How Voiceora Uses AI

How Voiceora uses AI

Voiceora turns content you provide into platform-native drafts using large language models.

What we use. Open-weight Qwen models, accessed through OpenRouter and run by DeepInfra on servers in the United States. We pin specific model versions rather than floating "latest" aliases, so your brand voice stays consistent between generations.

What we send. The content you submit for a generation — text, images, and (when Moment Bundle video is enabled) video — plus your brand voice configuration and the instructions needed to produce the requested format. We do not send your name, email address or billing details to AI providers.

Training. We do not train AI models on your content, and neither do our providers. We keep prompt logging disabled on our routing account and do not permit routing to providers that train on submitted data. Our inference provider may retain inputs and outputs briefly for debugging.

Human involvement. Voiceora produces drafts. You review, edit and decide what to publish. Editorial responsibility for anything you publish rests with you.

What AI can get wrong. Language models can state things that are false with apparent confidence, invent quotes or statistics, misread images, and produce phrasing similar to content generated for someone else. They can also reflect biases present in their training data. Please treat every output as a first draft from a fast but unreliable assistant.

Your choice. You choose what to submit. You can delete your content and your account at any time from Account Settings.

Questions: support@voiceora.io