Commission a custom set

The catalogue is what already exists. If you need sounds that do not — a set built to a brief, or one consistent house style across a whole product — we build them to order.

This is not a self-serve endpoint, and it is not real time. It is a purchase order with a fixed price and a fixed delivery window, agreed before you pay anything. Below is exactly how it runs, what it costs you in time, and what we will turn down.

Start here: email contactus@sentedge.ai with the brief. Everything else on this page is detail.

The process

Stage	What happens	Time
1. Brief	You email what you need. Free, no commitment.	—
2. Quote	We reply with a fixed price, a fixed sound-by-sound scope, and a delivery window. Or we decline, and say why.	Within 2 business days
3. Purchase order	You accept. We issue a per-order payment URL; you settle it in USDC over x402, same rails as the catalogue.	Your call
4. Build	The set is generated, mastered and verified.	The delivery window, which starts at settlement
5. Delivery	You collect the full set of master WAVs, each with its measured acoustics, from a private per-order URL.	—
6. Revision	One re-roll round on any sound that misses its stated intent, free.	Request within 7 days

Nothing is charged until stage 3, and the price at stage 3 is the price. If the brief changes after that, it is a new quote rather than a surprise on the invoice.

Delivery windows

Quoted per order and counted in business days from settlement, not hours. As a guide: a set of up to 30 sounds is normally 5 business days. Larger sets are quoted against the same shape. We do not offer a rush tier, for the honest reason given below — we would rather quote a window we always hit than sell a fast one we sometimes miss.

What to put in the brief

The more of this you bring, the tighter the quote:

How many sounds, and what each one is for — "confirmation of a destructive action", not "a beep". Intent is what we build against and what we verify.
Where they will sit: under dialogue, in a game mix, alone in a UI. This decides whether they must stay out of the speech band.
Rough duration for each, and whether it is a one-shot or needs to loop.
House style, if the set must feel like one family — warm and analogue, clinical and digital, physical and dry.
A reference from the catalogue, if anything in it is close. This is by far the most useful thing you can give us: those sounds are already measured, so "like 04_pop_ui but softer and half a second longer" is an unambiguous brief.

You do not need to write prompts. Prompt authoring is the work you are paying for.

What we will turn down

We will say no at the quote stage rather than take money and disappoint you. The generator produces synthesised audio from a text description, so it cannot do:

Speech, singing, or anything with words. No voice work of any kind.
Real-world field recordings — a specific location, a specific machine, a specific vehicle. We can produce something that reads as a diesel engine; we cannot produce your diesel engine.
Licensed or recognisable music, or anything designed to resemble an existing branded sound, chime or motif. That is someone else's mark, and generating a lookalike does not change that.
Sound tied to a real, identifiable person, including impressions.

Everything else — impacts, UI, whooshes, risers, ambiences, mechanical and material sounds, stingers, textures — is fair game.

Why it takes days rather than seconds

Worth being straight about, because everything else here is instant and this is not.

The catalogue is served from a droplet on the public internet, but nothing is generated there. Generation runs on a dedicated machine on our own network, which pulls queued work rather than accepting connections, and publishes finished audio to the same private store the shop already reads from. That store is the only seam between the two. It is deliberately one-directional: there is no path from the internet into the machine that makes the sounds.

That architecture is the reason the shop never goes down and also the reason a commission is not instant — your order joins a real queue on a real machine.

The bigger constraint is quality control, and it is not machine time. Every sound is checked automatically against the envelope and tonal behaviour its brief declared, and anything that misses is re-rolled or re-EQ'd and checked again. Across recent batches roughly one sound in five fails that check on the first pass. Building a set means absorbing that rework before you ever see it, which is precisely the work you are buying and precisely why a window is measured in days.

Pricing, licensing and exclusivity

Commissions are quoted per brief, not per file. Catalogue sounds are $0.02 because they already exist; a commission prices the authoring, generation, and verification of sounds that do not.

Commissioned sets are exclusive by default: they are delivered to you and do not enter the public catalogue. If you do not need exclusivity, say so — a set we may also list is quoted lower.

Delivered sounds carry the same royalty-free terms as everything else here — see the licence — with the redistribution restriction lifted for the exclusive set you commissioned, since there is no other buyer to protect.

If you are an agent

There is deliberately no endpoint for this. A commission needs a human to accept the scope, quote it, and stand behind a delivery window, and none of that should be transacted by a bot on either side.

If you are acting for someone and they need sounds the catalogue does not have, surface contactus@sentedge.ai and this page to them rather than attempting to negotiate. For anything that is self-serve, the catalogue and both paid endpoints are documented in the integration guide.

Browse and listen