Audio Node

AI Transparency Statement

Last updated: June 22, 2026

At MakeLabs, we believe that AI-powered tools come with a responsibility to be transparent about how they work, what data they use, and how they can be misused. This document explains how Audio Node's AI voice technology operates and our commitments to responsible deployment.

1. How Our AI Works

Audio Node uses open-source neural text-to-speech (TTS) models to convert text into natural-sounding speech. Specifically:

  • Kokoro ONNX — an open-source TTS model supporting 9 languages and 54 voices, deployed on RunPod serverless GPU infrastructure;
  • Qwen TTS — an open-source multilingual TTS model supporting additional voice styles and accents.

Neither model is a voice clone of any real individual. The voices are synthetic — trained on datasets of speech and designed to produce natural-sounding output, but they do not replicate any specific person's voice identity.

2. What We Do NOT Do

  • We do not train on your data: Your Input text and generated Output audio are never used to train, fine-tune, or improve any AI model — by us or any third party we work with;
  • We do not clone real voices: Audio Node does not currently offer voice cloning. The voices in our library are synthetic and do not replicate any named individual;
  • We do not create deepfakes: Our platform is not designed or intended for creating non-consensual synthetic media of real people. This is explicitly prohibited in our Acceptable Use Policy;
  • We do not retain your text: Input text is processed to generate audio and then discarded. We do not build profiles of what you type or create datasets from your submissions;
  • We do not surveil users: We do not use AI to monitor, score, or profile user behaviour beyond what is necessary to enforce plan limits and prevent abuse.

3. Our Stance on Deepfakes & Impersonation

Synthetic voice technology is powerful, and with that power comes the potential for harm. We take the following firm positions:

  • Creating audio that falsely represents a real person's voice without their consent is prohibited and constitutes grounds for immediate account termination;
  • Using Audio Node to generate disinformation, political manipulation content, or synthetic evidence of events that did not occur is prohibited;
  • Using generated audio to deceive, defraud, or harm any person is prohibited;
  • We support the development of AI voice detection tools and audio provenance standards and do not oppose efforts to detect or label AI-generated audio.

4. Disclosure & Consent Guidance

We recommend the following best practices for users of Audio Node:

  • Disclose AI origin when publishing AI-generated audio in contexts where listeners might reasonably assume they are hearing a human voice — especially in journalism, education, customer service, and public communications;
  • Label AI content in compliance with any applicable laws in your jurisdiction (such as the EU AI Act, US state-level AI disclosure laws, or India's emerging AI governance frameworks);
  • Obtain consent before using AI voice in any context involving a real individual's likeness, name, or identity.

5. Open-Source Model Transparency

We believe in transparency about the AI we deploy:

  • Kokoro ONNX is an open-source model. Its training data, architecture, and license are publicly documented by its developers;
  • Qwen TTS is developed by Alibaba DAMO Academy and available under a commercially permissive open-source license;
  • MakeLabs does not modify these models' core capabilities — we deploy them as-is through a serverless API (RunPod) and provide the user interface and management layer.

Because we use commercially licensed open-source models, we can grant users commercial rights to their Output without additional licensing fees. See our Commercial License Terms for details.

6. Limitations of AI Voice

We are transparent about what our AI cannot guarantee:

  • Output quality depends on input text, punctuation, language, and voice selection. Results may vary;
  • The model may mispronounce uncommon proper nouns, technical terms, or words in mixed-language input;
  • Output is not unique — other users submitting identical input may receive similar or identical audio;
  • The AI does not understand context or intent — it generates audio based purely on the text provided, regardless of meaning or consequence.

7. Reporting Misuse

If you encounter Audio Node output being used in a harmful, deceptive, or abusive way — or if you believe our platform is being misused to produce harmful synthetic audio — please report it to support@audionode.dev with subject “AI Misuse Report.” We investigate all reports and cooperate with law enforcement where appropriate.

8. Evolving Commitments

AI governance is a fast-moving area. We will update this statement as our practices evolve, as new regulations take effect, and as our understanding of responsible AI deployment deepens. We are committed to being honest about what we know and what we are still learning.

9. Contact

Questions about our AI practices? Contact us at support@audionode.dev.