ListenSpeedScoresπŸ—³ Vote β†—

Speed

TTFA = time to first audio (lower is better). RTF = real-time factor (Γ— realtime; higher is better). Cold = first run after load; warm = subsequent runs. Pick a rig below β€” each shows its default-voice and cloning runs. Tables default to warm RTF, fastest first; click any column header to re-sort. Audio is on the Listen page.
Voice:
Default voiceCloning
Rig:
windows-5090 linux-3090 mac-m4
Rig: windows-5090 β€” AMD Ryzen 9 9950X3D 16-Core Processor (16C) Β· NVIDIA GeForce RTX 5090 32GB Β· 126 GB RAM Β· Windows 11

Default voice Β· 38 models Β· sorted fastest first Β· full report β†—

ModelDeviceTTFA coldTTFA warmRTF coldRTF warmPeak RAMPeak VRAMSize
Kokorocuda895ms67ms8.32Γ—103.68Γ—2.43 GB925 MB82M
MeloTTScuda1.58s111ms5.08Γ—67.28Γ—2.89 GB1.16 GB~52M
Pipercpu160ms107ms38.94Γ—58.83Γ—470 MBβ€”~25MB
StyleTTS 2cpu1.74s251ms4.93Γ—33.76Γ—2.61 GBβ€”~148M
StyleTTS 2cuda1.76s265ms4.80Γ—32.15Γ—2.60 GB1.49 GB~148M
OpenVoice v2cuda1.85s361ms3.65Γ—16.93Γ—2.65 GB1.35 GB~100M
Kokorocpu609ms532ms12.13Γ—14.38Γ—1.82 GBβ€”82M
Supertonic 3cpu744ms741ms9.86Γ—9.92Γ—570 MBβ€”99M
OmniVoicecuda1.12s757ms6.44Γ—9.30Γ—2.07 GB2.16 GB~1B
MeloTTScpu1.96s876ms3.87Γ—9.27Γ—2.48 GBβ€”~52M
KittenTTS Nano 0.1cpu1.22s1.20s6.40Γ—6.39Γ—338 MBβ€”<100M
F5-TTS v1cuda1.31s845ms3.47Γ—5.32Γ—2.67 GB802 MB330M
Coqui XTTS-v2cuda2.05s1.87s3.63Γ—4.75Γ—2.10 GB2.14 GB750M
Chatterbox Turbocuda2.39s1.62s2.80Γ—4.28Γ—2.44 GB3.01 GB744M
OpenVoice v2cpu2.83s1.48s2.10Γ—4.10Γ—2.78 GBβ€”~100M
Pocket-TTScpu147ms123ms3.99Γ—4.06Γ—1.95 GBβ€”100M
Soprano 1.1 80Mcuda1.74s1.77s3.77Γ—3.76Γ—2.12 GB326 MB80M
Qwen3-TTS 1.7B (CUDA-graph)cuda6.51s1.60s0.90Γ—3.76Γ—2.48 GB4.89 GB1.7B
Echo-TTScuda2.83s2.15s2.63Γ—3.44Γ—1.94 GB9.38 GB2.8B
Soprano 1.1 80Mcpu1.97s2.00s3.40Γ—3.40Γ—1.34 GBβ€”80M
NeuTTS Nanocuda678ms258ms2.19Γ—2.76Γ—3.24 GB3.26 GB229M
DramaBoxcuda5.14s3.78s1.93Γ—2.58Γ—2.36 GB17.39 GB3.3B
VibeVoice Realtime 0.5Bcuda3.80s3.77s2.24Γ—2.39Γ—1.88 GB2.62 GB0.5B
Chatterboxcuda3.35s2.61s1.66Γ—2.24Γ—2.80 GB3.24 GB1.2B
NeuTTS Nanocpu698ms303ms1.73Γ—2.00Γ—5.03 GBβ€”229M
Magpie-TTScuda5.48s4.48s1.53Γ—1.93Γ—3.54 GB5.60 GB357M
NeuTTS Aircuda1.15s417ms1.34Γ—1.62Γ—3.60 GB3.26 GB748M
MOSS-TTS-Nanocuda5.83s4.92s1.36Γ—1.61Γ—2.42 GB777 MB100M
VibeVoice 1.5Bcuda4.99s5.32s1.44Γ—1.61Γ—2.04 GB5.26 GB1.5B
VibeVoice 7Bcuda5.67s6.17s1.39Γ—1.55Γ—2.05 GB17.63 GB7B
MOSS-TTS v1.0cuda5.23s4.74s1.34Γ—1.48Γ—2.08 GB22.83 GB8B
VoxCPM2 2Bcuda5.36s5.19s1.35Γ—1.35Γ—6.18 GB5.65 GB2B
IndexTTS-2cpu6.48s5.42s1.09Γ—1.31Γ—5.56 GBβ€”1.5B
NeuTTS Aircpu1.20s471ms1.16Γ—1.29Γ—5.37 GBβ€”748M
MOSS-TTS-Nanocpu6.38s5.64s1.08Γ—1.21Γ—3.14 GBβ€”100M
IndexTTS-2cuda7.20s6.03s0.93Γ—1.11Γ—5.88 GB7.60 GB1.5B
Parler-TTS Mini v1cuda8.29s8.14s0.96Γ—1.05Γ—3.19 GB2.63 GB878M
Fish Speech 1.5cuda8.53s7.84s0.81Γ—0.94Γ—3.83 GB1.80 GB~500M
Coqui XTTS-v2cpu9.92s9.73s0.87Γ—0.88Γ—3.23 GBβ€”750M
Chatterbox Turbocpu10.65s9.92s0.70Γ—0.73Γ—4.16 GBβ€”744M
Zonos v0.1cuda10.88s10.38s0.66Γ—0.71Γ—6.51 GB4.48 GB1.6B
Qwen3-TTS 1.7B Basecuda10.77s8.95s0.55Γ—0.70Γ—2.42 GB4.64 GB1.7B
VibeVoice Realtime 0.5Bcpu14.49s13.35s0.63Γ—0.66Γ—5.89 GBβ€”0.5B
ZipVoice 123M (4/5 ok)cuda68.83s86.51s0.35Γ—0.60Γ—25.87 GB53.16 GB123M
Maya1cuda16.57s14.80s0.53Γ—0.59Γ—2.42 GB6.72 GB3B
Dia 1.6B-0626cuda25.42s22.82s0.46Γ—0.55Γ—4.42 GB6.32 GB1.6B
Sesame CSM-1Bcuda12.01s12.45s0.52Γ—0.54Γ—2.38 GB3.51 GB1B
ZipVoice 123M (3/5 ok)cpu22.96s13.44s0.27Γ—0.45Γ—35.45 GBβ€”123M
VoxCPM2 2Bcpu15.35s14.85s0.46Γ—0.45Γ—10.14 GBβ€”2B
Chatterboxcpu14.47s14.10s0.40Γ—0.43Γ—4.24 GBβ€”1.2B
OmniVoicecpu16.45s15.98s0.38Γ—0.39Γ—3.05 GBβ€”~1B
OuteTTS 1.0 1Bcuda25.01s24.27s0.32Γ—0.33Γ—2.59 GB3.67 GB1B
Magpie-TTScpu37.17s33.72s0.29Γ—0.30Γ—6.10 GBβ€”357M
Mars5-TTScpu31.19s30.69s0.22Γ—0.24Γ—4.03 GBβ€”1.2B
Mars5-TTScuda31.20s31.06s0.23Γ—0.23Γ—2.25 GB6.81 GB1.2B
VibeVoice 1.5Bcpu39.59s45.31s0.19Γ—0.20Γ—11.62 GBβ€”1.5B
Qwen3-TTS 1.7B Basecpu34.81s30.07s0.18Γ—0.19Γ—10.40 GBβ€”1.7B
Fish Speech 1.5cpu45.73s45.34s0.17Γ—0.17Γ—4.46 GBβ€”~500M
Parler-TTS Mini v1cpu63.39s62.77s0.14Γ—0.14Γ—4.31 GBβ€”878M
Zonos v0.1cpu62.12s60.57s0.12Γ—0.12Γ—7.39 GBβ€”1.6B
Sesame CSM-1Bcpu50.90s57.54s0.11Γ—0.12Γ—5.77 GBβ€”1B
F5-TTS v1cpu58.77s60.21s0.07Γ—0.07Γ—2.59 GBβ€”330M
Maya1 (2/4 ok)cpu66.31s73.66s0.07Γ—0.07Γ—7.33 GBβ€”3B
LuxTTScpuLuxTTS install failed (piper-phonemize has no Windows wheels)