MasterTone

Build a premium Russian-language AI audio mastering web app called MasterTone.

Product summary:
A dark, modern, desktop-first SaaS for musicians, producers, and creators who make tracks in Suno and want professional release-ready mastering. The app must support single-track mastering and stem-based mastering, with a realtime AI agent UI that explains what is happening step by step in Russian.

Core features:
1. Auth and onboarding in Russian: email, Google, Telegram login.
2. Project creation: project name, genre, target platform, mastering target (streaming, loud club, cinematic, social).
3. Audio upload: drag-and-drop for one track or multiple stems, resumable uploads, file validation, clear error states.
4. AI mastering pipeline: analyze loudness, peak, spectrum, stereo width, dynamics, clipping; auto-select a mastering chain with EQ, compression, saturation, stereo widening, limiter, de-essing, tonal shaping.
5. Stem processing: detect stem type, process vocal/drums/bass/synths differently, then render final master.
6. Realtime agent panel: show current step, estimated time, tools being used, why the step is applied, and live progress updates.
7. Playback comparison: original vs processed vs alternate preset A/B switching, with synchronized playback and metrics like LUFS and true peak.
8. Export: WAV 24-bit and MP3 320 kbps, version history, metadata, and downloadable previews.
9. Full Russian localization and premium dark theme with neon accents, accessible contrast, and smooth audio visualizations.

Primary screens/flows:
Screen 1 Landing page with value prop and signup.
Screen 2 Dashboard with projects list and create project.
Screen 3 Project detail with upload, analysis status, agent timeline, waveform, and metrics.
Screen 4 Compare screen with A/B player and preset switcher.
Screen 5 Export modal/page with formats and file history.
Screen 6 Settings for profile, billing, language, and data deletion.

Data model:
User(id, name, email, authProvider, locale, plan)
Project(id, userId, name, genre, targetPlatform, masteringTarget, status, createdAt)
Asset(id, projectId, type: track|stem, filename, duration, format, storageUrl, validationStatus)
Analysis(id, projectId, loudnessLUFS, truePeak, spectrumSummary, stereoWidth, dynamicsSummary, confidence)
AgentStep(id, projectId, label, status, details, startedAt, endedAt)
RenderVersion(id, projectId, presetName, outputFormat, storageUrl, createdAt)
ExportJob(id, projectId, status, progress, errorMessage)
AuditEvent(id, userId, projectId, eventName, payload, createdAt)

Tech stack:
Next.js 14 or 15, TypeScript, Tailwind CSS, shadcn/ui, Framer Motion, Zustand or TanStack Query, Supabase or PostgreSQL for database, S3-compatible storage, Redis + BullMQ for jobs, WebSockets or SSE for realtime status, FFmpeg for audio processing orchestration, OpenTelemetry and Sentry for observability, Stripe for payments, PostHog for analytics.

Implementation requirements:
Use a clean SaaS architecture with separate frontend, API, worker queue, and storage layers. Make the UI polished, minimal, expensive-looking, and optimized for audio workflow. All copy, labels, errors, and empty states must be in Russian. Include seed data and mock states for development. Prioritize responsive desktop UX but support mobile for monitoring status and exports. Build with production-ready components, typed models, and realistic loading, empty, error, and success states. Create the app structure, main pages, reusable components, sample data, and a sensible design system for a premium music product.