TTS Bench — Speed — windows-es-default

Rig: windows-5090 — AMD Ryzen 9 9950X3D 16-Core Processor (16C) · NVIDIA GeForce RTX 5090 32GB · 126 GB RAM · Windows 11

Label: Spanish default voice

TTFA = time to first audio (ms; lower is better). RTFx = realtime speed (× 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 predefined-voice: Kokoro (cuda) — 93.41× warm RTFx, 67ms warm TTFA

Fastest cloning-capable: OpenVoice v2 (cuda) — 18.38× warm RTFx, 395ms warm TTFA

Model	Device	TTFA cold	TTFA warm	RTFx cold	RTFx warm	Peak RAM	Peak VRAM	Size	Released
Kokoro	cuda	15.30s	67ms	0.41×	93.41×	2.35 GB	682 MB	82M	Dec 2024
MeloTTS	cuda	13.32s	84ms	0.54×	85.54×	2.44 GB	443 MB	~52M	Feb 2024
Piper	cpu	74ms	34ms	31.73×	47.81×	240 MB	—	~25MB	Jan 2023
OpenVoice v2	cuda	6.89s	395ms	1.07×	18.38×	2.65 GB	705 MB	~100M	Apr 2024
Audio8 TTS 0.6B (compiled)	cuda	2.36s	542ms	3.37×	14.64×	3.04 GB	2.24 GB	601M	Aug 2026
Kokoro	cpu	810ms	472ms	7.69×	13.19×	1.64 GB	—	82M	Dec 2024
Supertonic 3	cpu	835ms	807ms	9.60×	9.93×	510 MB	—	99M	May 2026
MeloTTS	cpu	1.34s	842ms	5.50×	8.80×	1.68 GB	—	~52M	Feb 2024
Scylla's Band	cpu	2.54s	898ms	2.45×	6.93×	607 MB	—	~103M	Jul 2026
OpenVoice v2	cpu	2.88s	1.63s	2.51×	4.32×	1.95 GB	—	~100M	Apr 2024
Coqui XTTS-v2	cuda	17.82s	2.65s	0.40×	3.72×	2.09 GB	1.89 GB	750M	Oct 2023
Qwen3-TTS 1.7B (CUDA-graph)	cuda	23.09s	1.77s	0.30×	3.60×	2.52 GB	4.85 GB	1.7B	Jan 2026
Pocket-TTS	cpu	457ms	398ms	1.10×	1.16×	2.02 GB	—	100M	Jan 2026
Audio8 TTS Preview 0.6B	cuda	27.71s	9.33s	0.25×	0.75×	2.96 GB	2.18 GB	601M	Jul 2026
Audio8 TTS Preview 0.1B	cuda	30.77s	12.20s	0.25×	0.64×	2.96 GB	1.39 GB	170M	Aug 2026
Qwen3-TTS 0.6B Base	cuda	27.43s	8.87s	0.23×	0.63×	2.45 GB	2.68 GB	0.6B	Jan 2026
Coqui XTTS-v2	cpu	10.60s	12.04s	0.65×	0.60×	3.01 GB	—	750M	Oct 2023
Qwen3-TTS 1.7B Base	cuda	27.50s	11.50s	0.22×	0.54×	2.48 GB	4.56 GB	1.7B	Jan 2026
Audio8 TTS Preview 0.1B	cpu	24.61s	21.19s	0.31×	0.36×	3.64 GB	—	170M	Aug 2026
Audio8 TTS Preview 0.6B	cpu	56.83s	49.77s	0.14×	0.16×	5.45 GB	—	601M	Jul 2026
NeuTTS Nano	cuda	5.56s	165ms	0.00×	0.12×	2.53 GB	3.26 GB	229M	Dec 2025
NeuTTS Nano	cpu	733ms	206ms	0.03×	0.10×	5.02 GB	—	229M	Dec 2025
Qwen3-TTS 1.7B Base	cpu	Timed out after 10 min — model too slow at this prompt length
Qwen3-TTS 0.6B Base	cpu	Timed out after 10 min — model too slow at this prompt length

Model

Device

TTFA cold

TTFA warm

RTFx cold

RTFx warm

Peak RAM

Peak VRAM

Size

Released

Kokoro

cuda

15.30s

67ms

0.41×

93.41×

2.35 GB

682 MB

82M

Dec 2024

MeloTTS

cuda

13.32s

84ms

0.54×

85.54×

2.44 GB

443 MB

~52M

Feb 2024

Piper

cpu

74ms

34ms

31.73×

47.81×

240 MB

—

~25MB

Jan 2023

OpenVoice v2

cuda

6.89s

395ms

1.07×

18.38×

2.65 GB

705 MB

~100M

Apr 2024

Audio8 TTS 0.6B (compiled)

cuda

2.36s

542ms

3.37×

14.64×

3.04 GB

2.24 GB

601M

Aug 2026

Kokoro

cpu

810ms

472ms

7.69×

13.19×

1.64 GB

—

82M

Dec 2024

Supertonic 3

cpu

835ms

807ms

9.60×

9.93×

510 MB

—

99M

May 2026

MeloTTS

cpu

1.34s

842ms

5.50×

8.80×

1.68 GB

—

~52M

Feb 2024

Scylla's Band

cpu

2.54s

898ms

2.45×

6.93×

607 MB

—

~103M

Jul 2026

OpenVoice v2

cpu

2.88s

1.63s

2.51×

4.32×

1.95 GB

—

~100M

Apr 2024

Coqui XTTS-v2

cuda

17.82s

2.65s

0.40×

3.72×

2.09 GB

1.89 GB

750M

Oct 2023

Qwen3-TTS 1.7B (CUDA-graph)

cuda

23.09s

1.77s

0.30×

3.60×

2.52 GB

4.85 GB

1.7B

Jan 2026

Pocket-TTS

cpu

457ms

398ms

1.10×

1.16×

2.02 GB

—

100M

Jan 2026

Audio8 TTS Preview 0.6B

cuda

27.71s

9.33s

0.25×

0.75×

2.96 GB

2.18 GB

601M

Jul 2026

Audio8 TTS Preview 0.1B

cuda

30.77s

12.20s

0.25×

0.64×

2.96 GB

1.39 GB

170M

Aug 2026

Qwen3-TTS 0.6B Base

cuda

27.43s

8.87s

0.23×

0.63×

2.45 GB

2.68 GB

0.6B

Jan 2026

Coqui XTTS-v2

cpu

10.60s

12.04s

0.65×

0.60×

3.01 GB

—

750M

Oct 2023

Qwen3-TTS 1.7B Base

cuda

27.50s

11.50s

0.22×

0.54×

2.48 GB

4.56 GB

1.7B

Jan 2026

Audio8 TTS Preview 0.1B

cpu

24.61s

21.19s

0.31×

0.36×

3.64 GB

—

170M

Aug 2026

Audio8 TTS Preview 0.6B

cpu

56.83s

49.77s

0.14×

0.16×

5.45 GB

—

601M

Jul 2026

NeuTTS Nano

cuda

5.56s

165ms

0.00×

0.12×

2.53 GB

3.26 GB

229M

Dec 2025

NeuTTS Nano

cpu

733ms

206ms

0.03×

0.10×

5.02 GB

—

229M

Dec 2025

Qwen3-TTS 1.7B Base

cpu

Timed out after 10 min — model too slow at this prompt length

Qwen3-TTS 0.6B Base

cpu

Timed out after 10 min — model too slow at this prompt length