Local audio models
Text to speech (TTS) · Speak, transcribe, convert audio, music and sound effects.
Documented GPU memory
Lowest published path per model · GB, log scale · lower is better · 11 of 66 models
- Color = publisher
Source: publishers' documented requirements
Zonos-v0.1: 6 GB
66 models · 11 with documented memory
6 models list Portuguese
Choose your machine and the charts and tables mark what fits in its memory.
Audio models
| AuK | 1.5 | 25 | Neither the Hub card nor the publisher's pages name its languages | Sep 2026 | MIT |
|---|---|---|---|---|---|
| AuK-Flash | The Hub reports no parameter count for this repository | 24.97 | 2 | Sep 2026 | MIT |
| Breeze TTS 2 | 3.47 | 12 | 2 | Aug 2026 | BreezeBlue Research and Non-Commercial |
| Higgs Audio V3 TTS | 4 | 16 | 100 · PT | Jun 2026 | Boson Higgs TTS 3 Research and Non-Commercial |
| Voxtral TTS | 4 | 16 | 9 · PT | Mar 2026 | CC BY-NC 4.0 |
| Fish Audio S2 Pro | 4 | 24 | 83 · PT | Mar 2026 | Fish Audio Research |
| Magpie-Multilingual 357M | 0.364 | No publisher documents a memory path for this model | 12 · PT | Jan 2026 | NVIDIA Open Model |
| GLM-TTS | The Hub reports no parameter count for this repository | No publisher documents a memory path for this model | 2 | Dec 2025 | MIT |
| Step Audio EditX (Mar 2026) | 3 | 12 | 5 | Nov 2025 | Publisher terms |
| VibeVoice 1.5B | 1.5 | No publisher documents a memory path for this model | 2 | Aug 2025 | MIT |
| OpenAudio S1 Mini | 0.5 | 12 | 13 · PT | Jun 2025 | CC BY-NC-SA 4.0 |
| Chatterbox | 0.5 | No publisher documents a memory path for this model | 1 | May 2025 | MIT |
| Kimi-Audio-7B | 9.8 | No publisher documents a memory path for this model | 2 | Apr 2025 | MIT |
| Kimi-Audio-7B-Instruct | 9.8 | No publisher documents a memory path for this model | 2 | Apr 2025 | MIT |
| Zonos-v0.1 | 1.62 | 6 | 5 | Feb 2025 | Apache 2.0 |
| Kokoro 82M v1.0 | 0.082 | No publisher documents a memory path for this model | 1 | Jan 2025 | Apache 2.0 |
| OpenVoice v2 | Not published by the publisher | No publisher documents a memory path for this model | 2 | Apr 2024 | MIT |
| MetaVoice v1 | 1.2 | 12 | 1 | Feb 2024 | Apache 2.0 |
| XTTS v2 | Not published by the publisher | No publisher documents a memory path for this model | 17 · PT | Nov 2023 | Coqui Public Model |
| StyleTTS 2 | Not published by the publisher | No publisher documents a memory path for this model | Neither the Hub card nor the publisher's pages name its languages | Oct 2023 | MIT (code) |
| Maya1 | 3 | 16 | Neither the Hub card nor the publisher's pages name its languages | Release date not found in the publisher's pages | Apache 2.0 |
| csm-1b | 1.6 | No publisher documents a memory path for this model | Neither the Hub card nor the publisher's pages name its languages | No publisher release date recorded; the Hub repository date is not a release date | Apache 2.0 |
| Dia-1.6B | 1.6 | No publisher documents a memory path for this model | Neither the Hub card nor the publisher's pages name its languages | No publisher release date recorded; the Hub repository date is not a release date | Apache 2.0 |
| Dia2-2B | 1.9 | No publisher documents a memory path for this model | Neither the Hub card nor the publisher's pages name its languages | No publisher release date recorded; the Hub repository date is not a release date | Apache 2.0 |
24 of 43
23 more repositories with limited data
| bark | MIT |
|---|---|
| chatterbox | MIT |
| chatterbox-turbo-ONNX | MIT |
| CosyVoice2-0.5B | Apache 2.0 |
| Dramabox | Publisher terms |
| E2-TTS | CC BY-NC 4.0 |
| F5-TTS | CC BY-NC 4.0 |
| fish-speech-1.5 | CC BY-NC-SA 4.0 |
| Fun-CosyVoice3-0.5B-2512 | Apache 2.0 |
| IndexTTS-2 | Not stated |
| IndexTTS-2.5 | Publisher terms |
| kitten-tts-2 | Publisher terms |
| MaskGCT | CC BY-NC 4.0 |
| MeloTTS-English | MIT |
| mms-tts-eng | CC BY-NC 4.0 |
| MOSS-TTS-Nano-100M | Apache 2.0 |
| pocket-tts | CC BY 4.0 |
| Spark-TTS-0.5B | CC BY-NC-SA 4.0 |
| speecht5_tts | MIT |
| supertonic-3 | OpenRAIL |
| tts-1.6b-en_fr | CC BY 4.0 |
| VoxCPM-0.5B | Apache 2.0 |
| ZONOS2 | Apache 2.0 |
No Languages column: neither the Hub cards nor the publishers' pages name their languages.