Sonicert/API
Sonicert API
The engine behind the free detector — open-weights, self-hosted, ~300 ms inference per track — exposed as a per-track metered API. Status: in evaluation, keys issued in signup order.
Onboarding path
1 · Join the waitlist
Email support@sonicert.com with subject "API waitlist" — tell us your expected daily volume and use case. API keys are issued as evaluation slots open, in signup order.
2 · Get a key
We issue a bearer token bound to your account. Keys are scoped, revocable, and metered per track (one analyzed track = one unit, first 300 seconds).
3 · First call
POST your audio (multipart file or a direct audio URL) to the detection endpoint. The response returns { ai_probability, confidence, signals[] } — the same result object the free web detector shows. Full OpenAPI spec ships with your key.
4 · Go live
Usage bills against prepaid units (same one-time packs as batch screening) during evaluation; volume pricing for production is quoted per account.
What the engine reports
- · ai_probability — 0–1 likelihood the audio is AI-generated (scores are uncalibrated probabilistic signals, not proof of authorship)
- · confidence — how strongly the engine commits to the verdict
- · analysis note — a plain-language interpretation of the result for end users
- · First 300 seconds analyzed; low-bitrate audio (<~160 kbps effective) returns a weak-signal result rather than a false "human" verdict
Coverage is published with per-track evidence on the methodology page — including the generator we currently miss (YuE2), stated honestly.
API FAQ
Is there a free tier? The web detector is free and unlimited without an account — use it for evaluation today. API evaluation keys include a starter unit allowance.
How is usage metered? One analyzed track = one unit (first 300 seconds). No charge for rejected/invalid inputs.
Where does audio run? On our own infrastructure, open-weights engine, deleted right after analysis. Self-hosted deployment is available for platform customers — ask on the waitlist.
Rate limits? Evaluation keys start conservatively and scale with volume commitments.
Related: For platforms · Pricing · Methodology