How We Use AI

AI is changing how creative work gets made, and we believe families have the right to know what's made by machine and what isn't. This page is a plain-English account of how AI fits into the books, videos, and reels you find on Random Creative Sparks.

What's made by AI

Illustrations are generated using image-generation models (currently Z-Image, FLUX, and SDXL variants). Every illustration is reviewed and chosen by a human (Vijay) from a set of candidates the model proposes. If the picture isn't right, it doesn't ship.

Voice narration is produced using F5-TTS, an open-source neural voice model. All voice references are recorded by real people who have given us their express permission to use their voice for RCS narration. Voice contributors are credited where they wish to be credited and kept anonymous where they prefer privacy.

Video and reel composition is done programmatically (using ffmpeg) — not AI per se, but worth disclosing for transparency.

What's made by a human

• All story text is written by Vijay Kukreja, an actual person who tells bedtime stories to actual children.
• All story concepts, character designs, pacing, and emotional arcs are by Vijay.
• Editorial selection of every page's final image is done by Vijay. AI proposes; a human curates.

What we won't do with AI

• We will not generate likenesses of real public figures or deceased persons.
• We will not generate AI voices of people who haven't given explicit consent for voice cloning. For minors, voice use requires the express consent of a parent or legal guardian in addition to the child's own willingness.
• We will not pretend our content was made entirely by humans when AI was involved. That's why this page exists.

Voice consent

Every voice used for narration on this site was recorded by the voice owner specifically for the purpose of narration on RCS content, with their consent. We do not scrape voices from podcasts, YouTube videos, audiobooks, or any other source where the speaker did not consent to having their voice modeled.

No training on visitors

We don't collect data from visitors to this site (see our Privacy Policy), so there is no visitor data being fed into any AI model.

Quality and editorial control

Every book published here was reviewed page-by-page by Vijay before it went live. AI is a tool we use in production; the editorial responsibility is human.

Models we've used

• Image: Z-Image, FLUX, SDXL (and variants of each)
• Voice: F5-TTS
• Story-drafting assistance: large language models, used as collaborators in early drafting — not as autonomous authors

Questions or concerns

AI questions: contact@rcs.owltreeconsulting.com