diff --git a/python/llm/example/GPU/PyTorch-Models/Model/speech-t5/README.md b/python/llm/example/GPU/PyTorch-Models/Model/speech-t5/README.md
new file mode 100644
index 00000000..8c78e0d9
--- /dev/null
+++ b/python/llm/example/GPU/PyTorch-Models/Model/speech-t5/README.md
@@ -0,0 +1,121 @@
+# SpeechT5
+In this directory, you will find examples on how you could use BigDL-LLM `optimize_model` API to accelerate SpeechT5 models. For illustration purposes, we utilize the [microsoft/speecht5_tts](https://huggingface.co/microsoft/speecht5_tts) as reference SpeechT5 models.
+
+## Requirements
+To run these examples with BigDL-LLM, we have some recommended requirements for your machine, please refer to [here](../README.md#recommended-requirements) for more information.
+
+## Example: Synthesize speech with the given input text
+In the example [synthesize_speech.py](./synthesize_speech.py), we show a basic use case for SpeechT5 model to synthesize speech based on the given text, with BigDL-LLM INT4 optimizations.
+### 1. Install
+#### 1.1 Installation on Linux
+We suggest using conda to manage the Python environment. For more information about conda installation, please refer to [here](https://docs.conda.io/en/latest/miniconda.html#).
+
+After installing conda, create a Python environment for BigDL-LLM:
+```bash
+conda create -n llm python=3.9 # recommend to use Python 3.9
+conda activate llm
+
+# below command will install intel_extension_for_pytorch==2.1.10+xpu as default
+pip install --pre --upgrade bigdl-llm[xpu] -f https://developer.intel.com/ipex-whl-stable-xpu
+pip install datasets soundfile # additional package required for SpeechT5 to conduct generation
+```
+
+#### 1.2 Installation on Windows
+We suggest using conda to manage environment:
+```bash
+conda create -n llm python=3.9 libuv
+conda activate llm
+
+# below command will install intel_extension_for_pytorch==2.1.10+xpu as default
+pip install --pre --upgrade bigdl-llm[xpu] -f https://developer.intel.com/ipex-whl-stable-xpu
+pip install datasets soundfile # additional package required for SpeechT5 to conduct generation
+```
+
+### 2. Configures OneAPI environment variables
+#### 2.1 Configurations for Linux
+```bash
+source /opt/intel/oneapi/setvars.sh
+```
+
+#### 2.2 Configurations for Windows
+```cmd
+call "C:\Program Files (x86)\Intel\oneAPI\setvars.bat"
+```
+> Note: Please make sure you are using **CMD** (**Anaconda Prompt** if using conda) to run the command as PowerShell is not supported.
+### 3. Runtime Configurations
+For optimal performance, it is recommended to set several environment variables. Please check out the suggestions based on your device.
+#### 3.1 Configurations for Linux
+
+
+For Intel Arc™ A-Series Graphics and Intel Data Center GPU Flex Series
+
+```bash
+export USE_XETLA=OFF
+export SYCL_PI_LEVEL_ZERO_USE_IMMEDIATE_COMMANDLISTS=1
+```
+
+
+
+
+
+For Intel Data Center GPU Max Series
+
+```bash
+export LD_PRELOAD=${LD_PRELOAD}:${CONDA_PREFIX}/lib/libtcmalloc.so
+export SYCL_PI_LEVEL_ZERO_USE_IMMEDIATE_COMMANDLISTS=1
+export ENABLE_SDP_FUSION=1
+```
+> Note: Please note that `libtcmalloc.so` can be installed by `conda install -c conda-forge -y gperftools=2.10`.
+
+
+#### 3.2 Configurations for Windows
+
+
+For Intel iGPU
+
+```cmd
+set SYCL_CACHE_PERSISTENT=1
+set BIGDL_LLM_XMX_DISABLED=1
+```
+
+
+
+
+
+For Intel Arc™ A300-Series or Pro A60
+
+```cmd
+set SYCL_CACHE_PERSISTENT=1
+```
+
+
+
+
+
+For other Intel dGPU Series
+
+There is no need to set further environment variables.
+
+
+
+> Note: For the first time that each model runs on Intel iGPU/Intel Arc™ A300-Series or Pro A60, it may take several minutes to compile.
+### 4. Running examples
+
+```bash
+python ./synthesize_speech.py --text 'Artificial intelligence refers to the development of computer systems that can perform tasks that typically require human intelligence.'
+```
+
+In the example, several arguments can be passed to satisfy your requirements:
+
+- `--repo-id-or-model-path REPO_ID_OR_MODEL_PATH`: argument defining the huggingface repo id for the SpeechT5 model (e.g `microsoft/speecht5_tts`) to be downloaded, or the path to the huggingface checkpoint folder. It is default to be `'microsoft/speecht5_tts'`.
+- `--repo-id-or-vocoder-path REPO_ID_OR_VOCODER_PATH`: argument defining the huggingface repo id for the SpeechT5 vocoder (e.g `microsoft/speecht5_hifigan`, which generates audio from a spectrogram) to be downloaded, or the path to the huggingface checkpoint folder. It is default to be `'microsoft/speecht5_hifigan'`.
+- `--repo-id-or-data-path REPO_ID_OR_DATA_PATH`: argument defining the huggingface repo id for the audio dataset (e.g. `Matthijs/cmu-arctic-xvectors`, which decides voice characteristics) to be downloaded, or the path to the huggingface dataset folder. It is default to be `'Matthijs/cmu-arctic-xvectors'`.
+- `--text TEXT`: argument defining the text to synthesize speech. It is default to be `"Artificial intelligence refers to the development of computer systems that can perform tasks that typically require human intelligence."`.
+
+#### 4.1 Sample Output
+
+#### [microsoft/speecht5_tts](https://huggingface.co/microsoft/speecht5_tts)
+
+Text: Artificial intelligence refers to the development of computer systems that can perform tasks that typically require human intelligence.
+
+[Click here to hear sample output.](https://llm-assets.readthedocs.io/en/latest/_downloads/f0bebfbe8c350b71fe565a82192c079b/speech-t5-example-output.wav)
diff --git a/python/llm/example/GPU/PyTorch-Models/Model/speech-t5/synthesize_speech.py b/python/llm/example/GPU/PyTorch-Models/Model/speech-t5/synthesize_speech.py
new file mode 100644
index 00000000..342d6229
--- /dev/null
+++ b/python/llm/example/GPU/PyTorch-Models/Model/speech-t5/synthesize_speech.py
@@ -0,0 +1,101 @@
+#
+# Copyright 2016 The BigDL Authors.
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+# http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+#
+# Some parts of this file is adapted from
+# https://huggingface.co/microsoft/speecht5_tts
+#
+# MIT License
+#
+# Copyright (c) Microsoft Corporation.
+#
+# Permission is hereby granted, free of charge, to any person obtaining a copy
+# of this software and associated documentation files (the "Software"), to deal
+# in the Software without restriction, including without limitation the rights
+# to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
+# copies of the Software, and to permit persons to whom the Software is
+# furnished to do so, subject to the following conditions:
+#
+# The above copyright notice and this permission notice shall be included in all
+# copies or substantial portions of the Software.
+#
+# THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+# IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+# FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
+# AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+# LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
+# OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+# SOFTWARE
+
+
+import torch
+import time
+import argparse
+
+from bigdl.llm import optimize_model
+from transformers import SpeechT5Processor, SpeechT5HifiGan, SpeechT5ForTextToSpeech
+from datasets import load_dataset
+import soundfile as sf
+
+
+if __name__ == '__main__':
+ parser = argparse.ArgumentParser(description='Synthesize speech with the given input text using SpeechT5 model')
+ parser.add_argument('--repo-id-or-model-path', type=str, default='microsoft/speecht5_tts',
+ help='The huggingface repo id for the SpeechT5 model to be downloaded'
+ ', or the path to the huggingface checkpoint folder')
+ parser.add_argument('--repo-id-or-vocoder-path', type=str, default='microsoft/speecht5_hifigan',
+ help='The huggingface repo id for the vocoder model (which generates audio from a spectrogram) to be downloaded'
+ ', or the path to the huggingface checkpoint folder.')
+ parser.add_argument('--repo-id-or-data-path', type=str, default="Matthijs/cmu-arctic-xvectors",
+ help='The huggingface repo id for the audio dataset (which decides voice characteristics) to be downloaded'
+ ', or the path to the huggingface dataset folder')
+ parser.add_argument('--text', type=str, default="Artificial intelligence refers to the development of computer systems that can perform tasks that typically require human intelligence.",
+ help='Text to synthesize speech')
+
+ args = parser.parse_args()
+ model_path = args.repo_id_or_model_path
+ vocoder_path = args.repo_id_or_vocoder_path
+ dataset_path = args.repo_id_or_data_path
+ text = args.text
+
+ processor = SpeechT5Processor.from_pretrained(model_path)
+ model = SpeechT5ForTextToSpeech.from_pretrained(model_path)
+ vocoder = SpeechT5HifiGan.from_pretrained(vocoder_path)
+
+ # With only one line to enable BigDL-LLM optimization on model
+ # Skip optimizing these two modules to get higher audio quality
+ # When running LLMs on Intel iGPUs for Windows users, we recommend setting `cpu_embedding=True` in the optimize_model function.
+ # This will allow the memory-intensive embedding layer to utilize the CPU instead of iGPU.
+ model = optimize_model(model, modules_to_not_convert=["speech_decoder_postnet.feat_out",
+ "speech_decoder_postnet.prob_out"])
+ model = model.to('xpu')
+ vocoder = vocoder.to('xpu')
+
+ inputs = processor(text=text, return_tensors="pt").to('xpu')
+
+ # load xvector containing speaker's voice characteristics from a dataset
+ embeddings_dataset = load_dataset(dataset_path, split="validation")
+ speaker_embeddings = torch.tensor(embeddings_dataset[7306]["xvector"]).unsqueeze(0).to('xpu')
+
+ with torch.inference_mode():
+ # ipex model needs a warmup, then inference time can be accurate
+ speech = model.generate_speech(inputs["input_ids"], speaker_embeddings, vocoder=vocoder)
+
+ st = time.time()
+ speech = model.generate_speech(inputs["input_ids"], speaker_embeddings, vocoder=vocoder)
+ torch.xpu.synchronize()
+ end = time.time()
+ print(f"Inference time: {end-st} s")
+
+ sf.write("bigdl_llm_speech_t5_out.wav", speech.to('cpu').numpy(), samplerate=16000)