[ 01 ]What it looks like
A headphone icon that plays digits or words over noise, with a box to type what you heard.
Captcha type · Pattern
Audio captcha solver API
Send an mp3 or wav and get the spoken characters back as text. The audio path is first-class here, which matters for accessibility tooling.
Also searched asrecaptcha audio challengespeech to text captchamp3 captcha recognitionaccessibility captcha solveraudio challenge bypass
- Price
- $0.75
- per 1,000
- Token acceptance
- 98.9%
- Median solve
- 4.0s
- p95 solve
- 9.0s
[ 02 ]Why it is hard
The audio is degraded on purpose: overlapping speakers, pitch shifting, reverb and a noise floor tuned to break general-purpose ASR while staying intelligible to a person. It is also the accessibility path, so providers rate-limit it harder than the visual one and cap replays, which leaves very little room for a second attempt.
[ 03 ]How we solve it
Send the clip as a base64 body with a lang code and you get the spoken characters back as text. No sitekey and no proxy, the task is the audio file alone, so it sits in the cheapest tier alongside the image types. Recognition runs on models fine-tuned per language on captcha audio rather than clean speech, because the noise profile is nothing like ordinary dictation.
[ 02 ]Solve it in five lines
Solve it in five lines
Post the task, get the token. Our SDKs collapse creation, polling and retries into one call.
audio-captcha.sh
curl -s https://api.nexcaptcha.net/createTask \
-H "Content-Type: application/json" \
-d '{
"clientKey": "YOUR_API_KEY",
"task": {
"type": "AudioTask",
"body": "...",
"lang": "..."
}
}'Required parameters
- API task type
- AudioTask
- body
- required
- lang
- required
- Difficulty
- Moderate
[ 03 ]Where this one shows up
Where this one shows up
[ 05 ]Questions about this one
Questions about this one
$0.75 per 1,000 clips, the same tier as image-to-text and rotate. Failed solves are not billed, which matters here because the success rate is the lowest of the cheap tier at 98.9%.
The lang field takes en, fr, de, el, pt and ru, each on its own fine-tuned model. Send the code that matches the challenge rather than leaving it to detection, since a two-second clip of digits gives a detector almost nothing to work with.
Cheaper, not easier. Providers rate-limit the accessibility channel and cap replays, so you get fewer attempts per session; use it when the visual challenge is unavailable or when you are testing an accessibility flow, not as a general shortcut.
No. The task is the audio file, with no sitekey and no token, so nothing is bound to your IP. Download the clip in your own session, send it, type the result back.
Solve your first thousand on us.
Create a key, paste it into the sample above, and watch a real token come back. No card, no sales call.