Audio (TTS & STT)

Two audio endpoints: POST /v1/audio/speech converts text to speech (MP3 or PCM output depending on the model) and POST /v1/audio/transcriptions converts speech to text. Via multipart upload or JSON with base64 audio. Both are OpenAI-SDK compatible.

POST/v1/audio/speechBearer auth

Text-to-speech (TTS)

Convert text to spoken audio. Drop-in compatible with the OpenAI SDK's client.audio.speech.create(...). Billed per character of input, per second of generated audio, or both, as each model's page states; the charge settles when the audio has been sent.

Every text-to-speech model in GET /v1/models (api_shape speech) works here and on POST /v1/predictions. Voices are model-specific: each model's input_fields lists its own. response_format is mp3 or pcm (24 kHz mono); when omitted the audio is PCM, so send mp3 to get an MP3 file.

Request bodyrequired

Field	Type	Description
`model`required	string	Example: `x-ai/grok-voice-tts-1.0`
`input`required	string	Example: `Hello world`
`voice`required	string	Example: `eve`
`response_format`	string	Default: `pcm` Allowed values: `mp3` · `pcm`

Example

curlPythonJavaScriptGo

curl https://api.hormouz.net/v1/audio/speech \
  -H "Authorization: Bearer ng-YOUR_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "x-ai/grok-voice-tts-1.0",
    "input": "Welcome to Hormouz!",
    "voice": "eve"
  }' \
  --output speech.mp3

from openai import OpenAI

client = OpenAI(api_key="ng-YOUR_API_KEY", base_url="https://api.hormouz.net/v1")

with client.audio.speech.with_streaming_response.create(
    model="x-ai/grok-voice-tts-1.0",
    input="Welcome to Hormouz!",
    voice="eve",
) as resp:
    resp.stream_to_file("speech.mp3")

import OpenAI from "openai";
import { writeFile } from "node:fs/promises";

const client = new OpenAI({ apiKey: "ng-YOUR_API_KEY", baseURL: "https://api.hormouz.net/v1" });

const resp = await client.audio.speech.create({
  model: "x-ai/grok-voice-tts-1.0",
  input: "Welcome to Hormouz!",
  voice: "eve",
});
await writeFile("speech.mp3", Buffer.from(await resp.arrayBuffer()));

package main

import (
	"bytes"
	"encoding/json"
	"io"
	"net/http"
	"os"
)

func main() {
	body, _ := json.Marshal(map[string]any{
		"model": "x-ai/grok-voice-tts-1.0",
		"input": "Welcome to Hormouz!",
		"voice": "eve",
	})
	req, _ := http.NewRequest("POST", "https://api.hormouz.net/v1/audio/speech", bytes.NewReader(body))
	req.Header.Set("Authorization", "Bearer ng-YOUR_API_KEY")
	req.Header.Set("Content-Type", "application/json")

	resp, err := http.DefaultClient.Do(req)
	if err != nil {
		panic(err)
	}
	defer resp.Body.Close()

	f, _ := os.Create("speech.mp3")
	defer f.Close()
	io.Copy(f, resp.Body)
}

Responses

200

Audio bytestream.

POST/v1/audio/transcriptionsBearer auth

Speech-to-text (STT)

Transcribe audio to text. Drop-in compatible with the OpenAI SDK's client.audio.transcriptions.create(...) (multipart file upload). You may also send the JSON shape ({model, input_audio:{data:<base64>, format}}) directly. Models: every speech-to-text model in GET /v1/models (api_shape: transcriptions), e.g. openai/whisper-1, openai/gpt-4o-transcribe, elevenlabs/scribe-v2, microsoft/mai-transcribe-2. OpenAI's bare ids (whisper-1) work too. An unknown model is 404 MODEL_NOT_FOUND.

Price. Per minute of audio, or per token on the token-priced models (gpt-4o-transcribe, gpt-4o-mini-transcribe, gemini-3.5-transcribe); the rate is on each model's page. We measure the audio's length before reserving credit, so the reserve matches the recording; you are charged the reported usage.cost.

Inline audio only. Send the file (multipart file) or base64 input_audio.data; pcm is raw 16-bit mono at 16 kHz. A URL (input_audio.url) is refused here: send it to POST /v1/predictions (input.audio), which also suits long recordings. Audio whose length cannot be read is 400.

Some models cap audio duration: google/chirp-3 accepts at most 60 seconds per request and answers anything longer with 400 AUDIO_DURATION_LIMIT, before anything is reserved. Split long recordings, or use openai/whisper-large-v3 / openai/gpt-4o-transcribe, which take multi-minute audio in one call.

diarize: true (with response_format: verbose_json) labels speakers on the models that support it (ElevenLabs Scribe); others answer 400. For Persian on Deepgram Nova-3, set language: "fa".

Request body`multipart/form-data`required

Field	Type	Description
`file`required	string (binary)
`model`required	string	Example: `openai/gpt-4o-mini-transcribe`
`response_format`	string
`language`	string

Request body`application/json`required

Field	Type	Description
`model`required	string	Example: `openai/gpt-4o-mini-transcribe`
`input_audio`required	object
`data`required	string	base64-encoded audio
`format`required	string	Allowed values: `mp3` · `wav` · `ogg` · `flac` · `webm` · `m4a` · `pcm`
`language`	string	ISO-639-1, e.g. fa; detected when omitted
`response_format`	string	Allowed values: `json` · `verbose_json`
`diarize`	boolean	label speakers (verbose_json, supported models)

Example

curlPythonJavaScriptGo

curl https://api.hormouz.net/v1/audio/transcriptions \
  -H "Authorization: Bearer ng-YOUR_API_KEY" \
  -F model="openai/gpt-4o-mini-transcribe" \
  -F file=@recording.mp3

from openai import OpenAI

client = OpenAI(api_key="ng-YOUR_API_KEY", base_url="https://api.hormouz.net/v1")

with open("recording.mp3", "rb") as f:
    resp = client.audio.transcriptions.create(
        model="openai/gpt-4o-mini-transcribe",
        file=f,
    )
print(resp.text)

import OpenAI from "openai";
import { createReadStream } from "node:fs";

const client = new OpenAI({ apiKey: "ng-YOUR_API_KEY", baseURL: "https://api.hormouz.net/v1" });

const resp = await client.audio.transcriptions.create({
  model: "openai/gpt-4o-mini-transcribe",
  file: createReadStream("recording.mp3"),
});
console.log(resp.text);

package main

import (
	"bytes"
	"fmt"
	"io"
	"mime/multipart"
	"net/http"
	"os"
)

func main() {
	var buf bytes.Buffer
	w := multipart.NewWriter(&buf)
	w.WriteField("model", "openai/gpt-4o-mini-transcribe")
	part, _ := w.CreateFormFile("file", "recording.mp3")
	f, _ := os.Open("recording.mp3")
	defer f.Close()
	io.Copy(part, f)
	w.Close()

	req, _ := http.NewRequest("POST", "https://api.hormouz.net/v1/audio/transcriptions", &buf)
	req.Header.Set("Authorization", "Bearer ng-YOUR_API_KEY")
	req.Header.Set("Content-Type", w.FormDataContentType())

	resp, err := http.DefaultClient.Do(req)
	if err != nil {
		panic(err)
	}
	defer resp.Body.Close()
	raw, _ := io.ReadAll(resp.Body)
	fmt.Println(string(raw))
}

Responses

200

Transcription result.

Audio (TTS & STT)

Text-to-speech (TTS)

Request bodyrequired

Example

Responses

Speech-to-text (STT)

Request bodymultipart/form-datarequired

Request bodyapplication/jsonrequired

Example

Responses

Request body`multipart/form-data`required

Request body`application/json`required