click any column header to sort

TTS Bench — Speed — windows-cloning

Rig: windows-5090 — AMD Ryzen 9 9950X3D 16-Core Processor (16C) · NVIDIA GeForce RTX 5090 32GB · 126 GB RAM · Windows 11
Label: cloning · chris_hemsworth_15s — ref chris_hemsworth_15s.wav
TTFA = time to first audio (ms; lower is better). RTF = real-time factor (× realtime; higher is better; e.g. 10× means 10 sec of audio generated per 1 sec of compute). Cold = first run after process start; warm = subsequent runs.

Speed winners

Fastest (this cloning run): StyleTTS 2 (cuda) — 41.66× warm RTF, 212ms warm TTFA

ModelDeviceTTFA coldTTFA warmRTF coldRTF warmPeak RAMPeak VRAMSize
StyleTTS 2cuda1.66s212ms5.45×41.66×2.63 GB1.52 GB~148M
StyleTTS 2cpu1.70s218ms5.27×40.62×2.63 GB~148M
OpenVoice v2cuda1.72s361ms3.61×16.73×2.65 GB1.35 GB~100M
OmniVoicecuda1.20s803ms6.67×9.75×2.27 GB2.42 GB~1B
MiraTTScuda2.14s1.46s4.76×7.01×3.58 GB1.28 GB0.5B
LongCat-AudioDiT 1Bcuda1.66s1.36s4.72×5.67×2.36 GB5.78 GB1.42B
F5-TTS v1cuda1.56s1.06s3.75×5.50×2.67 GB883 MB330M
Coqui XTTS-v2cuda2.02s1.51s3.02×4.76×2.35 GB2.10 GB750M
OpenVoice v2cpu2.60s1.44s2.23×4.25×2.78 GB~100M
Chatterbox Turbocuda2.66s1.59s2.43×4.15×2.69 GB3.33 GB744M
Pocket-TTScpu136ms124ms4.11×4.10×2.29 GB100M
Qwen3-TTS 1.7B (CUDA-graph)cuda7.93s1.57s0.83×3.66×2.61 GB4.89 GB1.7B
LongCat-AudioDiT 3.5Bcuda2.59s2.32s3.13×3.49×2.26 GB15.14 GB3.83B
Echo-TTScuda2.84s2.23s2.62×3.39×1.94 GB9.38 GB2.8B
Chatterboxcuda4.01s2.88s1.63×2.33×3.02 GB3.63 GB1.2B
DramaBoxcuda6.46s4.99s1.66×2.14×2.45 GB17.39 GB3.3B
WavTTS 0.67Bcuda3.16s2.84s1.85×2.08×2.00 GB1.44 GB0.67B
NeuTTS Nanocuda854ms331ms1.60×2.00×2.59 GB3.26 GB229M
MOSS-TTS-Nanocuda5.59s4.87s1.34×1.66×2.42 GB910 MB100M
VibeVoice 7Bcuda5.22s4.86s1.41×1.58×2.04 GB17.63 GB7B
NeuTTS Nanocpu863ms387ms1.23×1.58×5.12 GB229M
IndexTTS-2cpu5.77s4.87s1.20×1.47×5.54 GB1.5B
MOSS-TTS v1.0cuda5.44s4.73s1.31×1.44×2.10 GB22.83 GB8B
VoxCPM2 2Bcuda6.36s4.72s0.96×1.33×2.49 GB5.71 GB2B
NeuTTS Aircuda1.46s492ms0.88×1.30×2.96 GB3.26 GB748M
IndexTTS-2cuda6.80s5.66s1.03×1.26×4.05 GB7.71 GB1.5B
MOSS-TTS-Nanocpu5.69s5.04s1.16×1.25×3.05 GB100M
MOSS-TTS v1.5cuda5.88s5.28s1.14×1.24×2.36 GB22.83 GB8B
NeuTTS Aircpu1.75s568ms0.75×0.96×5.47 GB748M
Fish Speech 1.5cuda8.04s8.21s0.85×0.95×3.82 GB1.76 GB~500M
Coqui XTTS-v2cpu9.93s8.76s0.84×0.85×3.43 GB750M
Zonos v0.1cuda10.36s9.85s0.67×0.73×6.55 GB4.45 GB1.6B
Chatterbox Turbocpu10.84s9.37s0.68×0.71×4.39 GB744M
ZipVoice 123M (4/5 ok)cuda68.19s105.99s0.38×0.60×25.56 GB53.24 GB123M
Sesame CSM-1Bcuda13.21s11.46s0.47×0.55×2.47 GB3.54 GB1B
Dia 1.6B-0626cuda14.04s11.22s0.33×0.47×8.53 GB4.82 GB1.6B
ZipVoice 123Mcpu50.03s47.06s0.30×0.43×39.66 GB123M
VoxCPM2 2Bcpu18.12s16.04s0.34×0.39×7.53 GB2B
Chatterboxcpu17.39s17.39s0.36×0.38×4.72 GB1.2B
Miso TTS 8Bcuda37.67s32.57s0.30×0.32×2.07 GB16.55 GB8.2B
OuteTTS 1.0 1Bcuda18.32s17.44s0.30×0.30×3.64 GB3.56 GB1B
Fish Speech 1.5cpu37.45s42.66s0.18×0.18×4.43 GB~500M
OmniVoicecpu54.74s53.28s0.14×0.14×3.98 GB~1B
Zonos v0.1cpu61.60s60.84s0.12×0.12×8.57 GB1.6B
Sesame CSM-1Bcpu48.90s48.64s0.12×0.12×6.32 GB1B
Mars5-TTScpu41.89s43.23s0.08×0.10×2.24 GB1.2B
F5-TTS v1cpu71.35s71.18s0.08×0.08×2.71 GB330M
Mars5-TTScuda41.12s39.53s0.09×0.07×2.23 GB7.23 GB1.2B
LuxTTScpuLuxTTS install failed (piper-phonemize has no Windows wheels)