af_heart
Female · American English
54 preset voices · 8 languages
Kokoro Voices — Full List of 54 Voices (Free, In-Browser)
Search the complete Kokoro v1.0 list and compare the official model grades. Play a small prerecorded sample for any of the 28 American and British English voices without loading the TTS model. The other 26 presets remain in the reference catalog without sample audio. Try a supported English voice with your own text in the free Kokoro TTS tool.
The source lists 54 voices in these groups: American English (20), British English (8), Japanese (5), Mandarin Chinese (8), Spanish (3), French (1), Hindi (4), Italian (2), Brazilian Portuguese (3). The catalog calls American and British English separate groups; the model release counts eight languages overall.
All 54 Kokoro Voices by Language
The official list publishes voice IDs, speaker labels, and some training-data grades, but it does not provide consistent subjective tone descriptions. The English previews are compact MP3 files loaded only when played; they do not run TTS inference or download the model. Other language rows are listed for reference without sample audio.
Start with af_heart · Grade A
Listen to a small, prerecorded sample. It streams directly from this site; the Kokoro model is only needed when you generate your own speech.
Choose an English sample to listen. Preview audio loads without downloading the TTS model.
Showing 54 of 54
American English — 20 voices (11 female, 9 male)
Grade A
af_alloy
Female · American English
af_aoede
Female · American English
af_bella
Female · American English
af_jessica
Female · American English
af_kore
Female · American English
af_nicole
Female · American English
af_nova
Female · American English
af_river
Female · American English
af_sarah
Female · American English
af_sky
Female · American English
am_adam
Male · American English
am_echo
Male · American English
am_eric
Male · American English
am_fenrir
Male · American English
am_liam
Male · American English
am_michael
Male · American English
am_onyx
Male · American English
am_puck
Male · American English
am_santa
Male · American English
British English — 8 voices (4 female, 4 male)
bf_alice
Female · British English
bf_emma
Female · British English
bf_isabella
Female · British English
bf_lily
Female · British English
bm_daniel
Male · British English
bm_fable
Male · British English
bm_george
Male · British English
bm_lewis
Male · British English
Japanese — 5 voices (4 female, 1 male)
jf_alpha
Female · Japanese
Official model preset · sample audio is not available for this voice
jf_gongitsune
Female · Japanese
Official model preset · sample audio is not available for this voice
jf_nezumi
Female · Japanese
Official model preset · sample audio is not available for this voice
jf_tebukuro
Female · Japanese
Official model preset · sample audio is not available for this voice
jm_kumo
Male · Japanese
Official model preset · sample audio is not available for this voice
Mandarin Chinese — 8 voices (4 female, 4 male)
zf_xiaobei
Female · Mandarin Chinese
Official model preset · sample audio is not available for this voice
zf_xiaoni
Female · Mandarin Chinese
Official model preset · sample audio is not available for this voice
zf_xiaoxiao
Female · Mandarin Chinese
Official model preset · sample audio is not available for this voice
zf_xiaoyi
Female · Mandarin Chinese
Official model preset · sample audio is not available for this voice
zm_yunjian
Male · Mandarin Chinese
Official model preset · sample audio is not available for this voice
zm_yunxi
Male · Mandarin Chinese
Official model preset · sample audio is not available for this voice
zm_yunxia
Male · Mandarin Chinese
Official model preset · sample audio is not available for this voice
zm_yunyang
Male · Mandarin Chinese
Official model preset · sample audio is not available for this voice
Spanish — 3 voices (1 female, 2 male)
ef_dora
Female · Spanish
Official model preset · sample audio is not available for this voice
em_alex
Male · Spanish
Official model preset · sample audio is not available for this voice
em_santa
Male · Spanish
Official model preset · sample audio is not available for this voice
French — 1 voice (1 female, 0 male)
ff_siwis
Female · French
Official model preset · sample audio is not available for this voice
Hindi — 4 voices (2 female, 2 male)
hf_alpha
Female · Hindi
Official model preset · sample audio is not available for this voice
hf_beta
Female · Hindi
Official model preset · sample audio is not available for this voice
hm_omega
Male · Hindi
Official model preset · sample audio is not available for this voice
hm_psi
Male · Hindi
Official model preset · sample audio is not available for this voice
Italian — 2 voices (1 female, 1 male)
if_sara
Female · Italian
Official model preset · sample audio is not available for this voice
im_nicola
Male · Italian
Official model preset · sample audio is not available for this voice
Brazilian Portuguese — 3 voices (1 female, 2 male)
pf_dora
Female · Brazilian Portuguese
Official model preset · sample audio is not available for this voice
pm_alex
Male · Brazilian Portuguese
Official model preset · sample audio is not available for this voice
pm_santa
Male · Brazilian Portuguese
Official model preset · sample audio is not available for this voice
The names, language groups, gender labels, and available overall grades follow the official Kokoro voice list. The English MP3 samples are prerecorded clips from the ONNX Community model card; each clip loads only when played. A missing grade means the source does not publish one for that voice.
Which Kokoro Voice Should You Pick?
Start by listening with the same sentence you plan to use. The published overall grades can help narrow the list: af_heart is rated A, af_bella A-, while af_nicole and British English bf_emma are B-. These grades estimate training-data quality and quantity; they do not measure whether a voice fits your audience, script, or preferred delivery.
For an audiobook, compare several voices across a longer passage and check names, dialogue, and transitions. For a short video narration or podcast intro, test the exact opening line and listen for pacing and pronunciation. A lower grade is not a reason to rule out a voice: use the sample and your target script to decide. If clarity changes with the text, review these Kokoro voice quality tips before exporting.
How to Use These Voices for Free
- Choose a prerecorded English sample above to compare voices without downloading the model.
- To make your own speech, open the Kokoro TTS online tool and paste or type your script.
- Choose one of the 28 available English presets in the selector and adjust speed for your audience.
- Generate and review the audio, then download the WAV when it sounds right.
No account is needed. The model runs on your device, so your text does not need to be sent to a speech-generation server. A model download is required only when generating custom speech; browser storage may keep model files available for a later visit.
What the Voice Grades Mean
The model author says each grade estimates the quality and quantity of that voice’s training data, which can affect inference quality. It is not a blind listening-test score or a promise that one voice will sound best to every person. Spanish and Brazilian Portuguese entries have no overall grade in the published table, so this page leaves them ungraded.
The upstream guidance says many voices perform best around 100–200 tokens. Very short input—especially under 10–20 tokens—may sound weaker; combine short lines for a fairer test. Long input above roughly 400 tokens may rush, so split it into sections or adjust speed. Read the full Kokoro voice grading and usage notes, and see our complete Kokoro TTS guide.
Kokoro-82M: The Model Behind These Voices
Kokoro v1.0 is an open-weight text-to-speech model with 82 million parameters and an Apache 2.0 license. Its voice list covers eight languages, with American and British English shown as separate accent groups. This catalog has prerecorded audio for the 28 English voices; the model itself includes 26 additional non-English presets, which are listed without audio here. Language support and training data vary: official notes describe Spanish, French, Hindi, Italian, and Portuguese eSpeak NG fallbacks for phonemization, and caution that non-English support can be thin.
Russian and Korean are not in the v1.0 voice list. A translated site page does not add a native Kokoro voice for that language, so we do not label these presets as Russian or Korean speech. This page previews the built-in voices; the tool does not offer voice cloning or custom voice creation.
Model details: official Kokoro-82M model card.
Kokoro Voices FAQ
How many voices does Kokoro v1.0 have?
Kokoro v1.0 lists 54 preset voices across eight languages. American and British English are shown as separate voice groups in this catalog.
Which languages are included, and which can I preview here?
The v1.0 list includes American and British English, Japanese, Mandarin Chinese, Spanish, French, Hindi, Italian, and Brazilian Portuguese. This page has prerecorded samples for its 28 English voices; the other 26 presets are listed for reference without sample audio. Russian and Korean are not in the v1.0 voice list.
Which Kokoro voices have the highest published grades?
The upstream voice list gives af_heart an overall grade of A, af_bella A-, and af_nicole and bf_emma B-. The grades estimate training-data quality and quantity; listeners may prefer a different voice.
What does a Kokoro voice grade mean?
The model author describes overall grades as estimates based on the quality of the reference voice and the amount of training audio. They are not a universal ranking of listener preference.
Does playing a sample upload my text?
The 28 English previews are prerecorded audio clips served by this site. Listening to them does not download the TTS model or upload any text. The model is downloaded only when you generate your own speech in the TTS tool.
Can I download a voice or create a custom voice here?
This page previews the built-in preset voices; it does not provide voice cloning or custom voice creation. To download generated speech, open the Kokoro Web tool, enter your own text, choose a preset voice, and export the WAV.
Does the Kokoro voice preview work offline?
The small sample files need an internet connection the first time they are played unless your browser already cached them. They do not need the TTS model. Generating your own speech still requires the model files.