Same /v1/transcriptions/jobs API shape, same $0.005/min list price, but no prepaid credits and no wallet-auth headers: each file is paid by a standard x402 402 → PAYMENT-SIGNATURE retry (USDC on Base). Machine-readable version: /migrate/dtelecom.md.
- BASE = "https://x402.dtelecom.org"
+ BASE = "https://speech.moneytools.dev"
- await gw.buyCredits({ amountUsd: 0.10 }) // prepaid credits: gone
- headers: Authorization/X-Wallet-Address/X-Wallet-Chain/X-Timestamp // wallet-auth headers: gone
+ payFetch = wrapFetchWithPayment(fetch, x402Client) // pays the exact 402 for each file
POST /v1/transcriptions/jobs multipart: file, language, diarize + Idempotency-Key (unchanged)
GET /v1/transcriptions/jobs/{id} (unchanged path)
poll auth: wallet signature headers from the paying wallet (still accepted, EVM)
+ or: Authorization: Bearer <job_token> (returned once in the 202)
DELETE /v1/transcriptions/jobs/{id} (unchanged)
- import { DtelecomGateway } from "@dtelecom/x402-client";
- const gw = new DtelecomGateway({ account });
- await gw.buyCredits({ amountUsd: 0.10 });
- const job = await gw.createFileTranscription({ file, idempotencyKey, language: "en" });
- const done = await gw.getFileTranscription(job.id);
+ import { wrapFetchWithPayment, x402Client } from "@x402/fetch";
+ import { ExactEvmScheme } from "@x402/evm";
+ const payFetch = wrapFetchWithPayment(fetch, new x402Client().register("eip155:8453", new ExactEvmScheme(account)));
+ const form = new FormData(); form.append("file", file, "meeting.mp3"); form.append("language", "en");
+ const job = await (await payFetch("https://speech.moneytools.dev/v1/transcriptions/jobs",
+ { method: "POST", headers: { "Idempotency-Key": idempotencyKey }, body: form })).json(); // 202 {id, job_token, status_url}
+ const done = await (await fetch(job.status_url, { headers: { Authorization: `Bearer ${job.job_token}` } })).json();
| dTelecom | MoneyTools Transcripts | |
|---|---|---|
| Price | $0.005 per decoded audio minute (seconds rounded up), from prepaid credits | $0.005 per started audio minute, paid per file — no credits, no minimum |
| Payment | buy credits first (min $0.10), then wallet-signed requests | x402 v2 per request (USDC on Base, CDP facilitator); any x402 client |
| Auth headers | Authorization + X-Wallet-Address + X-Wallet-Chain + X-Timestamp | none to submit; poll with Bearer job_token or the same EVM wallet-auth headers from the paying wallet |
| Endpoint / body | POST /v1/transcriptions/jobs, multipart file, language, diarize, Idempotency-Key | same, plus JSON {"url"} import, optional webhook and summary |
| Max file | 256 MiB / 2 h | 100 MB / 3 h (MP3/WAV segmented; other containers up to 10 min) |
| Language | default en, auto not supported | auto-detect by default, or pass a Whisper code |
| Speaker diarization | included | NOT available (speaker:null, speaker_segments:[]) — stay on dTelecom if you need speakers |
| Outputs | text, segments, media | same shape + SRT, VTT, TXT (?format=), optional summary |
| Bad files | hold released | first ~5 min segment is transcribed before settlement: unreadable files are never charged |
| Retention | 24 h | 24 h |
| Capacity | admission-bounded (429) | shared daily audio-minute budget (live in https://speech.moneytools.dev/health → usage); paid jobs over the budget wait and resume after 00:00 UTC |
Rounding: we bill per started minute (61 s = 2 min = 0.004 USDC). Free price check: GET https://speech.moneytools.dev/v1/speech/quote?minutes=60; the unpaid POST also returns the exact price for your file before you pay.
# 1) unpaid upload -> HTTP 402 + PAYMENT-REQUIRED (exact price for THIS file) + {id, quote}
curl -i -X POST https://speech.moneytools.dev/v1/transcriptions/jobs \
-H 'Idempotency-Key: meeting-42' -F file=@meeting.mp3 -F language=en
# 2) sign the PAYMENT-REQUIRED header locally (nothing is sent; refuses above MAX_USD)
curl -sO https://speech.moneytools.dev/examples/x402_sign.py && pip install "x402[evm]" eth-account
SIG=$(X402_PRIVATE_KEY=0x... python3 x402_sign.py "<PAYMENT-REQUIRED header value>")
# 3) same request, same Idempotency-Key (stored upload reused) + PAYMENT-SIGNATURE -> 202 {id, job_token, status_url}
curl -X POST https://speech.moneytools.dev/v1/transcriptions/jobs -H 'Idempotency-Key: meeting-42' \
-H "PAYMENT-SIGNATURE: $SIG" -F file=@meeting.mp3 -F language=en
# 4) poll (every ~15 s) and fetch subtitles
curl -H "Authorization: Bearer $JOB_TOKEN" https://speech.moneytools.dev/v1/transcriptions/jobs/$ID
curl -H "Authorization: Bearer $JOB_TOKEN" "https://speech.moneytools.dev/v1/transcriptions/jobs/$ID?format=srt"
# whole flow as one script: curl -sO https://speech.moneytools.dev/examples/transcribe.sh && bash transcribe.sh meeting.mp3 srt
# pip install "x402[requests,evm]" eth-account
import uuid, time, requests
from eth_account import Account
from x402 import x402ClientSync, max_amount
from x402.http.clients import x402_requests
from x402.mechanisms.evm import EthAccountSigner
from x402.mechanisms.evm.exact import ExactEvmScheme
client = x402ClientSync()
client.register("eip155:8453", ExactEvmScheme(signer=EthAccountSigner(Account.from_key(PRIVATE_KEY))))
client.register_policy(max_amount(500_000)) # never pay more than 0.50 USDC per file
s = x402_requests(client) # pays the 402 and resends the same request
r = s.post("https://speech.moneytools.dev/v1/transcriptions/jobs",
headers={"Idempotency-Key": f"py-{uuid.uuid4().hex}"},
files={"file": ("meeting.mp3", open("meeting.mp3", "rb").read())}, data={"language": "en"})
job = r.json() # 202 {id, status:"running", job_token, status_url, payment}
auth = {"Authorization": f"Bearer {job['job_token']}"}
while (st := requests.get(job["status_url"], headers=auth).json())["status"] not in ("completed", "failed"):
time.sleep(15)
print(st["result"]["text"]) # also: result.segments, result.srt, result.vtt
# full script: https://speech.moneytools.dev/examples/transcribe.py
// npm i @x402/fetch @x402/evm viem
import { wrapFetchWithPayment, x402Client } from "@x402/fetch";
import { ExactEvmScheme } from "@x402/evm";
import { privateKeyToAccount } from "viem/accounts";
import { readFile } from "node:fs/promises";
const client = new x402Client()
.register("eip155:8453", new ExactEvmScheme(privateKeyToAccount(process.env.X402_PRIVATE_KEY as `0x${string}`)))
.registerPolicy((_v, reqs) => reqs.filter((r) => BigInt(r.amount) <= 500_000n)); // cap 0.50 USDC per file
const payFetch = wrapFetchWithPayment(fetch, client);
const form = new FormData();
form.append("file", new Blob([await readFile("meeting.mp3")]), "meeting.mp3");
form.append("language", "en");
const res = await payFetch("https://speech.moneytools.dev/v1/transcriptions/jobs", {
method: "POST", headers: { "Idempotency-Key": `ts-${crypto.randomUUID()}` }, body: form,
});
const job = await res.json(); // 202 {id, job_token, status_url, payment:{amountUsd, transaction}}
const auth = { Authorization: `Bearer ${job.job_token}` };
let st;
do { await new Promise((r) => setTimeout(r, 15000)); st = await (await fetch(job.status_url, { headers: auth })).json(); }
while (st.status !== "completed" && st.status !== "failed");
console.log(st.result.text); // also st.result.segments / srt / vtt
// full script: https://speech.moneytools.dev/examples/transcribe.ts
import { DtelecomGateway } from "@dtelecom/x402-client"; // unchanged
import { installTranscriptsFetch, TRANSCRIPTS_URL } from "@moneytools-dev/transcripts";
import { privateKeyToAccount } from "viem/accounts";
const account = privateKeyToAccount(process.env.PRIVATE_KEY);
installTranscriptsFetch(account, { maxUsdPerRequest: 0.5 }); // + 1 line: pays the 402 of each file (only for https://speech.moneytools.dev)
const gw = new DtelecomGateway({ account, gatewayUrl: TRANSCRIPTS_URL }); // + base URL
// no gw.buyCredits(...)
const job = await gw.createFileTranscription({ file, filename: "meeting.mp3", idempotencyKey: "meeting-42", language: "en" });
const done = await gw.getFileTranscription(job.id); // the client's own wallet-auth headers work for polling
console.log(done.status, done.result?.text);
Tested against @dtelecom/x402-client 0.6.2: create → poll (its own EVM wallet-auth headers are accepted from the wallet that paid) → delete. Solana wallet auth is not supported.
import { TranscriptsGateway } from "@moneytools-dev/transcripts"; // same method names as DtelecomGateway
const tc = new TranscriptsGateway({ account, maxUsdPerRequest: 0.5 });
const job = await tc.createFileTranscription({ file, idempotencyKey: "meeting-42", language: "en" });
const done = await tc.waitForFileTranscription(job.id);
const srt = await tc.getFileTranscriptionText(job.id, "srt");from eth_account import Account
from moneytools_transcripts import TranscriptsClient # pip install moneytools-transcripts (not published yet; see below)
tc = TranscriptsClient(Account.from_key(PRIVATE_KEY), max_usd_per_request=0.5)
job = tc.create_file_transcription(open("meeting.mp3", "rb"), idempotency_key="meeting-42", language="en")
done = tc.wait_for_file_transcription(job["id"]) # Bearer job_token, or wallet auth from the paying wallet
print(done["result"]["text"])TypeScript: npm i @moneytools-dev/transcripts. The PyPI package moneytools-transcripts is built and tested but not published yet; until then use the copy-paste Python snippet above or the raw x402 examples below. Any x402 client works.
| Sample (public domain) | Length | Segments | Ref. words | WER | Processing | × real time |
|---|---|---|---|---|---|---|
| Ozymandias (Shelley), LibriVox | 1.5 min | 1 | 111 | 0.9% | 2.8 s | 0.032 |
| Gettysburg Address, LibriVox | 2.7 min | 1 | 268 | 1.9% | 6.2 s | 0.039 |
| The Raven (Poe), LibriVox | 9.5 min | 2 | 1090 | 5.4% | 31.3 s | 0.055 |
Word error rate after lower-casing and stripping punctuation, aligned so the LibriVox intro/outro (not in the reference text) is not counted. Reference texts: Project Gutenberg #4 and #1065, standard 1818 Ozymandias text. Readers may follow a slightly different edition, and archaic spellings count as errors (e.g. 'visiter' → 'visitor', 'Pallas' → 'palace'), so these are sanity numbers, not a leaderboard.
Processing time = paid submit → last segment done, measured with our engine code on our build box calling Workers AI remotely (adds network hops; in production the worker calls the model from inside Cloudflare). Segments of one job run one after another, so long files take roughly (minutes ÷ 5) × 8–16 s.
3 clean single-speaker English recordings only. No diarization. Noisy, multi-speaker or non-English audio will score worse. We have not benchmarked dTelecom and make no claims about its quality or speed.
Links: llms.txt · OpenAPI · .well-known/x402 · MCP https://speech.moneytools.dev/mcp · health / daily capacity