: the open models.dev dataset (names, dates, token prices where they exist). Many media models are billed per image, per second or per character; those prices are not in the catalog yet.
