Update Docker Docs and Examples (Published Images, GGUF, Qwen3 TTS)

README.md

@@ -103,8 +103,8 @@ PocketTTS language selection is a model-load option. When the model path points
 
 ## Docker
 
-Docker CPU and CUDA images are available for both CLI and server use. See [Docker.md](Docker.md) for build commands and working Docker examples.
-
+Docker CUDA and CPU images are available for both CLI and server use. See [docker.md](docs/docker.md) for
+available images, build commands and working Docker examples.
 
 ## Model Manager and GGUF Downloads
 

docs/docker.md

@@ -1,68 +1,111 @@
 # Running audio.cpp in Docker
 
+## Table of Contents
+
+- [Prerequisites](#prerequisites)
+- [Image Variants](#image-variants)
+- [Published Images](#published-images)
+- [Build Images locally](#build-images-locally)
+- [Usage](#usage)
+- [Examples](#examples)
+
 ## Prerequisites
 
-- Docker must be installed and running on your system
+- Docker must be installed and running on your system.
 - For CUDA:
-  - The [NVIDIA container toolkit](https://github.com/NVIDIA/nvidia-container-toolkit) must be installed
+  - The [NVIDIA container toolkit](https://github.com/NVIDIA/nvidia-container-toolkit) must be installed.
+
+## Image Variants
+
+The following image variants are available:
+
+- **full**: Provides the main tools **cli** and **server** and test binaries in one image. When running the container, the first argument selects the tool to execute.
+
+The following backends are supported:
+- **cuda12**
+- **cuda13**
+- **cpu**
+
+The following architectures are supported:
+- **amd64**
+- **arm64**
 
-## Images
+## Published Images
 
-The following image flavors are available for this project:
-- **full**: This image provides the main tools ***cli*** and ***server***
-  and test binaries in one image. When running the container the first argument
-  selects the tool to execute.
+Docker images are published daily when new commits are available. The images are provided
+as multiarch images (amd64/arm64).
 
-## Build the image
+Pull the latest images using these tags:
+- **cuda12**: `ghcr.io/0xshug0/audio.cpp:full-cuda12`
+- **cuda13**: `ghcr.io/0xshug0/audio.cpp:full-cuda13`
+- **cpu**: `ghcr.io/0xshug0/audio.cpp:full-cpu`
+
+Images for a specific day/commit can be found in the
+[versions](https://github.com/0xShug0/audio.cpp/pkgs/container/audio.cpp/versions?filters%5Bversion_type%5D=tagged)
+history.
+The format is: `full-<backend>-<date>-<shortsha>`, e.g. `full-cuda12-20260725-db7d2c4`
+
+
+## Build Images locally
+
+If you would like to build the images locally, you can use the available
+Dockerfiles in `.devops`.
 
 ### CUDA
 
-Build with default CUDA 12.x version (see *.devops/cuda.Dockerfile*):
+Build with the default CUDA 12.x version. See `.devops/cuda.Dockerfile`.
 
 ```bash
-docker build -f .devops/cuda.Dockerfile -t local/audiocpp:full-cuda .
+docker build -f .devops/cuda.Dockerfile -t local/audio.cpp:full-cuda12 .
 ```
 
-Build with a specific CUDA version, e.g. 13.2.0:
+Build with a specific CUDA version, for example 13.3.0:
+
 ```bash
-docker build -f .devops/cuda.Dockerfile -t local/audiocpp:full-cuda --build-arg CUDA_VERSION=13.2.0 .
+docker build -f .devops/cuda.Dockerfile -t local/audio.cpp:full-cuda13 --build-arg CUDA_VERSION=13.3.0 .
 ```
 
 ### CPU
 
 ```bash
-docker build -f .devops/cpu.Dockerfile -t local/audiocpp:full-cpu .
+docker build -f .devops/cpu.Dockerfile -t local/audio.cpp:full-cpu .
 ```
 
 ## Usage
 
 The model directory `<models-dir>` must be mounted into the container.
-An additional `<output-dir>` should be mounted for TTS tasks. 
+An additional `<output-dir>` should be mounted for TTS tasks.
 
 ### CUDA
+
 ```bash
-docker run --rm --gpus all -v "<models-dir>:/models:ro" local/audiocpp:full-cuda <cli|server> <...>
+docker run --rm --gpus all -v "<models-dir>:/models:ro" ghcr.io/0xshug0/audio.cpp:full-cuda12 <cli|server> --model /models/<model> <...>
 ```
 
 ### CPU
+
 ```bash
-docker run --rm -v "<models-dir>:/models:ro" local/audiocpp:full-cpu <cli|server> <...>
+docker run --rm -v "<models-dir>:/models:ro" ghcr.io/0xshug0/audio.cpp:full-cpu <cli|server> --model /models/<model> <...>
 ```
 
-See the fully working [examples](#examples) below. 
+See the fully working [examples](#examples) below.
 
 ## Examples
 
-Examples for Docker (CUDA and CPU) are available in `examples/docker`.
+Examples for Docker, including CUDA and CPU, are available in `examples/docker`.
 
-### Pocket TTS via docker run
+### CLI
 
-Generate speech via audio.cpp cli running in Docker.
+The **[examples](../examples/docker/cli/EXAMPLE.md)** in `examples/docker/cli`
+demonstrate how to run the audio.cpp CLI with `docker run`. The examples include:
 
-See [EXAMPLE.md](../examples/docker/cli/EXAMPLE.md) in `examples/docker/cli`.
+- **PocketTTS:** Text-to-Speech
+- **Qwen3-TTS:** Text-to-Speech with Voice Cloning
 
-### Pocket TTS via Docker Compose
+### Server
 
-Generate speech via audio.cpp server running in Docker (Compose).
+The **[examples](../examples/docker/server/EXAMPLE.md)** in `examples/docker/server`
+demonstrate how to run the audio.cpp server with `docker compose`. The examples include:
 
-See [EXAMPLE.md](../examples/docker/server/EXAMPLE.md) in `examples/docker/server`.
+- **PocketTTS:** Text-to-Speech
+- **Qwen3-TTS:** Text-to-Speech with Voice Cloning

examples/docker/cli/EXAMPLE.md

@@ -2,47 +2,72 @@
 
 Run the audio.cpp CLI tool inside a Docker container for text-to-speech.
 
-## Setup
+## PocketTTS
 
-**1. Download the PocketTTS model**
+### 1. Download the PocketTTS model
 
-Get the English model from [kyutai/pocket-tts](https://huggingface.co/kyutai/pocket-tts/) on Hugging Face.
-Place the files from `languages/english/` into `../models/pocket-tts/languages/english/`:
+Get the English q8 model from
+[audio-cpp/audio.cpp-gguf](https://huggingface.co/audio-cpp/audio.cpp-gguf/tree/main/PocketTTS-GGUF)
+on Hugging Face.
+Place the files from `english/` into `../models/PocketTTS-GGUF/english/`:
 
 The directory should look like:
 ```
-../models/pocket-tts/languages/english/
-├── model.safetensors
-├── tokenizer.model
+../models/PocketTTS-GGUF/english/
+├── pocket-tts-english-q8_0.gguf
 └── embeddings/
     ├── alba.safetensors
     └── ...
 ```
 
-**2. Build the image**
+### 2. Run
 
-Run from repository root:
-
-CPU:
+Run one of:
 
 ```bash
-docker build -f .devops/cpu.Dockerfile -t local/audiocpp:full-cpu .
+./pocket-tts-cuda12.sh
+./pocket-tts-cuda13.sh
+./pocket-tts-cpu.sh
 ```
 
-GPU (CUDA):
+See the scripts for details. The scripts reference the published docker image
+and save the generated speech to `output/speech.wav`.
 
-```bash
-docker build -f .devops/cuda.Dockerfile -t local/audiocpp:full-cuda .
+## Qwen3-TTS (with voice cloning)
+
+### 1. Download the Qwen3-TTS model
+
+Get the Qwen3-TTS-12Hz-1.7B-Base-GGUF q8 model from
+[audio-cpp/audio.cpp-gguf](https://huggingface.co/audio-cpp/audio.cpp-gguf/tree/main/Qwen3-TTS-12Hz-1.7B-Base-GGUF)
+on Hugging Face.
+Place the file into `../models/Qwen3-TTS-12Hz-1.7B-Base-GGUF/`:
+
+The directory should look like:
+```
+../models/Qwen3-TTS-12Hz-1.7B-Base-GGUF/
+└── qwen3-tts-12hz-1.7b-base-q8_0_v2.gguf
 ```
 
-## Usage
+### 2. Add reference audio and transcription
+
+Put a `ref_audio.wav` and `ref_text.txt` in `../references/`.
+
+The directory should look like:
+```
+../references/
+├── ref_audio.wav
+└── ref_text.txt
+```
+
+### 3. Run
 
 Run one of:
 
 ```bash
-./cpu-tts.sh
-./cuda-tts.sh
+./qwen3-tts-cuda12.sh
+./qwen3-tts-cuda13.sh
+./qwen3-tts-cpu.sh
 ```
 
-Speech is saved to `output/speech.wav`.
-
+See the scripts for details. The scripts reference the published docker image
+and save the generated speech based on the provided reference audio in `output/speech.wav`.

examples/docker/cli/cpu-tts.sh → examples/docker/cli/pocket-tts-cpu.sh

@@ -9,11 +9,11 @@ docker run --rm \
   --user "$(id -u):$(id -g)" \
   -v "$MODEL_DIR:/models:ro" \
   -v "$OUTPUT_DIR:/output" \
-  local/audiocpp:full-cpu \
+  ghcr.io/0xshug0/audio.cpp:full-cpu \
   cli \
   --task tts \
   --family pocket_tts \
-  --model /models/pocket-tts \
+  --model /models/PocketTTS-GGUF/english/pocket-tts-english-q8_0.gguf \
   --backend cpu \
   --text "You are successfully running a text-to-speech model using audio.cpp, a pure C++ inference engine for audio models." \
   --voice-id alba \

examples/docker/cli/cuda-tts.sh → examples/docker/cli/pocket-tts-cuda12.sh

@@ -10,11 +10,11 @@ docker run --rm \
   --user "$(id -u):$(id -g)" \
   -v "$MODEL_DIR:/models:ro" \
   -v "$OUTPUT_DIR:/output" \
-  local/audiocpp:full-cuda \
+  ghcr.io/0xshug0/audio.cpp:full-cuda12 \
   cli \
   --task tts \
   --family pocket_tts \
-  --model /models/pocket-tts \
+  --model /models/PocketTTS-GGUF/english/pocket-tts-english-q8_0.gguf \
   --backend cuda \
   --device 0 \
   --text "You are successfully running a text-to-speech model using audio.cpp, a pure C++ inference engine for audio models." \

examples/docker/cli/pocket-tts-cuda13.sh

@@ -0,0 +1,24 @@
+#!/usr/bin/env bash
+set -e
+
+OUTPUT_DIR="$(cd "$(dirname "$0")" && pwd)/output"
+mkdir -p "$OUTPUT_DIR"
+MODEL_DIR="$(cd "$(dirname "$0")/../models" && pwd)"
+
+docker run --rm \
+  --gpus all \
+  --user "$(id -u):$(id -g)" \
+  -v "$MODEL_DIR:/models:ro" \
+  -v "$OUTPUT_DIR:/output" \
+  ghcr.io/0xshug0/audio.cpp:full-cuda13 \
+  cli \
+  --task tts \
+  --family pocket_tts \
+  --model /models/PocketTTS-GGUF/english/pocket-tts-english-q8_0.gguf \
+  --backend cuda \
+  --device 0 \
+  --text "You are successfully running a text-to-speech model using audio.cpp, a pure C++ inference engine for audio models." \
+  --voice-id alba \
+  --out /output/speech.wav
+
+echo "Saved to: $OUTPUT_DIR/speech.wav"

examples/docker/cli/qwen3-tts-cpu.sh

@@ -0,0 +1,25 @@
+#!/usr/bin/env bash
+set -e
+
+OUTPUT_DIR="$(cd "$(dirname "$0")" && pwd)/output"
+mkdir -p "$OUTPUT_DIR"
+MODEL_DIR="$(cd "$(dirname "$0")/../models" && pwd)"
+REFERENCES_DIR="$(cd "$(dirname "$0")/../references" && pwd)"
+
+docker run --rm \
+  --user "$(id -u):$(id -g)" \
+  -v "$MODEL_DIR:/models:ro" \
+  -v "$REFERENCES_DIR:/references:ro" \
+  -v "$OUTPUT_DIR:/output" \
+  ghcr.io/0xshug0/audio.cpp:full-cpu \
+  cli \
+  --task tts \
+  --family qwen3_tts \
+  --model /models/Qwen3-TTS-12Hz-1.7B-Base-GGUF/qwen3-tts-12hz-1.7b-base-q8_0_v2.gguf \
+  --backend cpu \
+  --text "You are successfully running a text-to-speech model using audio.cpp, a pure C++ inference engine for audio models." \
+  --voice-ref /references/ref_audio.wav \
+  --reference-text "$(cat "$REFERENCES_DIR/ref_text.txt")" \
+  --out /output/speech.wav
+
+echo "Saved to: $OUTPUT_DIR/speech.wav"