AI product engineering
VoiceFans: AI personas with usage controls
A V1 AI persona platform covering discovery, conversations and usage controls. The case study distinguishes demonstrated screens from planned or illustrative flows.
ContributionSource-reported lead development, architecture and project management; pre-launch V1
View the product
A convincing AI conversation is only one part of the product.
How does someone discover a persona worth talking to? What context shapes the reply? What happens when voice generation costs more than a plan allows? VoiceFans brings those product and engineering questions together in a creator/persona experience.
The supplied dossier attributes lead development, project management and architecture work to HINTX, including AI, payments and cost controls. It describes V1 in pre-launch/final testing. This case study therefore explains the product work without inventing an active subscriber base, revenue or a public launch.
01 / Give discovery a purpose
The actual discovery interface groups persona demonstrations by topic and presents cards before a conversation starts. The shown historical-character catalogue is seeded demo content, not evidence of creator endorsement, real customer reviews or market traction.
Authentic discovery-screen crop with seeded demonstration personas. Unverified headline totals are excluded; remaining demo cards are not real creator endorsements or customer evidence. Select the image to enlarge.
02 / Connect persona context to text and voice
The source describes persona identity, backstory, tone, knowledge and testing alongside text and voice features. OpenAI models and ElevenLabs capabilities are listed in the engineering account. The useful experience is a coherent conversation with explicit product boundaries, not simply placing a model response inside a chat bubble.
The new visual uses a short sample exchange to explain the intended flow. It replaces the old chat capture in this gallery because that capture still contained personal-name references. Original evidence is retained privately; the illustration does not pretend to be an actual accepted voice interaction.
Illustrative conversation and usage-control UX. Sample dialogue, abstract waveform and interface layout are explanatory, not a real user conversation, captured VoiceFans screen or generated audio result. Select the image to enlarge.
03 / Treat AI usage as a first-class system concern
An AI product needs to understand who can request a response, which allowance applies and what was consumed. The dossier describes server-side caps, top-ups, a usage ledger and model/voice cost considerations. It also describes idempotent billing webhooks so a retried event should not be counted twice.
These controls belong alongside role access and data isolation. The reported stack includes Next.js/TypeScript, Supabase, Stripe/Connect and the AI providers. Actual live payouts, quota enforcement and production billing acceptance are not established by the supplied screenshots.
Illustrative request-control sequence based on the engineering dossier. No live payment, quota test, provider cost or performance result is fabricated. Select the image to enlarge.
04 / Build the operations around the experience
Creator, user, conversation and settings interfaces form the administrative side of the product. The module map separates those responsibilities from the member journey and commercial integration. This makes the less visible engineering legible without turning an admin navigation item into a delivery claim.
Illustrative module map of the reported V1 scope. Pre-launch status is retained; the board is not a production release receipt. Select the image to enlarge.
What would complete this case study?
Final acceptance captures of persona creation, a permitted voice interaction, usage-limit behaviour and sandbox billing would strengthen the story. The intended value is a sustainable AI experience; no unmeasured engagement, latency or revenue improvement is claimed.
Explore a related engineering problem
Voice and conversational AI · Separate voice and messaging engagements · Discuss your workflow with HINTX.
AI + software engineering
Make your systems work together.
Tell us about the workflow, the people using it and the constraints that matter.
Discuss your project



