remiqora
Remiqora is a local GPU-powered music studio unifying ACE-Step 1.5 and YuE2-3B in one Vue interface with built-in multitrack DAW, stem separation, MIDI transcription, and LoRA fine-tuning.
Ficha actualizada 23 sept 2026, 12:35 · upstream 23 sept 2026, 12:18 · release sin release detectada
Qué es remiqora
Descripción editorial construida a partir del repositorio, su documentación y sus señales de mantenimiento.
Remiqora solves the problem of running two independent music generation engines (ACE-Step and YuE2) simultaneously on a single consumer GPU. The platform provides a unified user interface that orchestrates both models with mutual exclusion — when one model starts, the other stops automatically. All tracks are stored in a shared SQLite database accessible from every module: Demucs for stem separation, MuScriptor for MIDI transcription, and the built-in DAW for mixing. A key feature is the integrated LoRA training pipeline that allows users to fine-tune ACE-Step on their own voice or style directly from the browser without needing a console environment.
The architecture consists of two backend services: FastAPI (Python) handles model orchestration, process lifecycle management, and API routing for both models. The frontend is built with Vue 3 + TypeScript using Tailwind CSS v4, Pinia state management, and vue-router for navigation. Audio processing runs entirely in the browser via a custom Web Audio engine that includes a mixer, timeline editor, effects rack, MIDI parser, and synth engine.
The platform supports both text-to-music generation (ACE-Step) and full-length composition with CoT planning (YuE2), along with stem separation, MIDI transcription, and LoRA fine-tuning. The interface is fully bilingual (English/Russian) with a system language switcher that overrides the header toggle.
Remiqora is currently in active development with breaking changes expected. It targets Windows, macOS, and Linux platforms.
- Generate complete songs from text prompts using ACE-Step or YuE2-3B models
- Create remixes of existing tracks by selecting a reference and applying transformations (cover, repaint sections, extract parts)
- Separate vocals, drums, bass, and other instruments from any uploaded track using Demucs
- Transcribe audio recordings into MIDI notation for sheet music or instrument practice
- Fine-tune ACE-Step on custom datasets to develop a unique voice or musical style
- Build complete multitrack projects in the built-in DAW with effects, BPM control, and export options
- Single unified interface instead of two separate interfaces for ACE-Step and YuE2
- Mutually exclusive model orchestration — one process starts while the other stops automatically
- Shared SQLite database stores all tracks regardless of how they were created (generation, upload, or DAW assembly)
- Built-in LoRA training pipeline allows fine-tuning ACE-Step on user-provided datasets without a console environment
- Full multitrack DAW with real-time effects, stereo mastering, MIDI playback, and loop editing
- Fully bilingual interface (English/Russian) with system language override capability
- Currently only supports Windows, macOS, and Linux platforms — no mobile or web browser support yet
- ACE-Step generation is limited to ~300 seconds duration per job; longer compositions require multiple jobs
- YuE2 CoT planning mode requires YuE2 to be the active model while other models are stopped
- LoRA training requires a GPU with sufficient VRAM (at least 8GB recommended) and local storage for datasets
- MIDI transcription is only available when YuE2 is running as the active model, not independently
- The built-in DAW lacks advanced features like granular editing, complex automation, or multi-track playback
Capturas del proyecto
Imágenes reales detectadas en la documentación del repositorio.
Perfil de infraestructura
Estimación inicial derivada del repositorio y de los servicios detectados.
Despliegue detectado
Qué señales reales de instalación hemos encontrado.
Remiqora uses FastAPI (Python) for the backend and Vue.js frontend. The architecture separates concerns: `app/orchestrator/` manages model process lifecycles with mutual exclusion on a single GPU via IPC communication. The shared SQLite database is organized per-model, so track data persists across different generation or editing sessions. For deployment, install FastAPI and uvicorn on the backend server, then serve the Vue frontend from a local directory (e.g., `npm run dev`). The API endpoints `/api/ace/*` and `/api/yue2/*` are reverse-proxied through `app/api/routes_proxy.py`. For production use, consider containerizing with Docker Compose or deploying to a cloud platform like AWS ECS. Note: The project is actively in development with breaking changes expected. Always check the GitHub repository for the latest version before using it.
Stack y dependencias
Componentes inferidos a partir de manifests, configuración y documentación.
Puede sustituir a
Relaciones detectadas por el análisis del repositorio y su categoría.
Experiencia real de la comunidad
Ayuda a contrastar la estimación automática con instalaciones reales.
Todavía no hay datos comunitarios para este proyecto.