whatmodel.sh / tts / F5-TTS Base (v0)

F5-TTS Base (v0)

+ contribute
Update model informationAdd a place to run itAdd a test resultPropose a benchmarkCorrect a benchmark scoreEdit data on GitHub

F5-TTS Base (v0) is an open-weights text-to-speech model from SWivid (SJTU X-LANCE), released October 2024. Its published size is 0.3358B parameters. Licence: cc-by-nc-4.0.

Catalogue identity notes

SWivid/F5-TTS F5TTS_Base/model_1200000.safetensors, the original (v0) Base release. Distinct from F5TTS_v1_Base (2025-03).

Model information sources

· Parameter count ↗

Maker paper (SWivid team). Matches the catalogue checkpoint F5TTS_Base/model_1200000 (1.2M updates; F5TTS_Base.yaml dim 1024, depth 22, heads 16). Covers the F5-TTS DiT + ConvNeXt text model only; excludes the separate Vocos vocoder.

56.9±2.4#48 of 118 text-to-speech modelsfrom 10 benchmark results across 3 sources

Score within the Vapi Humanness · Bradley–Terry rating, 2026-09-25 snapshot comparison set only.

$ how it compares

The text-to-speech models ranked either side of F5-TTS Base (v0), out of 148 with a score, then more from SWivid (SJTU X-LANCE).

#modelscorewhereRAMreleased
58GPT-4o mini TTS OpenAI58±2comparison only2025-03
59Sonic 2 Cartesia58±3comparison only2025-03
60F5-TTS Base (v0) SWivid (SJTU X-LANCE)57±2local~1.4GB2024-10
61TTS-1 HD OpenAI57±7api $30.00/M chars2023-11
62Deepgram Aura 2 Deepgram57±3api $30.00/M chars2025-04
more from SWivid (SJTU X-LANCE)
40F5-TTS v1 Base SWivid (SJTU X-LANCE)62±3local~1.4GB2025-03
50E2 TTS Base (F5-TTS reproduction) SWivid (SJTU X-LANCE)60±2local~1.3GB2024-10

see all text-to-speech models · find the ones that fit your machine

$ what it's good at

voice cloning▁·▁▁▁▁▁▁▇█▃▃ #24 of 37 · 38% of its score · 4 tests

Beats GLM-TTS, CosyVoice2-0.5B, MaskGCT here, though they rank higher overall.

testresultvs the others in its rundifficultycountswhy
Seed-TTS test-en · speaker similarity (WavLM-large SV) lab67.0% ✎▁·▂▂▂█▆▃▅··▂ beats 7/30medium · avg 53%▇▇▇▇▇▇ 15%at its level
Seed-TTS test-zh · speaker similarity (WavLM-large SV) lab76.0% ✎▁·▂··▂·█▄·▃▄ beats 10/25medium · avg 62%▇▇▇▇▇ 13%at its level
Seed-TTS test-en · Whisper large-v3 WER lab1.8% ✎▁···▂▂▂▂█▆▄▂ beats 16/35avg 3.6%▇▇ 6.2%blunt
Seed-TTS test-zh · Paraformer CER lab1.6% ✎▁·▂▁▁▁▃▃▅█·▂ beats 6/30avg 2.1%▇ 3.4%blunt
other tests6 not in a task · 59% of its score
testresultvs the others in its rundifficultycountswhy
TTSDS2 listening test · speaker-similarity MOS (4 English domains)4.10 / 5 ✎▅·█▅▅·█▃▃▅·▅ beats 15/17easy · avg 4.03 / 5▇▇▇▇▇▇▇▇▇▇ 26%typical
LibriSpeech-PC test-clean (F5-TTS subset) · speaker similarity (WavLM SV) lab66.0% ✎▄·········█▄ beats 1/3medium · avg 52%▇▇▇▇▇▇▇ 17%at its level
LibriSpeech-PC test-clean (F5-TTS 4-10 s subset) · Whisper large-v3 WER lab2.4% ✎█·····█·█··█ beats 3/3avg 4.1%▇▇▇ 7.5%typical
TTSDS2 listening test · naturalness MOS (4 English domains)3.33 / 5 ✎▅··▃▅··▅█▅██ beats 15/17medium · avg 3.30 / 5▇ 3.6%blunt · at its level
TTSDS2 multilingual benchmark · Chinese overall (YouTube, 2025 round)89.4% ✎▄▄·····▄··▄█ beats 4/5very easy · avg 88%▇ 3.0%blunt
TTSDS2 multilingual benchmark · English overall (YouTube, 2025 round)89.6% ✎▂▂█·▂·█▆▂·▄▂ beats 16/17very easy · avg 89%▇ 2.1%blunt

$ where to run it

Other ways to run it

  1. Run it as published · needs ~1.35 GB · from github.com/SWivid (official)
    pip install f5-tts && f5-tts_infer-cli --model F5TTS_Base --gen_text "Hello, world."

Official weights and documentation: huggingface.co/SWivid ↗

$ how the score was worked out · F5-TTS: A Fairytaler that Fak… 63%, TTSDS2 listening test results… 30%, TTSDS2 multilingual benchmark… 5%

Each result implies a score on its own, after adjusting for how hard its test is and who ran it. The score is their weighted average, with a small pull toward the middle for thinly tested models. Each test's weight is its counts column above; here it's added up by who ran the tests. How it works

sourceresultsshare of the scorelicence
F5-TTS: A Fairytaler that Fakes Fluent and Faithful Speech with Flow Matching (arXiv:2410.06885v1) · lab-reported6▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇ 63%CC BY 4.0
TTSDS2 listening test results (ttsds/listening_test)2▇▇▇▇▇▇▇▇▇ 30%MIT
TTSDS2 multilingual benchmark results (ttsds/leaderboard_data)2▇▇ 5%MIT
prior · pull toward the middle▇ 2%
weighted average= 56.9

$ improve this page

Update model informationAdd a place to run itAdd a test resultPropose a benchmarkCorrect a benchmark scoreEdit data on GitHub

Use a short GitHub form or edit the data through a pull request. Every number needs a source; here's which ones we can use.

data updated · sources and licences · json