Frequently Asked Questions

Duel has up to seven rounds. Interrogation is a five-round thriller against Graves. Tribunal is a six-round showdown against a rival and unlocks after defeating AXIOM.

The Training Arena is coming soon. No training is required for live play. Arena opponents unlock through wins; Tribunal unlocks after defeating AXIOM.

Yes. Speak naturally, then submit your move. Tap the mic to interrupt opponent playback and speak. Voice input depends on your browser and microphone permission; text is always available. Graves and Tribunal responses target 25–45 spoken words, with a 60-word ceiling.

Voice generation is the highest ongoing infrastructure cost. Limits keep plans sustainable while preserving quality and responsiveness.

Moves are judged on relevance, scene consistency, tactical leverage, and creativity—not word count or legal jargon. The AI proposes a structured result; the server validates it and applies mode-specific rules. A short explanation follows each move. Invalid AI output does not spend a turn or damage your score.

Interrogation: earn all displayed breakthroughs and keep suspicion below 60 within five rounds. Tribunal: earn all breakthroughs and reach 70 panel favor or reduce the rival's grip to 10 within six rounds. Partial progress is not completion. If you miss either requirement by the final round, the rival wins. Tribunal uses one aggregate favor meter, not three independent AI judges.

Declaring victory does not earn it. Player text cannot directly write scores or change the server's rules. Fixed scene facts, validated structured results, and guarded turn writes reduce prompt-injection and cheating risks. AI judging can still make mistakes; no system here is claimed to be injection-proof.

Provider capacity, connection problems, or anti-abuse controls can temporarily pause a turn. Shorter replies and bounded conversation context reduce the workload. Retry the same move safely; a saved response is not scored twice. Refresh the battle URL to recover saved turns. Voice allowances are separate from text-provider capacity.

Text responses and judging currently use Groq's openai/gpt-oss-120b model. Spoken opponent audio uses ElevenLabs. The game supports imaginative fantasy conflict; that is not a promise of unrestricted content or an exemption from provider safeguards.

Challenger ($0): unlimited text battles, all modes, 5 voice battles to start then 2 per week, post-game arc summary. No card required.

Duelist ($12.99/mo or $99/yr): around 30 voice battles per month, unlimited text, God Mode, priority leaderboard ranking, full battle history and stats, epithet identity, and first access to Worlds 3–5.

Registered plans have no text-battle allowance to spend. Guest allowances are smaller, and anti-abuse controls and provider capacity limits still apply. These are separate from your plan's voice allowance.

Your profile tracks record, win rate, history entries, mode performance, and recent form. It also stores settings like avatar, difficulty, and voice preferences.

Battle and profile data is retained so progression, rankings, and history features work as expected.

Prompt Duel is founded and operated by Michael Langer.

For partnerships, press, investor questions, and product feedback: michael@promptduelai.com.

The Founder Beta is a limited 30-day access program for early testers. Apply at promptduelai.com/v2/beta/. Accepted applicants receive a promo code that unlocks Founder tier — including 10 voice battles per week and God Mode.

In return, Founder testers commit to completing at least 5 voice battles, reporting bugs and broken experiences, and filling out a short exit survey at the end of 30 days.

God Mode removes standard battle constraints — no rules, no limits. It is available exclusively to Duelist subscribers and Founder Beta testers. Enable it in Settings under Gameplay.

Yes. You can manage or cancel your subscription from your account billing settings.

Billing is processed by Stripe. Prompt Duel does not directly store full card details.