Review signal
Use the result to guide review, not as proof of identity.
AI Voice Detector APIAI Voice Detector API
Analyze uploaded audio for signs of AI-generated speech. Get a clear result your team can review inside your own product.
Audio sample
4.2 seconds
voice-sample.mp3
Illustrative audio analysis
Example result
- Result 92% AI probability
- Duration 4.2 s
- Billed seconds 5
- Credits used 50
Request
4.2 seconds
curl -X POST https://developer-portal.walterwrites.ai/api/voice-detector/detect/ -H "X-API-Key: YOUR_API_KEY" -F "[email protected]"
Response excerpt
{
"status": "success",
"verdict": "ai_voice",
"ai_probability": 0.92,
"natural_probability": 0.08,
"duration_seconds": 4.2,
"segments": [],
"billed_seconds": 5,
"credits_used": 50
}Make audio review part of your product.
Bring audio analysis into your existing upload, review and evaluation workflows.
Moderation queues
Add a review signal to submitted audio. Give moderators the original clip, the result and the context they need to apply your policies.
Editorial verification
Add a check when assessing contributed interviews or voice clips. Keep source verification and editorial judgment in the workflow.
Internal review tools
Bring audio analysis into a private dashboard. Track reviewer decisions and investigate patterns across the recordings your team handles.
Upload workflows
Request an analysis after an audio upload and show the result alongside the original recording.
Reviewer dashboards
Help reviewers find recordings that need attention. Keep human decisions and supporting context together.
Evaluation pipelines
Compare results with a labeled set of natural and synthetic recordings before choosing review thresholds.
Voice detectionand the rest of Walter.
Use one developer portal and scoped API keys for humanization, AI text detection, grammar correction, plagiarism scanning, and AI image detection.
AI Detector API
LiveAnalyze text for AI-generated content with confidence scores and sentence-level details.
AI Humanizer API
LiveTransform AI-generated text into natural, human-like content through a documented REST endpoint.
AI Voice Detector API
LiveAnalyze uploaded audio for synthetic speech and review the returned evidence in your product.
Plagiarism Checker API
LiveScan text against web sources and return a score, matched URLs, and highlighted character ranges.
Grammar Checker API
LiveCorrect grammar, spelling, and punctuation and return corrected text plus a complete word-level diff.
AI Image Detector API
LiveUpload an image and receive a real, fake, or inpainting verdict with confidence and per-class probabilities.
Clear costs. Pay by audio duration.
10 credits per audio second, rounded up. These examples show credits per recording, not subscription prices.
Need enterprise pricing?
Discuss higher-volume audio review and your integration requirements with our team.
Frequently asked questions.
Practical answers for teams evaluating and integrating voice detection.
What does the AI Voice Detector API return?
A verdict, probabilities and audio duration. Longer recordings also include segment results. See the response reference.
Which audio files can I upload?
WAV, FLAC, OGG, MP3, M4A and AAC, up to 150 MB and 10 minutes. Extract audio from video before uploading.
Can it detect every cloned or AI-generated voice?
No universal detection guarantee is made here. Evaluate the API against the generators, languages and recording conditions relevant to your product. Independent audio-deepfake research shows why benchmark performance alone is insufficient for real-world deployment.
Does a high score prove impersonation?
No. A detection score should prompt investigation, not identify a speaker or establish intent. Check the recording’s source and corroborating evidence before drawing conclusions.
Can I analyze a live audio stream?
The documented endpoint accepts uploaded files. It does not document a streaming interface. Design this integration around completed recordings.
Why can a timeline omit the end of a recording?
A trailing span shorter than the analysis window can be skipped. Do not present unprocessed audio as verified natural speech.
How should I protect my API key?
Make requests from your backend. Keep the key out of browser code, logs and public repositories. Use a secrets manager and separate environment keys. See authentication guidance.
What should happen when a request fails?
Explain the failure and preserve the upload. Correct invalid input before retrying; use controlled retries for temporary errors. The error guide describes response codes and recovery.
Is uploaded audio retained or used for training?
This page makes no zero-retention or no-training promise. Review Walter’s privacy policy and Trust Center, then confirm any audio-specific requirements with the team before integrating sensitive data.
How should I evaluate accuracy for my use case?
Create a labeled test set representative of your users. Measure false positives and missed detections separately, including changes in quality and recording conditions. Treat published research as evaluation guidance, not as a Walter accuracy claim.
Ship your first AI Voice Detector API request.
Create an API key with the required scope, follow the documented request format, and keep the returned evidence in your existing review workflow.
Developer resources.
From the blog
Latest Research & Insights
Guides, comparisons and practical advice from Walter.
View all-
How to Humanize AI Text Inside Codex
Connect Walter to OpenAI Codex over MCP in three commands. Humanize drafts and run AI detection without leaving…
Read more -
How to Humanize AI Text Inside Claude
One prompt. The whole content workflow. Walter’s MCP connector brings humanization, detection, and batch processing directly into Claude.…
Read more -
ChatGPT vs ChatGPT Plus: My Real Experience After Using Both Every Day
One of the most common questions I get as a heavy AI user is the old “ChatGPT vs…
Read more
