Short version: there is no public voice endpoint on the TextSight API today. The public API covers text — detection, scoring and rewriting — and voice detection currently runs only through the web tool. Rather than publish a page implying otherwise and collect sign-ups for something that does not exist, here is exactly what is available, what is not, and how to reach us if you have a volume use case.
Plainly, so you can decide in thirty seconds whether to keep reading.
POST /v2/detect — AI detection on text, with per-sentence scoresPOST /v2/score — the same engine, scores only, lighter payloadPOST /v2/rewrite — humanize textVoice detection is available through the web tool: upload MP3, WAV, M4A, OGG or FLAC up to 10 MB and get an AI-likelihood score and a human-versus-synthetic verdict. Three checks a day, free, whether or not you are signed in. Voice has its own plans, separate from our text subscriptions — Echo $9/mo for 100 checks a month, Amplify $29 for 1,000, Broadcast $99 for 10,000 — plus one-time Soundbites credit packs from $5. That is a manual workflow, not something you can put a queue through.
If text detection is also useful to you, this part is live and documented.
Base URL is https://api.textsight.ai/v2. Authenticate with an API
key — either an Authorization: Bearer header or an
x-api-key header. Live keys are prefixed
sk_live_.
# AI detection on text — returns overall + per-sentence scores
curl -X POST https://api.textsight.ai/v2/detect \
-H "Authorization: Bearer sk_live_YOUR_KEY" \
-H "Content-Type: application/json" \
-d '{"text": "The text you want to check."}'
The response carries a humanization_score, an
ai_probability, a verdict, a
confidence value, per-sentence breakdown and a
request_id. /v2/score returns the same
numbers without the sentence array when you do not need it.
Scopes are enforced per key: detection and scoring need the scan
scope, rewriting needs humanize. Quota is counted in API calls per
month and varies by plan. Full parameters, error codes and rate limits are in
the API documentation.
/v2/detect will not work — it takes text.
What would actually help us give you a straight answer.
If you have a genuine volume use case for voice detection — a fraud queue, a moderation pipeline, a research corpus — get in touch. We would rather have a concrete conversation than a waitlist. The things that let us respond usefully rather than vaguely:
The technical guide: how detection works, and why it generalises poorly to unseen generators.
Read the technical guide →Why we publish no single accuracy percentage, what false positives look like, and how to weigh a result.
See our position →Parameters, error codes, rate limits and authentication for the public text API.
Read the docs →Three checks a day, free, no signup. Your audio is never stored.