Add Seed-VC and Applio studio modes

2026-09-09 20:58:32 +02:00
parent 51ed201f6c
commit 13a6714b2b
16 changed files with 410 additions and 33 deletions
@@ -0,0 +1,25 @@
+# syntax=docker/dockerfile:1
+FROM python:3.12-trixie
+
+ARG APPLIO_COMMIT=7fa68ec2166ab1331c539704159fa14901e94e5a
+ENV PATH=/app/.venv/bin:$PATH \
+    HF_HOME=/models/huggingface \
+    PIP_DISABLE_PIP_VERSION_CHECK=1
+
+RUN apt-get update && apt-get install -y --no-install-recommends \
+      ca-certificates curl ffmpeg git libportaudio2 \
+    && rm -rf /var/lib/apt/lists/*
+
+WORKDIR /app
+RUN git clone https://github.com/IAHispano/Applio.git . \
+    && git checkout "$APPLIO_COMMIT" \
+    && python3 -m venv /app/.venv \
+    && pip install --no-cache-dir --upgrade pip \
+    && pip install --no-cache-dir python-ffmpeg \
+    && pip install --no-cache-dir torch==2.7.1 torchvision==0.22.1 torchaudio==2.7.1 \
+       --index-url https://download.pytorch.org/whl/cu128 \
+    && sed -i '/^torch==/d;/^torchvision==/d;/^torchaudio==/d' requirements.txt \
+    && pip install --no-cache-dir -r requirements.txt
+
+EXPOSE 6969
+CMD ["python3", "app.py", "--server-name", "0.0.0.0", "--port", "6969"]
@@ -0,0 +1,14 @@
+# Applio / RVC Studio
+
+Reproduzierbarer, experimenteller Applio-Worker mit der offiziellen
+Weboberfläche. Der Build ist auf Upstream-Commit
+`7fa68ec2166ab1331c539704159fa14901e94e5a` festgeschrieben.
+
+- Dashboard-Modus: `Applio / RVC`
+- WireGuard-URL: `http://192.168.1.212:8011/`
+- GPU: RTX 5080, exklusiv zu LLM, Musik- und anderen Voice-Modi
+- Persistenz: Basisgewichte, importierte/trainierte Modelle, Konfiguration,
+  Logs und Hugging-Face-Cache unter `/data/voice/applio`
+
+Applio stellt die RVC-Werkzeuge und deren Oberfläche bereit. Eine konkrete
+Zielstimme wird anschließend in der Oberfläche importiert oder trainiert.
@@ -0,0 +1,41 @@
+services:
+  applio-studio:
+    build: .
+    image: mike-ai/applio-studio:7fa68ec
+    container_name: mike-ai-applio-studio
+    restart: "no"
+    labels:
+      com.mike-ai.applio-worker: applio
+    environment:
+      NVIDIA_VISIBLE_DEVICES: ${VOICE_GPU_UUID:?set VOICE_GPU_UUID to the RTX 5080 UUID}
+      NVIDIA_DRIVER_CAPABILITIES: compute,utility
+      HF_HOME: /models/huggingface
+      PYTORCH_CUDA_ALLOC_CONF: expandable_segments:True
+    ports:
+      - "127.0.0.1:8011:6969"
+    volumes:
+      - /data/voice/applio/huggingface:/models/huggingface
+      - /data/voice/applio/logs:/app/logs
+      - /data/voice/applio/models:/app/rvc/models
+      - /data/voice/applio/config.json:/app/assets/config.json
+    healthcheck:
+      test: ["CMD-SHELL", "curl -fsS http://127.0.0.1:6969/ >/dev/null"]
+      interval: 5s
+      timeout: 3s
+      start_period: 900s
+      retries: 3
+    deploy:
+      resources:
+        reservations:
+          devices:
+            - driver: nvidia
+              device_ids: ["${VOICE_GPU_UUID:?set VOICE_GPU_UUID to the RTX 5080 UUID}"]
+              capabilities: [gpu]
+    networks:
+      frontend:
+        aliases: [applio-studio]
+
+networks:
+  frontend:
+    name: mike-ai_frontend
+    external: true
@@ -0,0 +1,23 @@
+FROM nvidia/cuda:12.8.1-cudnn-runtime-ubuntu22.04
+
+ARG SEED_VC_COMMIT=51383efd921027683c89e5348211d93ff12ac2a8
+ENV DEBIAN_FRONTEND=noninteractive \
+    PIP_DISABLE_PIP_VERSION_CHECK=1 \
+    HF_HOME=/models/huggingface
+
+RUN apt-get update && apt-get install -y --no-install-recommends \
+      build-essential ca-certificates curl ffmpeg git libsndfile1 python3 python3-dev python3-pip \
+    && rm -rf /var/lib/apt/lists/*
+
+WORKDIR /app
+RUN git clone https://github.com/Plachtaa/seed-vc.git . \
+    && git checkout "$SEED_VC_COMMIT" \
+    && sed -i '/^torch\($\|[= <>=]\)/d;/^torchvision\($\|[= <>=]\)/d;/^torchaudio\($\|[= <>=]\)/d' requirements.txt \
+    && python3 -m pip install --no-cache-dir \
+       torch==2.7.1 torchvision==0.22.1 torchaudio==2.7.1 \
+       --index-url https://download.pytorch.org/whl/cu128 \
+    && python3 -m pip install --no-cache-dir -r requirements.txt \
+    && sed -i 's/demo\.launch()/demo.launch(server_name="0.0.0.0", server_port=7860)/' app.py
+
+EXPOSE 7860
+CMD ["python3", "app.py", "--enable-v1"]
@@ -0,0 +1,14 @@
+# Seed-VC Studio
+
+Reproduzierbarer, experimenteller Seed-VC-v2-Worker mit der offiziellen
+Gradio-Oberfläche. Der Build ist auf Upstream-Commit
+`51383efd921027683c89e5348211d93ff12ac2a8` festgeschrieben.
+
+- Dashboard-Modus: `Seed-VC`
+- WireGuard-URL: `http://192.168.1.212:8010/`
+- GPU: RTX 5080, exklusiv zu LLM, Musik- und anderen Voice-Modi
+- Persistenz: Hugging-Face-Cache unter `/data/huggingface`
+
+Das Image enthält den Programmcode. Die benötigten Modellgewichte werden beim
+ersten tatsächlichen Einsatz in den persistenten Cache geladen.
+
@@ -0,0 +1,39 @@
+services:
+  seed-vc-studio:
+    build: .
+    image: mike-ai/seed-vc-studio:51383efd
+    container_name: mike-ai-seed-vc-studio
+    restart: "no"
+    labels:
+      com.mike-ai.seed-vc-worker: seed-vc
+    environment:
+      NVIDIA_VISIBLE_DEVICES: ${VOICE_GPU_UUID:?set VOICE_GPU_UUID to the RTX 5080 UUID}
+      NVIDIA_DRIVER_CAPABILITIES: compute,utility
+      HF_HOME: /models/huggingface
+      PYTORCH_CUDA_ALLOC_CONF: expandable_segments:True
+    ports:
+      - "127.0.0.1:8010:7860"
+    volumes:
+      - /data/voice/seed-vc/huggingface:/models/huggingface
+      - /data/voice/seed-vc/output:/app/results
+    healthcheck:
+      test: ["CMD-SHELL", "curl -fsS http://127.0.0.1:7860/ >/dev/null"]
+      interval: 5s
+      timeout: 3s
+      start_period: 900s
+      retries: 3
+    deploy:
+      resources:
+        reservations:
+          devices:
+            - driver: nvidia
+              device_ids: ["${VOICE_GPU_UUID:?set VOICE_GPU_UUID to the RTX 5080 UUID}"]
+              capabilities: [gpu]
+    networks:
+      frontend:
+        aliases: [seed-vc-studio]
+
+networks:
+  frontend:
+    name: mike-ai_frontend
+    external: true