Added performance test values

README.md

@@ -27,12 +27,28 @@ The HiFi-GAN vocoder (`hift.pt`) is used unchanged from the base model.
 
 ---
 
-## Requirements
+## Quickstart with Docker
 
-### 1. Clone CosyVoice at the correct commit
+The easiest way to use this model is via the official Docker container:
+
+```bash
+docker run -p 8000:8000 \
+  -v cosyvoice_models:/app/CosyVoice/pretrained_models \
+  thorstenvoice/cosyvoice-tts
+
+# Then generate audio:
+curl -X POST http://localhost:8000/tts \
+     -F "text=Hallo, ich bin Thorsten. Schön, dass du da bist." \
+     --output thorsten.wav
+```
+
+→ [Docker Hub: thorstenvoice/cosyvoice-tts](https://hub.docker.com/r/thorstenvoice/cosyvoice-tts)
+
+---
 
-This model was trained and tested against a specific version of CosyVoice.
-Using a different commit may cause compatibility issues.
+## Manual Installation
+
+### 1. Clone CosyVoice at the correct commit
 
 ```bash
 git clone https://github.com/FunAudioLLM/CosyVoice.git
@@ -43,17 +59,12 @@ git submodule update --init --recursive
 
 ### 2. Install dependencies
 
-Python 3.10 or 3.11 is recommended. With Python 3.12 additional patches are required (see below).
+Python 3.10 or 3.11 recommended.
 
 ```bash
-# System dependencies
 sudo apt-get install -y sox libsox-fmt-all ffmpeg
-
-# Install openai-whisper first (before requirements.txt)
 pip install setuptools --upgrade
 pip install openai-whisper
-
-# Install remaining dependencies
 grep -v "openai-whisper" requirements.txt > requirements_fixed.txt
 pip install -r requirements_fixed.txt
 ```
@@ -64,31 +75,22 @@ pip install -r requirements_fixed.txt
 export PYTHONPATH=/path/to/CosyVoice:/path/to/CosyVoice/third_party/Matcha-TTS:$PYTHONPATH
 ```
 
-### 4. Download the base model
+### 4. Download models
 
 ```bash
 pip install huggingface_hub
 
+# Base model
 hf download FunAudioLLM/Fun-CosyVoice3-0.5B-2512 \
   --local-dir pretrained_models/CosyVoice3-0.5B
-```
-
-### 5. Download this model
 
-```bash
+# Thorsten fine-tuned weights
 hf download Thorsten-Voice/CosyVoice3 \
   --local-dir pretrained_models/CosyVoice3-0.5B \
   --include "llm.pt" "flow.pt" "spk2info.pt" "infer_thorsten.py"
 ```
 
-This overwrites only `llm.pt` and `flow.pt` in the base model directory.
-All other files (`hift.pt`, `campplus.onnx`, etc.) remain from the base model.
-
----
-
-## Inference
-
-### Quick start
+### 5. Generate audio
 
 ```bash
 python3 infer_thorsten.py \
@@ -96,27 +98,31 @@ python3 infer_thorsten.py \
   --output thorsten.wav
 ```
 
-Optional parameters:
+---
 
-```bash
-python3 infer_thorsten.py \
-  --text "Ein längerer Text hier." \
-  --model_dir pretrained_models/CosyVoice3-0.5B \
-  --output output.wav \
-  --speed 0.9
-```
+## Performance
+
+Benchmarked with these two test texts:
+
+**Short** (~8 words):
+> "Hallo, hier ist Thorsten. Schön, dass Du da bist."
+
+**Long** (~80 words):
+> "Für mich sind alle Menschen gleich, unabhängig von Geschlecht, sexueller Orientierung, Religion, Hautfarbe oder Geokoordinaten der Geburt. Ich glaube an eine globale Welt, wo jeder überall willkommen ist und freies Wissen und Bildung kostenfrei für jeden zur Verfügung steht. Ich habe meine Stimme der Allgemeinheit gespendet, in der Hoffnung darauf, dass sie in diesem Sinne genutzt wird."
+
+| Hardware | Short text | Long text |
+|----------|-----------|-----------|
+| MacBook Air M1 (CPU) | 47s | 4:30 min |
+| QNAP NAS Intel (CPU) | 50s | — |
+| RunPod RTX 4090 (GPU) | **2.9s** | **12.9s** |
 
 ---
 
 ## Python 3.12 patches
 
-When using Python 3.12, two additional source patches are required.
-
-**1. `cosyvoice/flow/flow.py`** — add these lines inside `CausalMaskedDiffWithDiT.forward()`,
-directly after `conds = conds.transpose(1, 2)` and before `loss, _ = self.decoder.compute_loss(...)`:
+**1. `cosyvoice/flow/flow.py`** — add after `conds = conds.transpose(1, 2)` in `CausalMaskedDiffWithDiT.forward()`:
 
 ```python
-# Alignment fix
 min_len = min(h.shape[1], feat.shape[1])
 h = h[:, :min_len, :]
 feat = feat[:, :min_len, :]
@@ -124,7 +130,7 @@ conds = conds[:, :, :min_len]
 mask = mask[:, :min_len]
 ```
 
-**2. `third_party/Matcha-TTS/matcha/utils/__init__.py`** — clear this file:
+**2. `third_party/Matcha-TTS/matcha/utils/__init__.py`:**
 
 ```bash
 echo "" > third_party/Matcha-TTS/matcha/utils/__init__.py
@@ -146,9 +152,8 @@ echo "" > third_party/Matcha-TTS/matcha/utils/__init__.py
 
 ## License
 
-The fine-tuned weights follow the [Apache 2.0](https://www.apache.org/licenses/LICENSE-2.0) license of the base model.
-
-The Thorsten-Voice dataset is licensed under [Creative Commons Zero (CC0)](https://creativecommons.org/publicdomain/zero/1.0/).
+Apache 2.0 — same as the base model.
+The Thorsten-Voice dataset is licensed under [CC0](https://creativecommons.org/publicdomain/zero/1.0/).
 
 ---
 
@@ -168,5 +173,7 @@ The Thorsten-Voice dataset is licensed under [Creative Commons Zero (CC0)](https
 ## Links
 
 - [Thorsten-Voice Website](https://www.thorsten-voice.de)
+- [Docker Container](https://hub.docker.com/r/thorstenvoice/cosyvoice-tts)
 - [CosyVoice GitHub](https://github.com/FunAudioLLM/CosyVoice)
-- [Base Model on HuggingFace](https://huggingface.co/FunAudioLLM/Fun-CosyVoice3-0.5B-2512)
+- [Base Model](https://huggingface.co/FunAudioLLM/Fun-CosyVoice3-0.5B-2512)
+- [Source on GitHub](https://github.com/thorstenMueller/Thorsten-Voice/tree/main/docker/cosyvoice)