Kokoro TTS API — voices, pricing & examples

Q: Do OpenAI voice names work?

Yes. alloy, echo, fable, onyx and nova map to the Kokoro voices of the same name (af_alloy, am_echo, bm_fable, am_onyx, af_nova); shimmer, ash, ballad, coral, sage and verse map to the closest character. The response header X-Voice-Mapped says which voice was used. They are Kokoro voices, not OpenAI's, so they sound different.

Models/Speech/Kokoro 82M

Text to speech54 voicesMP3 / WAV8 languages

Modality

Text → Audio

Price

$5.20 / 1M characters

300 chars → audio

0.84 s

Model ID

hexgrad/Kokoro-82M

Quickstart

You need an API key to call this model. Create a free account or log in.

Set INFERENCE_API_KEY to your key from the API Keys page, then run:

curl https://api.inferenceapis.com/v1/audio/speech \
  -H "Authorization: Bearer $INFERENCE_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "hexgrad/Kokoro-82M",
    "input": "Hello! This is Kokoro 82M speaking.",
    "voice": "af_heart",
    "response_format": "mp3"
  }' --output speech.mp3

import os
from openai import OpenAI

client = OpenAI(
    base_url="https://api.inferenceapis.com/v1",
    api_key=os.environ["INFERENCE_API_KEY"],
)
with client.audio.speech.with_streaming_response.create(
    model="hexgrad/Kokoro-82M",
    voice="af_heart",
    input="Hello! This is Kokoro 82M speaking.",
    response_format="mp3",
) as response:
    response.stream_to_file("speech.mp3")

import fs from "node:fs";
import OpenAI from "openai";

const client = new OpenAI({
  baseURL: "https://api.inferenceapis.com/v1",
  apiKey: process.env.INFERENCE_API_KEY,
});
const audio = await client.audio.speech.create({
  model: "hexgrad/Kokoro-82M",
  voice: "af_heart",
  input: "Hello! This is Kokoro 82M speaking.",
  response_format: "mp3",
});
fs.writeFileSync("speech.mp3", Buffer.from(await audio.arrayBuffer()));

When to use it

A good fit when

Interactive speech: sub-second responses for a paragraph in our measurement
You need voices in several languages at a low price per character

Look elsewhere when

You need emotional delivery or non-verbal sounds such as laughter

What it costs in practice

Workload	Cost
One 1,000-word article (about 6,000 characters)	$0.03
One hour of narration (about 54,000 characters)	$0.28
One million characters (about 18 hours of speech)	$5.20

Computed from the live rate below. There is no subscription or minimum; new accounts start with free credit.

Endpoint

POST https://api.inferenceapis.com/v1/audio/speech

Headers Authorization: Bearer $INFERENCE_API_KEY
Content-Type: application/json

Compatible OpenAI wire format. /openai/v1/… is accepted too, so Groq-style base URLs work with only the host changed.

Model IDs and aliases

Send any of these as model; they all resolve to this model and bill at its rate.

ID	Note
`hexgrad/Kokoro-82M`	Canonical
`kokoro`	Alias
`kokoro-82m`	Alias
`tts-1`	OpenAI's id. OpenAI voice names are mapped too, so OpenAI speech code runs with only the base URL changed

Request body

Parameter	Type	Description
`model` required	string	Model ID: `hexgrad/Kokoro-82M`
`input` required	string	The text to speak. Up to 20,000 characters per request.
`voice`	string	Voice id. Default `af_heart`. See the voice list below.
`response_format`	string	`mp3` (default), `wav` or `raw` (PCM).
`speed`	number	Playback speed multiplier, e.g. `1.1`.
`language`	string	Language hint for multilingual voices, e.g. `en`, `ja`.

Voices

Click a voice to copy its id. Hear them in the playground.

Response

200 OK · audio

Binary audio (Content-Type: audio/mpeg for mp3, audio/wav for wav).
Headers:
  X-Usage-Characters: 42        # characters billed for this request

Pricing

Price	$5.20 / 1M characters
Billing	Pay as you go from prepaid credits. No subscription, no daily request or token cap. See all pricing.

Notes

Voice ids are prefixed by language and gender: af_ American female, am_ American male, bf_/bm_ British, jf_/jm_ Japanese, zf_/zm_ Mandarin, ef_/em_ Spanish, ff_ French, hf_/hm_ Hindi, if_/im_ Italian, pf_/pm_ Portuguese.
Billing is per character of input text; the X-Usage-Characters response header reports the metered count.

Compare Kokoro 82M with…

Orpheus 3B OpenAI TTS (tts-1)OpenAI gpt-4o-mini-tts

Languages

Per-language voice lists and samples: text-to-speech languages.

Questions about Kokoro 82M

Do OpenAI voice names work?

Yes. alloy, echo, fable, onyx and nova map to the Kokoro voices of the same name (af_alloy, am_echo, bm_fable, am_onyx, af_nova); shimmer, ash, ballad, coral, sage and verse map to the closest character. The response header X-Voice-Mapped says which voice was used. They are Kokoro voices, not OpenAI's, so they sound different.

How long can the input be?

20,000 characters per request. For longer text, split on sentence boundaries, request each part and concatenate the audio; mp3 parts can be joined as they are.

Does it support SSML?

No. Pacing follows punctuation: commas and full stops produce pauses, and a new paragraph produces a longer one. speed changes the overall rate.

Is the audio streamed?

No, the response is the complete file. A paragraph returns in under a second in our measurement, so generating sentence by sentence works for interactive use.

Which languages sound best?

English. Kokoro's authors grade its own voices and the American and British English ones rate highest; the other languages were trained on far less data. Listen before you commit: voices and samples by language.

Can I use the audio commercially?

Kokoro is released under Apache 2.0, which permits commercial use of the model and its output.

Licence

Kokoro 82M is published by Hexgrad under the Apache 2.0. Your use of its output is also subject to that licence.

Errors

Errors use the OpenAI envelope: {"error": {"message", "type", "code"}}. 401 missing or invalid key · 402 insufficient_balance · 404 model_not_found · 503 model_unavailable · 502 backend error, safe to retry. Full table with what to retry: API documentation. Errors from other providers: provider error reference.

Kokoro 82M API