CapyCue uses artificial intelligence to turn raw screen recordings into editable product walkthroughs. This disclosure explains the current processing path and the choices customers should understand before selecting Enhance with Capy AI.
1. What is processed
After you request enhancement, CapyCue uses media tools in our Azure environment to prepare the recording. The current workflow:
- extracts an audio track from the recording;
- extracts a limited number of representative screenshots in chronological order;
- sends the audio to OpenAI's audio transcription API;
- sends the recording title, transcript, and representative screenshots to an OpenAI text-and-vision model to draft the title, summary, timed steps, bullets, and narration;
- sends the narration text and selected synthetic voice to OpenAI's speech API; and
- combines the approved assets into a rendered walkthrough in our Azure processing environment.
The original source video is stored and processed in CapyCue's Azure environment and is not currently sent to OpenAI as one complete video file. Product changes may alter this architecture; we will update this disclosure before materially expanding provider processing.
2. Provider data use and retention
OpenAI's official API documentation states that API data is not used to train or improve its models unless the API customer explicitly opts in. OpenAI may retain some API inputs and outputs in abuse-monitoring logs for up to 30 days by default, depending on the endpoint and the account's data-control configuration. Audio transcription, speech, image, and response endpoints can have different retention behavior.
CapyCue does not represent that your processing uses OpenAI Zero Data Retention unless your written CapyCue order expressly states that the relevant account, project, endpoints, and workflow are approved and configured for that control.
3. Human review and accuracy
AI output is a draft. It may omit actions, misunderstand a screen, create awkward timing, or produce inaccurate text. You can edit walkthrough text and narration, select a different voice, save revisions, and request a new render. Review the complete result before sharing or embedding it.
4. Synthetic voice disclosure
CapyCue-generated narration uses a synthetic voice. CapyCue discloses AI-generated voice to viewers in supported players and exports. You may not use CapyCue to imitate a real person without permission, misrepresent a person's endorsement, or deceive viewers about the origin of synthetic speech.
5. Sensitive and confidential information
Do not submit recordings containing secrets, passwords, private keys, full payment-card data, protected health information, or other highly sensitive information unless you have authority, a lawful purpose, and appropriate contractual and technical safeguards. Pause notifications, close unrelated applications, choose the narrowest recording source, and review extracted content before processing when possible.
6. Cost and usage
AI enhancement and reprocessing can consume plan minutes or create overage. CapyCue maintains a per-video processing ledger for transcription, analysis, speech, image, and compute costs. Customer-facing usage and provider-cost entries may be estimated until provider billing is reconciled.
7. Questions and enterprise controls
Business customers that need specific retention, region, security, or subprocessor commitments should contact support@capycue.com before processing regulated or sensitive content.