Speech to Text API

Upload one audio file and receive a completed transcript as JSON or plain text. The multipart fields and response use a familiar OpenAI transcription shape, but the endpoint is different, so changing only an OpenAI SDK Base URL is not supported.

API overview

Item	Value
Base URL	https://api.kitschlabs.com
Endpoint	POST /v1/speech-to-text
Authentication	Authorization: Bearer <API_KEY> or xi-api-key: <API_KEY>
Request	multipart/form-data
Model	kitsch-stt-v1
Maximum file size	100 MiB
Default response	JSON

Speech to Text is enabled by default for every authenticated account.

How do I transcribe a file?

cURL

curl --request POST \
  "https://api.kitschlabs.com/v1/speech-to-text" \
  --header "Authorization: Bearer <API_KEY>" \
  --form "file=@speech.wav" \
  --form "model=kitsch-stt-v1" \
  --form "language=en" \
  --form "response_format=json"

Python

import os

import requests

with open("speech.wav", "rb") as audio:
    response = requests.post(
        "https://api.kitschlabs.com/v1/speech-to-text",
        headers={"Authorization": f"Bearer {os.environ['KITSCH_API_KEY']}"},
        data={
            "model": "kitsch-stt-v1",
            "language": "en",
            "response_format": "json",
        },
        files={"file": audio},
        timeout=120,
    )

response.raise_for_status()
print(response.json()["text"])

JavaScript

import { readFile } from "node:fs/promises";

const form = new FormData();
form.append("file", new Blob([await readFile("speech.wav")]), "speech.wav");
form.append("model", "kitsch-stt-v1");
form.append("language", "en");
form.append("response_format", "json");

const response = await fetch("https://api.kitschlabs.com/v1/speech-to-text", {
  method: "POST",
  headers: { Authorization: `Bearer ${process.env.KITSCH_API_KEY}` },
  body: form,
});

if (!response.ok) {
  throw new Error(`${response.status}: ${await response.text()}`);
}

console.log((await response.json()).text);

What fields can I send?

Field	Type	Required	Description
file	file	Yes	Audio file to transcribe, up to 100 MiB
model	string	Yes	kitsch-stt-v1
language	string	No	Expected ISO 639-1 language code, such as ko, en, or ja
prompt	string	No	Short hint for proper nouns or context
response_format	string	No	json (default) or text
stream	boolean	No	Only false is supported
kitsch_options	JSON string	No	Advanced recognition options

temperature, stream=true, other model IDs, and unsupported response formats return an error. To send audio as it is produced, use the Streaming API.

Advanced options

Send kitsch_options as a JSON string. Supported keys are:

• language_hints, language_hints_strict
• enable_speaker_diarization
• enable_language_identification
• context
• translation

--form 'kitsch_options={"language_hints":["ko","en"],"enable_speaker_diarization":true}'

Unknown keys return an error.

What does the response look like?

Response with response_format=json:

{
  "text": "Hello.",
  "usage": {
    "type": "duration",
    "seconds": 1.25
  }
}

With response_format=text, the API returns UTF-8 plain text. Record the x-kitsch-request-id response header for support and incident investigation.

How is usage billed?

Successful requests are billed at 0.1 credits per started second of input audio. For example, 4.2 and 5.0 seconds each cost 0.5 credits, while 5.1 seconds costs 0.6 credits. Failed requests and requests rejected during input validation are not charged.

How should I handle errors?

The error body includes error.message, error.type, error.param, and error.code.

{
  "error": {
    "message": "The uploaded file is too large.",
    "type": "invalid_request_error",
    "param": "file",
    "code": "file_too_large"
  }
}

Status	Meaning	Recommended action
400 or 422	Invalid file, model, or option	Check the fields and audio file
401	Missing or invalid API key	Check the authentication header
402	Insufficient credits	Check the balance and billing status
413	File exceeds the size limit	Reduce the file to 100 MiB or less
429	Request or processing capacity limit	Follow Retry-After and apply backoff
503	Temporarily unavailable	Record the request ID and retry selectively