AI USE & TRANSPARENCY POLICY
Voocaly is an artificial intelligence-powered speech synthesis platform. This AI Use & Transparency Policy discloses the nature of our neural speech generation pipelines, clarifies data boundaries, and details the transparency considerations applicable to creators distributing synthetic audio under global regulations, including transparency principles reflected in Article 50 of the European Union Artificial Intelligence Act (Regulation (EU) 2024/1689). Voocaly is evaluating and implementing transparency measures applicable to AI-generated audio in jurisdictions where such requirements apply.
Table of Contents
01. Synthetic Nature of Generated Outputs
Every audio file, speech stream, vocal track, and audition sample generated on Voocaly is artificially generated synthetic media produced using deep neural networks and machine learning acoustic models.
While Voocaly voices are trained to sound organic and realistic, they do not represent authentic live recordings of human speakers during generation.
02. Inference Architecture & Model Provenance
Voocaly uses a multi-provider neural inference pipeline:
- Neural Studio Voices & Persona Profiles: Certain Neural Studio voice labels (including catalogue profiles designated in technical registries as Kokoro, Orpheus, Qwen, and Voxtral) are routed through OpenRouter to OpenAI GPT-Audio models. Product voice labels represent curated acoustic presentation-layer personas and stylistic configurations rather than separately hosted foundation models.
- Zero-Shot Voice Cloning: Synthesized via Fish Audio S2.1 Pro on OpenRouter (primary), with direct Fish Audio API fallback.
- Grok TTS: Synthesized via direct xAI REST API (primary) with OpenRouter fallback.
- Edge Demo Voices: Synthesized via Microsoft Edge TTS using local speech synthesis libraries connecting to Microsoft Speech Services.
- Luna AI Copilot & Emotions: LLM chat and emotion tag analysis routed via OpenRouter.
No Foundation Model Training: Voocaly does not train its own foundation models on customer scripts or voice recordings. Provider-specific retention and model-training practices are governed by applicable provider terms, as detailed in our Subprocessor Register. Voocaly does not independently represent that all third-party providers maintain identical retention or training practices.
03. Audio Provenance & Machine-Readable Metadata
Generated audio is delivered to users in standard MP3 and WAV container formats.
Voocaly is evaluating and implementing transparency measures applicable to AI-generated audio in jurisdictions where such requirements apply. Voocaly does not currently embed proprietary inaudible acoustic watermarks, C2PA cryptographic provenance manifests, or machine-readable AI markers directly into downloadable audio files. Creators and publishers should take account of these technical characteristics when assessing downstream legal and platform disclosure requirements.
04. Downstream Creator Disclosure Obligations (EU AI Act Article 50)
Under Article 50 of the European Union Artificial Intelligence Act (Regulation (EU) 2024/1689) and comparable synthetic media laws globally:
- Deceptive Media & Deepfakes: Deployers who generate or manipulate audio content constituting a deepfake must disclose that the content has been artificially generated or manipulated.
- Public Interest & News Content: Where applicable law requires disclosure of artificially generated speech representing real individuals or used in matters of public interest, the deployer or publisher is responsible for providing the required disclosure.
Include an audible audio disclaimer or visible caption: "Voice generated by AI" or "Speech synthesized using Voocaly AI".