Speech to Text (Whisper Transcription)

Speech to text for agents: transcribe a voice note, meeting, podcast or call recording into text with timestamps, in about 100 languages with automatic language detection. Runs OpenAI Whisper (small by default, base for speed) on agentsvc.io itself, the audio is deleted after the call. Send audio_url or audio_base64 (mp3, mp4, m4a, wav, ogg, webm, flac), up to 10 minutes and 24 MB per call. Returns text, detected language, duration and segments with start and end seconds. Also available OpenAI-compatible at POST /v1/audio/transcriptions (multipart file, model whisper-1).

Run free trial ↗3 free calls per day with the example input. Paid: $0.010 USDC, no limit.

Call it

import { wrapFetchWithPayment, x402Client } from "@x402/fetch";
import { registerExactEvmScheme } from "@x402/evm/exact/client";
import { privateKeyToAccount } from "viem/accounts";

const client = new x402Client();
registerExactEvmScheme(client, { signer: privateKeyToAccount(process.env.EVM_PRIVATE_KEY) });
const payFetch = wrapFetchWithPayment(fetch, client);

const res = await payFetch("https://agentsvc.io/api/v1/proxy/audio-transcribe", {
  method: "POST",
  headers: { "Content-Type": "application/json" },
  body: JSON.stringify({"audio_url":"https://agentsvc.io/samples/speech-en.mp3"}),
});
const { data, payment } = await res.json();  // payment.status "settling"; header X-Payment-Settle: sync returns payment.transaction

Input

Field	Type	Description
audio_url	string	Public URL of the audio file
audio_base64	string	Audio file as base64; or use audio_url
language	string	ISO 639-1 code like en or de, default auto-detect
model	small \| base = "small"	small is more accurate, base about twice as fast
timestamps	boolean = true	Include segments with start and end seconds

Output

Field	Type	Description
text	string
language	string
language_probability	number
duration_seconds	number
model	string
processing_ms	integer
segments	array
model_license	string

Example response (data)

{
  "text": "This is a sample recording for the agents.io speech to text service. It turns spoken audio into text with time stamps in about 100 languages.",
  "language": "en",
  "language_probability": 0.998,
  "duration_seconds": 9.17,
  "model": "whisper-small",
  "processing_ms": 3688,
  "segments": [
    {
      "start": 0,
      "end": 4.4,
      "text": "This is a sample recording for the agents.io speech to text service."
    },
    {
      "start": 4.4,
      "end": 9,
      "text": "It turns spoken audio into text with time stamps in about 100 languages."
    }
  ],
  "model_license": "MIT (OpenAI Whisper weights, faster-whisper runtime), runs on agentsvc.io, audio is deleted after the call"
}