Bad Times AI: Lower Entry Price, One Photo, and Credit Refunds That Build Trust

A developer built Bad Times AI, a Cloudflare Workers-hosted tool that turns a single person or pet photo into a roughly 15-second 9:16 quarrel or lip-sync dance video, using TanStack Start for the front end, Cloudflare D1 for credits and task records, Stripe Checkout for one-time credit packs starting at $6.90, and TryAllAPI's Seedance integration for asynchronous video generation. The build emphasizes honest paid pricing and automatic credit refunds on failed generations, with the developer arguing that restoring credits on supply, safety, or blurry-input failures avoids chargebacks and makes the low entry price feel safe.

How I shipped a one-photo → quarrel / lip-sync dance product on Cloudflare Workers — honest “not free” pricing, Stripe credits, Seedance via TryAllAPI. badtimesai.cc https://badtimesai.cc/?utm source=devto&utm medium=blog&utm campaign=builder takes 1 clear person or pet photo → ~15s 9:16 quarrel / lip-sync dance MP4 with original AI ambient audio. Browser + Stripe. Independent fan tool — not affiliated with song/artists/TikTok creators named in the meme. Person templates 18+ with likeness rights. Angle: Stripe credits + accessible entry pricing from $6.90 , plus explicit “failed generations refund credits” next to the sell. Prices and model ids change; treat live site / console as source of truth. The hero says packs from $6.90 one-time, no subscription and repeats not free . That filters tire-kickers who only want infinite free GPU and sets expectations for a paid render. Other tiers visible in page copy while writing include $9.90 and $16.90 -style packs — verify live before quoting. Guest checkout path keeps friction low for “I just want one quarrel clip.” Unlike strict two-face booths, Bad Times leans one reference with alternate scenes person / pet / character messaging and a “describe any idea” adjacent path on the page. The credit system still has to: Why refunds are marketing: face video fails for boring reasons — supply, safety, blurry input. If credits vanish silently, you get chargebacks. If credits restore automatically, the $6.90 experiment feels safe. | Layer | Choice | Role | |---|---|---| | Front / SSR | TanStack Start Vite | Landing, upload panel, generation status | | Runtime | Cloudflare Workers | Edge deploy, secrets, API routes | | DB | Cloudflare D1 | Credits, orders, task rows table prefixes if DBs are shared | | Payments | Stripe Checkout | One-time packs + optional larger packs; guest checkout | | Video upstream | TryAllAPI → Seedance | Async create + poll | | Analytics | GA | Funnel sanity check | 1 photo + scene choice → quarrel / lip-sync template → check credits else Stripe → Worker creates Seedance task via TryAllAPI → poll until MP4 URL → store / show → on failure: refund credits Operational checklist after register / top-up: badtimes-prod vs sandbox . TRYALLAPI API KEY . Never in the frontend bundle or a blog post. Conceptually the call shape looks like this field names follow the aggregator’s v3 contents API; always re-check current docs : POST /api/v3/contents/generations/tasks Authorization: Bearer <TRYALLAPI API KEY Content-Type: application/json Accept-Encoding: identity GET /api/v3/contents/generations/tasks/{id} I send Accept-Encoding: identity from the Worker because some gateways behave badly with transparent decompression. Model ids used in production look like Seedance 2.x family strings — treat console ids as source of truth, not this paragraph. A few details that bit me across this family of sites: duration as an explicit integer. resolution and ratio . generate audio . queued → running → succeeded / failed . Success URLs are often supply unavailable or hard failures: refund credits, show a short human message, log the detailed upstream error server-side, optionally fall back to another Seedance tier. Rough mental model for upstream cost illustrative only — use your own usage fields and invoices : tokens ≈ duration sec × width × height × fps / 1024 cost ≈ tokens × price per million tokens Price the site after you measure a few real runs with retry margin, Stripe fees, and failed-job refunds — not from a spreadsheet fantasy of 0% failure. Keep TryAllAPI USD as COGS; site credits as the customer meter. Rehost signed MP4 URLs. Lower entry pricing only survives if you measure real runs including refunds. Demo UIs are easy. The line between toy and product is: task id in D1. Two books of account stay separate: USD balance at TryAllAPI your cost vs site credits what you sold . Mixing them in one mental ledger is how you underprice or invent “mysterious” losses. When several products share one Stripe account, isolate webhooks by endpoint or by metadata site / product . Callbacks pointed at the wrong apex domain are a classic multi-site footgun. Secrets stay in Workers: TRYALLAPI API KEY , STRIPE SECRET KEY , webhook secret, auth secrets. The browser only ever sees publishable Stripe keys if needed. Typical loop on this template family: pnpm install wire D1 database id, production env, custom domain in wrangler pnpm cf:deploy Before calling a deploy “done”: AUTH TRUSTED ORIGINS and Stripe return / webhook URLs match the production HTTPS origin. Production origin: https://badtimesai.cc . Isolate webhook metadata if Bad Times shares a Stripe account with sister dance domains. Face-based generation is not “just another SaaS CRUD app”: Those constraints belong in the product, not in a footnote you add after the first dispute email. Consumer: upload one clear authorized photo at badtimesai.cc https://badtimesai.cc/?utm source=devto&utm medium=blog&utm campaign=builder . Builders: TryAllAPI https://tryallapi.com/register?aff=5A6A&utm source=devto&utm medium=blog&utm campaign=builder , measure COGS, then set an entry pack that still survives refunds. Cheap only works when failure is refundable.