refactor: consolidate Docker files into docker/ and merge streaming repo

- Move full 4-service compose config/docker-compose.yml → docker/docker-compose.yml
- Move single-service quickstart docker-compose.yml → docker/docker-compose.simple.yml
- Replace private /home/sparky paths with /path/to/ placeholders in docker/docker-compose.yml
- Merge martinb78/qwen3-tts-streaming-dgx-spark into martinb78/faster-qwen3-tts-dgx-spark:streaming tag
- Update all image references: v4 → latest, streaming image → :streaming tag
- Update README and DOCKERHUB_STREAMING.md to reflect new structure

Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>

DOCKERHUB_STREAMING.md

@@ -0,0 +1,112 @@
+# Qwen3-TTS Streaming — DGX Spark (GB10)
+
+Low-latency, OpenAI-compatible streaming TTS server for the **NVIDIA DGX Spark GB10**
+(ARM64 / SM 121 / CUDA 13), powered by [faster-qwen3-tts](https://github.com/andimarafioti/faster-qwen3-tts)
+with CUDA graph acceleration.
+
+Streams WAV audio chunks to the client while generation is still running —
+first audio arrives in under a second for typical sentences.
+
+Part of a four-backend TTS stack documented on the NVIDIA Developer Forum:
+[Three times (VoiceClone | VoiceDesign | CustomVoice) — Faster-Qwen3-TTS for NVIDIA DGX Spark (GB10)](https://forums.developer.nvidia.com/t/three-times-voiceclone-voicedesign-customvoice-faster-qwen3-tts-for-nvidia-dgx-spark-gb10/370530)
+
+---
+
+## Quick start
+
+```bash
+docker run -d \
+  --runtime nvidia \
+  --name qwen3-tts-streaming \
+  -p 8023:8000 \
+  -e NVIDIA_VISIBLE_DEVICES=all \
+  -v /path/to/Qwen3-TTS-12Hz-1.7B-Base:/models/Qwen3-TTS:ro \
+  -v /path/to/faster-qwen3-tts/config:/config:rw \
+  -v /path/to/active_voices:/voices:ro \
+  martinb78/faster-qwen3-tts-dgx-spark:streaming \
+  /bin/bash -c "
+    python3 /config/generate_voices.py &&
+    python3 /config/run_server.py
+      --model /models/Qwen3-TTS
+      --voices /config/voices.json
+      --port 8000
+      --max-seq-len 4096
+  "
+```
+
+Check it's running:
+
+```bash
+curl http://localhost:8023/health
+```
+
+---
+
+## Voice configuration
+
+Create a `voices.json` in your config directory. Each entry maps a voice ID to a
+reference audio file and transcript:
+
+```json
+{
+  "william": {
+    "ref_audio": "/voices/william.wav",
+    "ref_text": "The quick brown fox jumps over the lazy dog.",
+    "language": "English",
+    "temperature": 0.75,
+    "top_k": 40,
+    "top_p": 0.85
+  },
+  "natasha": {
+    "ref_audio": "/voices/natasha.wav",
+    "ref_text": "She sells seashells by the seashore.",
+    "language": "English"
+  }
+}
+```
+
+`temperature`, `top_k`, and `top_p` are optional — defaults are `0.8 / 50 / 0.9`.
+
+---
+
+## API
+
+OpenAI-compatible `/v1/audio/speech`:
+
+```bash
+curl http://localhost:8023/v1/audio/speech \
+  -H "Content-Type: application/json" \
+  -d '{"model": "tts-1", "input": "Hello world!", "voice": "william", "response_format": "wav"}' \
+  --output speech.wav
+```
+
+| Endpoint | Method | Description |
+|---|---|---|
+| `/v1/audio/speech` | POST | Generate speech (WAV / PCM / MP3) |
+| `/v1/models` | GET | List available voice IDs |
+| `/v1/audio/voices` | GET | Voice list (OpenWebUI fallback) |
+| `/speakers` | GET | Voice list (SillyTavern) |
+| `/health` | GET | Liveness check |
+
+Works with **OpenWebUI**, **SillyTavern**, **llama-swap**, and any OpenAI-compatible client.
+
+---
+
+## Requirements
+
+- NVIDIA DGX Spark GB10 or another ARM64 system with CUDA 13
+- CUDA driver 580+
+- Docker + NVIDIA Container Toolkit
+- [Qwen3-TTS-12Hz-1.7B-Base](https://huggingface.co/Qwen/Qwen3-TTS-12Hz-1.7B-Base) weights downloaded locally
+
+---
+
+## Related images
+
+| Image | Description |
+|---|---|
+| `martinb78/faster-qwen3-tts-dgx-spark:latest` / `:v5` | VoiceClone, VoiceDesign, and CustomVoice backends |
+| `martinb78/faster-qwen3-tts-dgx-spark:streaming` | This tag — streaming VoiceClone |
+
+Full four-backend `docker-compose` setup in `docker/docker-compose.yml` and detailed documentation on GitHub and the
+[NVIDIA Developer Forum](https://forums.developer.nvidia.com/t/three-times-voiceclone-voicedesign-customvoice-faster-qwen3-tts-for-nvidia-dgx-spark-gb10/370530).

README.md

@@ -8,17 +8,17 @@ This repo packages the DGX Spark fixes plus four OpenAI-compatible TTS backends:
 
 | Backend | Port | Image | Voice source |
 |---|---:|---|---|
-| VoiceClone | `8020` | `martinb78/faster-qwen3-tts-dgx-spark:v5` | Reference audio plus transcript |
+| VoiceClone | `8020` | `martinb78/faster-qwen3-tts-dgx-spark:latest` | Reference audio plus transcript |
-| VoiceDesign | `8021` | `martinb78/faster-qwen3-tts-dgx-spark:v5` | Text prompt describes the voice; no reference needed |
+| VoiceDesign | `8021` | `martinb78/faster-qwen3-tts-dgx-spark:latest` | Text prompt describes the voice; no reference needed |
-| CustomVoice | `8022` | `martinb78/faster-qwen3-tts-dgx-spark:v5` | Separate CustomVoice model variant |
+| CustomVoice | `8022` | `martinb78/faster-qwen3-tts-dgx-spark:latest` | Separate CustomVoice model variant |
-| Streaming | `8023` | `martinb78/qwen3-tts-streaming-dgx-spark:latest` | Same voices as `8020`, but streams WAV chunks while generating |
+| Streaming | `8023` | `martinb78/faster-qwen3-tts-dgx-spark:streaming` | Same voices as `8020`, but streams WAV chunks while generating |
 
 All four backends expose the OpenAI `/v1/audio/speech` contract and work with **OpenWebUI**, **SillyTavern**, **llama-swap**, `curl`, or any OpenAI-compatible client.
 
-Both Docker images are published and publicly available:
+One Docker image covers all four backends:
 
-- `martinb78/faster-qwen3-tts-dgx-spark:v5` - used by VoiceClone, VoiceDesign, and CustomVoice.
+- `martinb78/faster-qwen3-tts-dgx-spark:v5` (or `:latest`) — VoiceClone, VoiceDesign, and CustomVoice.
-- `martinb78/qwen3-tts-streaming-dgx-spark:latest` - used by the streaming service.
+- `martinb78/faster-qwen3-tts-dgx-spark:streaming` — streaming VoiceClone.
 
 ## What this solves
 
@@ -31,7 +31,7 @@ The DGX Spark GB10 has a unique ARM64 Grace CPU plus Blackwell GPU stack (SM 121
 
 ## Quick start: VoiceClone only
 
-Use the root `docker-compose.yml` when you only need voice cloning on port `8020`.
+Use `docker/docker-compose.simple.yml` when you only need voice cloning on port `8020`.
 
 ```bash
 docker pull martinb78/faster-qwen3-tts-dgx-spark:latest
@@ -39,11 +39,9 @@ docker pull martinb78/faster-qwen3-tts-dgx-spark:latest
 mkdir -p models
 huggingface-cli download Qwen/Qwen3-TTS-12Hz-1.7B-Base --local-dir ./models/Qwen3-TTS
 
-cp .env.example .env
-# Edit .env and set MODEL_PATH to your local Qwen3-TTS-12Hz-1.7B-Base directory.
-
 # Add reference audio and transcripts to config/speakers/ first.
-docker compose up -d
+cd docker
+MODEL_PATH=/path/to/Qwen3-TTS-12Hz-1.7B-Base docker compose -f docker-compose.simple.yml up -d
 ```
 
 Build the image locally instead of pulling Docker Hub:
@@ -66,7 +64,7 @@ curl http://localhost:8020/health
 
 ## Full stack: VoiceClone, VoiceDesign, CustomVoice, Streaming
 
-Use `config/docker-compose.yml` when you want all four OpenAI-compatible backends side by side:
+Use `docker/docker-compose.yml` when you want all four OpenAI-compatible backends side by side:
 
 ```text
 8020  ->  VoiceClone   (/v1/audio/speech, reference audio)
@@ -83,21 +81,21 @@ huggingface-cli download Qwen/Qwen3-TTS-12Hz-1.7B-VoiceDesign --local-dir /path/
 huggingface-cli download Qwen/Qwen3-TTS-12Hz-1.7B-CustomVoice --local-dir /path/to/Qwen3-TTS-12Hz-1.7B-CustomVoice
 ```
 
-2. Edit `config/docker-compose.yml` and adjust the volume paths for your machine:
+2. Edit `docker/docker-compose.yml` and adjust the volume paths for your machine:
 
 ```yaml
 volumes:
   - /path/to/Qwen3-TTS-12Hz-1.7B-Base:/models/Qwen3-TTS:ro
   - /path/to/Qwen3-TTS-12Hz-1.7B-VoiceDesign:/models/Qwen3-TTS-VoiceDesign:ro
   - /path/to/Qwen3-TTS-12Hz-1.7B-CustomVoice:/models/Qwen3-TTS-CustomVoice:ro
-  - /path/to/this/repo/config:/config:rw
+  - /path/to/faster-qwen3-tts/config:/config:rw
 ```
 
 3. Make sure the external Docker network exists, then start the stack:
 
 ```bash
 docker network create dgx_net 2>/dev/null || true
-cd config
+cd docker
 docker compose up -d
 ```
 
@@ -338,6 +336,13 @@ The first request after container startup can be slower because CUDA graph captu
 
 ## Changelog
 
+### v6 — 2026-05-30
+**Consolidate Docker files into `docker/` folder**
+
+- Moved full 4-service compose from `config/docker-compose.yml` → `docker/docker-compose.yml`
+- Moved single-service quickstart from root `docker-compose.yml` → `docker/docker-compose.simple.yml`
+- Consolidated streaming image into `martinb78/faster-qwen3-tts-dgx-spark:streaming` tag; removed separate `qwen3-tts-streaming-dgx-spark` repository
+
 ### v5 — 2026-05-30
 **Fix: voice drifts and gender changes on long paragraphs (VoiceClone)**
 

docker/docker-compose.simple.yml

@@ -12,11 +12,11 @@ services:
       - "8020:8000"
     volumes:
       # Mount your local Qwen3-TTS model (download first, see README)
-      - ${MODEL_PATH:-./models/Qwen3-TTS}:/models/Qwen3-TTS:ro
+      - ${MODEL_PATH:-../models/Qwen3-TTS}:/models/Qwen3-TTS:ro
       # Config directory (voices.json generated on startup)
-      - ./config:/config:rw
+      - ../config:/config:rw
       # Speaker reference audio files
-      - ./config/speakers:/voices:ro
+      - ../config/speakers:/voices:ro
     command: >
       /bin/bash -c "
       python3 /config/generate_voices.py &&

docker/docker-compose.yml

@@ -7,9 +7,12 @@
 #    8022 -> CustomVoice model
 #    8023 -> Low-latency streaming voice clone endpoint
 #
-#  Voice files:
+#  Before starting: adjust the volume paths below to match your setup.
-#    The voice clone and streaming services scan active voices from:
+#    - /path/to/Qwen3-TTS-12Hz-1.7B-Base       -> local Qwen3-TTS base model
-#    /home/sparky/Projekte/TTS_Voices/active_voices
+#    - /path/to/Qwen3-TTS-12Hz-1.7B-VoiceDesign -> local VoiceDesign model
+#    - /path/to/Qwen3-TTS-12Hz-1.7B-CustomVoice -> local CustomVoice model
+#    - /path/to/faster-qwen3-tts/config          -> this repo's config directory
+#    - /path/to/active_voices                    -> directory with reference .wav files
 #
 #  Portainer note:
 #    Set HF_TOKEN in the stack environment if the models ever need Hugging Face access.
@@ -18,7 +21,7 @@
 services:
   # Standard voice cloning backend used by TTS Voice Creator and other OpenAI-compatible clients.
   faster-qwen3-tts-voiceclone:
-    image: martinb78/faster-qwen3-tts-dgx-spark:v4
+    image: martinb78/faster-qwen3-tts-dgx-spark:latest
     container_name: faster-qwen3-tts-voiceclone
     restart: unless-stopped
     runtime: nvidia
@@ -28,9 +31,9 @@ services:
     ports:
       - "8020:8000"
     volumes:
-      - /home/sparky/LLMs/vllm/Alibaba/Qwen3-TTS-12Hz-1.7B-Base:/models/Qwen3-TTS:ro
+      - /path/to/Qwen3-TTS-12Hz-1.7B-Base:/models/Qwen3-TTS:ro
-      - /home/sparky/Docker/faster-qwen3-tts/config:/config:rw
+      - /path/to/faster-qwen3-tts/config:/config:rw
-      - /home/sparky/Projekte/TTS_Voices/active_voices:/voices:ro
+      - /path/to/active_voices:/voices:ro
     command: >
       /bin/bash -c "
       python3 /config/generate_voices.py &&
@@ -52,7 +55,7 @@ services:
 
   # VoiceDesign backend: creates speech from a descriptive voice prompt instead of a reference clip.
   faster-qwen3-tts-voicedesign:
-    image: martinb78/faster-qwen3-tts-dgx-spark:v4
+    image: martinb78/faster-qwen3-tts-dgx-spark:latest
     container_name: faster-qwen3-tts-voicedesign
     restart: unless-stopped
     runtime: nvidia
@@ -62,8 +65,8 @@ services:
     ports:
       - "8021:8000"
     volumes:
-      - /home/sparky/LLMs/vllm/Alibaba/Qwen3-TTS-12Hz-1.7B-VoiceDesign:/models/Qwen3-TTS-VoiceDesign:ro
+      - /path/to/Qwen3-TTS-12Hz-1.7B-VoiceDesign:/models/Qwen3-TTS-VoiceDesign:ro
-      - /home/sparky/Docker/faster-qwen3-tts/config:/config:rw
+      - /path/to/faster-qwen3-tts/config:/config:rw
     command: >
       /bin/bash -c "
       python3 /config/run_voicedesign_server.py
@@ -84,7 +87,7 @@ services:
 
   # CustomVoice backend: keeps the separate CustomVoice model available on port 8022.
   faster-qwen3-tts-customvoice:
-    image: martinb78/faster-qwen3-tts-dgx-spark:v4
+    image: martinb78/faster-qwen3-tts-dgx-spark:latest
     container_name: faster-qwen3-tts-customvoice
     restart: unless-stopped
     runtime: nvidia
@@ -94,8 +97,8 @@ services:
     ports:
       - "8022:8000"
     volumes:
-      - /home/sparky/LLMs/vllm/Alibaba/Qwen3-TTS-12Hz-1.7B-CustomVoice:/models/Qwen3-TTS-CustomVoice:ro
+      - /path/to/Qwen3-TTS-12Hz-1.7B-CustomVoice:/models/Qwen3-TTS-CustomVoice:ro
-      - /home/sparky/Docker/faster-qwen3-tts/config:/config:rw
+      - /path/to/faster-qwen3-tts/config:/config:rw
     command: >
       /bin/bash -c "
       python3 /config/run_customvoice_server.py
@@ -116,7 +119,7 @@ services:
 
   # Streaming voice clone backend: same active voices as 8020, but streams WAV chunks while generating.
   faster-qwen3-tts-streaming:
-    image: martinb78/qwen3-tts-streaming-dgx-spark:latest
+    image: martinb78/faster-qwen3-tts-dgx-spark:streaming
     container_name: faster-qwen3-tts-streaming
     restart: unless-stopped
     runtime: nvidia
@@ -131,9 +134,9 @@ services:
     ports:
       - "8023:8000"
     volumes:
-      - /home/sparky/LLMs/vllm/Alibaba/Qwen3-TTS-12Hz-1.7B-Base:/models/Qwen3-TTS:ro
+      - /path/to/Qwen3-TTS-12Hz-1.7B-Base:/models/Qwen3-TTS:ro
-      - /home/sparky/Docker/faster-qwen3-tts/config:/config:rw
+      - /path/to/faster-qwen3-tts/config:/config:rw
-      - /home/sparky/Projekte/TTS_Voices/active_voices:/voices:ro
+      - /path/to/active_voices:/voices:ro
     command: >
       /bin/bash -c "
       python3 /config/generate_voices.py &&