SFX Gallery — integration guide

A catalogue of verified sound effects, priced for machine consumption. Discovery is free and unauthenticated; masters are paid per file in USDC on Base using x402. The point of the service is the metadata: every sound ships with measured acoustics, so an agent can place it on a cut without ever decoding the audio.

Endpoints

Payment is USDC on Base mainnet (eip155:8453), asset 0x833589fCD6eDb6E08f4c7C32D4f71b54bdA02913. The live pay_to address is published in the GET /api/sfx response — read it from there rather than hardcoding it.

Paying for a sound

Both paid routes follow the standard x402 handshake. Call the endpoint with no payment and you get 402 with a PAYMENT-REQUIRED header carrying a base64 challenge — scheme, network, amount, asset, and payTo. Settle it and repeat the request with the payment attached.

# 1. Unpaid — returns 402 plus the challenge
curl -X POST https://sentedge.ai/api/sfx/01_whoosh_soft/wav

# 2. Any x402-capable client handles the handshake for you, e.g.
#    npm i @x402/fetch

import { wrapFetchWithPayment } from '@x402/fetch';
import { createWalletClient, http } from 'viem';
import { privateKeyToAccount } from 'viem/accounts';
import { base } from 'viem/chains';

const account = privateKeyToAccount(process.env.PRIVATE_KEY);
const wallet  = createWalletClient({ account, chain: base, transport: http() });
const pay     = wrapFetchWithPayment(fetch, wallet);

const res   = await pay('https://sentedge.ai/api/sfx/01_whoosh_soft/wav', { method: 'POST' });
const sound = await res.json();

// sound.data is base64 WAV; sound.acoustic is the measured metadata
fs.writeFileSync(sound.filename, Buffer.from(sound.data, 'base64'));

The metadata is the product

Each sound carries a measured acoustic block. These are not tags someone typed — they are computed from the rendered audio, and each sound is checked against its declared intent before it enters the catalogue.

Filtering the catalogue

The full catalogue is 1000 sounds and about 635 KB. GET /api/sfx accepts filters so you can ask for the handful you want instead:

Filters combine with AND. The response reports count (returned), matched (before paging), catalogue_total, the query actually applied, and supported_filters — so one call tells you everything the endpoint can do.

# A short, bright one-shot that will not fight a voiceover
curl "https://sentedge.ai/api/sfx?occupies=high&masks_dialogue=false&max_duration_s=1&limit=5"

Choosing a sound without listening

A worked heuristic: for a hard cut needing emphasis that must not fight a voiceover, filter for align_on: "start", masks_dialogue: false, and a short attack_s. For a transition that lands on the cut rather than starting at it, take align_on: "peak" and offset your placement by attack_s.

Licence

Everything here is royalty-free. A paid wav gives you a perpetual, worldwide, non-exclusive licence to use the sound in unlimited projects, commercial or not, with no attribution required and nothing further to pay — including in works your own systems generate automatically.

The single restriction: do not redistribute the sound as a sound. Baked into a mix, a game build, a video or a UI is fine; shipping the WAV as a standalone file, in a sample pack, a sound library, or a machine-learning training dataset is not. Free preview MP3s are watermarked and are not licensed for use in finished work.

Every sound is generated, not recorded — Stable Audio output from a prompt and a fixed seed. There is no underlying sample, recording or performance owned by anyone else, so there is nothing to clear. Full terms, including the buyout position and the honest note on whether copyright subsists in AI-generated audio at all, are on the licence page.

What a buyout does and does not mean

A buyout delists one specific file and retires its generation recipe, so that file is never sold again. It does not mean no similar sound can exist — a comparable one could be generated from a different seed. We would rather state that plainly than sell an exclusivity we cannot enforce.

MCP

SentEdge also runs an MCP server at https://sentedge.ai/mcp (Streamable HTTP) for clients that prefer tool calls to raw HTTP.

Custom work, bundles and other enquiries

Not everything is self-serve. If you need sounds the catalogue does not have, we build custom sets to a brief: the brief becomes a fixed quote and a fixed delivery window, settled as a purchase order over the same x402 rails, then delivered from a private per-order URL. It runs in business days, not seconds, and the reasons why are written up on the commissions page.

There is deliberately no endpoint for commissions, and none is planned — a human has to accept the scope and stand behind the window. Email contactus@sentedge.ai for that, for bundle pricing on a large selection, for licensing questions, or for anything the API does not cover.

Browse and listen