Skip to content

MVP スコープ / MVP Scope

MVP Goal

"3 characters moving in a studio apartment, tips trigger instant reactions, free comments get responses every 10 minutes, streaming on YouTube + TikTok"

「1Kの部屋で3体が動き、投げ銭に即時反応、無料コメントは10分ごとに対応、YouTube+TikTokで同時配信」

MVP に含めるもの / MVP Features

#FeatureDetail
1Room rendering1K studio, 1 screen. Tilemap with basic furniture + dog bed
23 characters displayed & moving32×32 sprites. 4-direction walk animation. A* pathfinding
3Basic animationsWalk, stand, sit, sleep, eat (5 types)
410-min cycle systemEvery 10 min: AI decides action + responds to free comments
5Instant tip reactionsTips from any platform → immediate character response
610 action categoriesSleep, eat, cook, clean, work, rest, talk, play with Eve, walk, idle
7Conversation systemCharacter-to-character dialogue in speech bubbles. LLM-generated
8Status managementHunger, mood, energy, stress. Changes over time
9YouTube integrationSuper Chat detection + Live Chat polling
10TikTok integrationGift detection + Live Comment via WebSocket
11Unified tip processorConvert YT Super Chat + TikTok Gifts → LifeCoin
12Tip effectsBanner notification + item spawn animation
13Tier 1 items100〜500 LC items (10 types functional)
14Comment responsesPick notable comments, characters reply in speech bubbles
15Bilingual dialogueJP/EN response based on comment language
16UI overlayStatus bar, clock, game day count, LifeCoin balance
17OBS-readyBrowser source compatible. 1280×720px
18Multi-platform streamOBS → YouTube + TikTok simultaneously

MVP に含めないもの / NOT in MVP (Later Phases)

FeatureReasonPhase
Voice synthesisStreaming works without itPhase 2
Outing scene graphics"Out" text display as placeholderPhase 2
Room expansion (moving)1 room map is enoughPhase 2
Seasonal changesEasy to add laterPhase 2
Large events (wedding)Tier 1-2 is enough for launchPhase 3
RAG-based long-term memoryJSON last-N entries as substitutePhase 3
Mobile appNot needed for MVPPhase 3

Phase Plan

Phase 1: MVP(Launch)

  • All MVP features above
  • 1 room, basic actions, Tier 1 tips, comment responses, dual-platform

Phase 2: Expansion(1-2 months after launch)

  • Voice synthesis (VOICEVOX)
  • Outing scene graphics (dates, walks)
  • Tier 2-3 items & events
  • Seasonal system
  • Moving system
  • Advanced comment reactions (remember viewer names)

Phase 3: Full(3+ months)

  • Tier 4-5 large events
  • RAG-based long-term memory
  • Mobile viewer app
  • Live2D narrator character
  • Additional platform support

MVP Development Estimate

TaskDays
Project setup (monorepo, TS config)0.5
Pixel art assets (3 characters + room tiles + furniture)3〜5
Backend: Game loop & 10-min turn scheduler1.5
Backend: Tip interrupt handler (immediate reactions)1
Backend: AI agent integration (Claude API)2
Backend: Memory management (SQLite)1
Backend: YouTube API integration1.5
Backend: TikTok Live Connector integration1.5
Backend: Unified tip processor0.5
Backend: Comment processor & filtering1
Backend: Bilingual prompt system0.5
Frontend: Tilemap & sprite rendering2
Frontend: Animation & movement1.5
Frontend: UI (status, speech bubbles, tip effects, i18n)2.5
WebSocket integration & testing1
Multi-platform streaming setup (OBS + Restream)0.5
End-to-end testing1
Total~23-25 days (including assets)

Pixel Art Assets

AI image generation tools (Stable Diffusion + PixelArt LoRA) can significantly reduce asset creation time.