add openvoice backend (#2334)

Wip openvoice

.github/workflows/test-extra.yml

@@ -123,6 +123,28 @@ jobs:
         run: |
            make --jobs=5 --output-sync=target -C backend/python/parler-tts
            make --jobs=5 --output-sync=target -C backend/python/parler-tts test
+  
+  tests-openvoice:
+    runs-on: ubuntu-latest
+    steps:
+      - name: Clone
+        uses: actions/checkout@v4
+        with: 
+          submodules: true
+      - name: Dependencies
+        run: |
+          sudo apt-get update
+          sudo apt-get install build-essential ffmpeg
+          # Install UV
+          curl -LsSf https://astral.sh/uv/install.sh | sh
+          sudo apt-get install -y ca-certificates cmake curl patch python3-pip
+          sudo apt-get install -y libopencv-dev
+          pip install --user grpcio-tools==1.63.0
+
+      - name: Test openvoice
+        run: |
+           make --jobs=5 --output-sync=target -C backend/python/openvoice
+           make --jobs=5 --output-sync=target -C backend/python/openvoice test
 
   tests-transformers-musicgen:
     runs-on: ubuntu-latest

Dockerfile

@@ -12,7 +12,7 @@ ARG TARGETARCH
 ARG TARGETVARIANT
 
 ENV DEBIAN_FRONTEND=noninteractive
-ENV EXTERNAL_GRPC_BACKENDS="coqui:/build/backend/python/coqui/run.sh,huggingface-embeddings:/build/backend/python/sentencetransformers/run.sh,petals:/build/backend/python/petals/run.sh,transformers:/build/backend/python/transformers/run.sh,sentencetransformers:/build/backend/python/sentencetransformers/run.sh,rerankers:/build/backend/python/rerankers/run.sh,autogptq:/build/backend/python/autogptq/run.sh,bark:/build/backend/python/bark/run.sh,diffusers:/build/backend/python/diffusers/run.sh,exllama:/build/backend/python/exllama/run.sh,vall-e-x:/build/backend/python/vall-e-x/run.sh,vllm:/build/backend/python/vllm/run.sh,mamba:/build/backend/python/mamba/run.sh,exllama2:/build/backend/python/exllama2/run.sh,transformers-musicgen:/build/backend/python/transformers-musicgen/run.sh,parler-tts:/build/backend/python/parler-tts/run.sh"
+ENV EXTERNAL_GRPC_BACKENDS="coqui:/build/backend/python/coqui/run.sh,huggingface-embeddings:/build/backend/python/sentencetransformers/run.sh,petals:/build/backend/python/petals/run.sh,transformers:/build/backend/python/transformers/run.sh,sentencetransformers:/build/backend/python/sentencetransformers/run.sh,rerankers:/build/backend/python/rerankers/run.sh,autogptq:/build/backend/python/autogptq/run.sh,bark:/build/backend/python/bark/run.sh,diffusers:/build/backend/python/diffusers/run.sh,exllama:/build/backend/python/exllama/run.sh,openvoice:/build/backend/python/openvoice/run.sh,vall-e-x:/build/backend/python/vall-e-x/run.sh,vllm:/build/backend/python/vllm/run.sh,mamba:/build/backend/python/mamba/run.sh,exllama2:/build/backend/python/exllama2/run.sh,transformers-musicgen:/build/backend/python/transformers-musicgen/run.sh,parler-tts:/build/backend/python/parler-tts/run.sh"
 
 ARG GO_TAGS="stablediffusion tinydream tts"
 
@@ -306,6 +306,9 @@ RUN if [[ ( "${EXTRA_BACKENDS}" =~ "coqui" || -z "${EXTRA_BACKENDS}" ) && "$IMAG
 RUN if [[ ( "${EXTRA_BACKENDS}" =~ "vall-e-x" || -z "${EXTRA_BACKENDS}" ) && "$IMAGE_TYPE" == "extras" ]]; then \
         make -C backend/python/vall-e-x \
     ; fi && \
+    if [[ ( "${EXTRA_BACKENDS}" =~ "openvoice" || -z "${EXTRA_BACKENDS}" ) && "$IMAGE_TYPE" == "extras" ]]; then \
+        make -C backend/python/openvoice \
+    ; fi && \
     if [[ ( "${EXTRA_BACKENDS}" =~ "petals" || -z "${EXTRA_BACKENDS}" ) && "$IMAGE_TYPE" == "extras" ]]; then \
         make -C backend/python/petals \
     ; fi && \

Makefile

@@ -454,10 +454,10 @@ protogen-go-clean:
 	$(RM) bin/*
 
 .PHONY: protogen-python
-protogen-python: autogptq-protogen bark-protogen coqui-protogen diffusers-protogen exllama-protogen exllama2-protogen mamba-protogen petals-protogen rerankers-protogen sentencetransformers-protogen transformers-protogen parler-tts-protogen transformers-musicgen-protogen vall-e-x-protogen vllm-protogen
+protogen-python: autogptq-protogen bark-protogen coqui-protogen diffusers-protogen exllama-protogen exllama2-protogen mamba-protogen petals-protogen rerankers-protogen sentencetransformers-protogen transformers-protogen parler-tts-protogen transformers-musicgen-protogen vall-e-x-protogen vllm-protogen openvoice-protogen
 
 .PHONY: protogen-python-clean
-protogen-python-clean: autogptq-protogen-clean bark-protogen-clean coqui-protogen-clean diffusers-protogen-clean exllama-protogen-clean exllama2-protogen-clean mamba-protogen-clean petals-protogen-clean sentencetransformers-protogen-clean rerankers-protogen-clean transformers-protogen-clean transformers-musicgen-protogen-clean parler-tts-protogen-clean vall-e-x-protogen-clean vllm-protogen-clean
+protogen-python-clean: autogptq-protogen-clean bark-protogen-clean coqui-protogen-clean diffusers-protogen-clean exllama-protogen-clean exllama2-protogen-clean mamba-protogen-clean petals-protogen-clean sentencetransformers-protogen-clean rerankers-protogen-clean transformers-protogen-clean transformers-musicgen-protogen-clean parler-tts-protogen-clean vall-e-x-protogen-clean vllm-protogen-clean openvoice-protogen-clean
 
 .PHONY: autogptq-protogen
 autogptq-protogen:
@@ -571,6 +571,14 @@ vall-e-x-protogen:
 vall-e-x-protogen-clean:
 	$(MAKE) -C backend/python/vall-e-x protogen-clean
 
+.PHONY: openvoice-protogen
+openvoice-protogen:
+	$(MAKE) -C backend/python/openvoice protogen
+
+.PHONY: openvoice-protogen-clean
+openvoice-protogen-clean:
+	$(MAKE) -C backend/python/openvoice protogen-clean
+
 .PHONY: vllm-protogen
 vllm-protogen:
 	$(MAKE) -C backend/python/vllm protogen
@@ -594,6 +602,7 @@ prepare-extra-conda-environments: protogen-python
 	$(MAKE) -C backend/python/transformers-musicgen
 	$(MAKE) -C backend/python/parler-tts
 	$(MAKE) -C backend/python/vall-e-x
+	$(MAKE) -C backend/python/openvoice
 	$(MAKE) -C backend/python/exllama
 	$(MAKE) -C backend/python/petals
 	$(MAKE) -C backend/python/exllama2

backend/python/openvoice/Makefile

@@ -0,0 +1,25 @@
+.DEFAULT_GOAL := install
+
+.PHONY: install
+install: protogen
+	bash install.sh
+
+.PHONY: protogen
+protogen: backend_pb2_grpc.py backend_pb2.py
+
+.PHONY: protogen-clean
+protogen-clean:
+	$(RM) backend_pb2_grpc.py backend_pb2.py
+
+backend_pb2_grpc.py backend_pb2.py:
+	python3 -m grpc_tools.protoc -I../.. --python_out=. --grpc_python_out=. backend.proto
+
+.PHONY: clean
+clean: protogen-clean
+	rm -rf venv __pycache__
+
+.PHONY: test
+test: protogen
+	@echo "Testing openvoice..."
+	bash test.sh
+	@echo "openvoice tested."

backend/python/openvoice/backend.py

@@ -0,0 +1,158 @@
+#!/usr/bin/env python3
+"""
+Extra gRPC server for OpenVoice models.
+"""
+from concurrent import futures
+
+import argparse
+import signal
+import sys
+import os
+import torch
+from openvoice import se_extractor
+from openvoice.api import ToneColorConverter
+from melo.api import TTS
+
+import time
+import backend_pb2
+import backend_pb2_grpc
+
+import grpc
+
+
+_ONE_DAY_IN_SECONDS = 60 * 60 * 24
+
+# If MAX_WORKERS are specified in the environment use it, otherwise default to 1
+MAX_WORKERS = int(os.environ.get('PYTHON_GRPC_MAX_WORKERS', '1'))
+
+# Implement the BackendServicer class with the service methods
+class BackendServicer(backend_pb2_grpc.BackendServicer):
+    """
+    A gRPC servicer for the backend service.
+
+    This class implements the gRPC methods for the backend service, including Health, LoadModel, and Embedding.
+    """
+    def Health(self, request, context):
+        """
+        A gRPC method that returns the health status of the backend service.
+
+        Args:
+            request: A HealthRequest object that contains the request parameters.
+            context: A grpc.ServicerContext object that provides information about the RPC.
+
+        Returns:
+            A Reply object that contains the health status of the backend service.
+        """
+        return backend_pb2.Reply(message=bytes("OK", 'utf-8'))
+
+    def LoadModel(self, request, context):
+        """
+        A gRPC method that loads a model into memory.
+
+        Args:
+            request: A LoadModelRequest object that contains the request parameters.
+            context: A grpc.ServicerContext object that provides information about the RPC.
+
+        Returns:
+            A Result object that contains the result of the LoadModel operation.
+        """
+        model_name = request.Model
+        try:
+
+            self.clonedVoice = False
+            # Assume directory from request.ModelFile.
+            # Only if request.LoraAdapter it's not an absolute path
+            if request.AudioPath and request.ModelFile != "" and not os.path.isabs(request.AudioPath):
+                # get base path of modelFile
+                modelFileBase = os.path.dirname(request.ModelFile)
+                request.AudioPath = os.path.join(modelFileBase, request.AudioPath)
+            if request.AudioPath != "":
+                self.clonedVoice = True
+
+            self.modelpath = request.ModelFile
+            self.speaker = request.Type
+            self.ClonedVoicePath = request.AudioPath
+            
+            ckpt_converter = request.Model+'/converter'
+            device = "cuda:0" if torch.cuda.is_available() else "cpu"
+            self.device = device
+            self.tone_color_converter = None
+            if self.clonedVoice:
+                self.tone_color_converter = ToneColorConverter(f'{ckpt_converter}/config.json', device=device)
+                self.tone_color_converter.load_ckpt(f'{ckpt_converter}/checkpoint.pth')
+       
+        except Exception as err:
+            return backend_pb2.Result(success=False, message=f"Unexpected {err=}, {type(err)=}")
+
+        return backend_pb2.Result(message="Model loaded successfully", success=True)
+
+    def TTS(self, request, context):
+        model_name = request.model
+        if model_name == "":
+            return backend_pb2.Result(success=False, message="request.model is required")
+        try:
+            # Speed is adjustable
+            speed = 1.0
+            voice = "EN"
+            if request.voice:
+                voice = request.voice
+            model = TTS(language=voice, device=self.device)
+            speaker_ids = model.hps.data.spk2id
+            speaker_key = self.speaker
+            modelpath = self.modelpath
+            for s in speaker_ids.keys():
+                print(f"Speaker: {s} - ID: {speaker_ids[s]}")
+            speaker_id = speaker_ids[speaker_key]
+            speaker_key = speaker_key.lower().replace('_', '-')
+            source_se = torch.load(f'{modelpath}/base_speakers/ses/{speaker_key}.pth', map_location=self.device)
+            model.tts_to_file(request.text, speaker_id, request.dst, speed=speed)
+            if self.clonedVoice:
+                reference_speaker = self.ClonedVoicePath
+                target_se, audio_name = se_extractor.get_se(reference_speaker, self.tone_color_converter, vad=False)
+                # Run the tone color converter
+                encode_message = "@MyShell"
+                self.tone_color_converter.convert(
+                    audio_src_path=request.dst, 
+                    src_se=source_se, 
+                    tgt_se=target_se, 
+                    output_path=request.dst,
+                    message=encode_message)
+           
+            print("[OpenVoice] TTS generated!", file=sys.stderr)
+            print("[OpenVoice] TTS saved to", request.dst, file=sys.stderr)
+            print(request, file=sys.stderr)
+        except Exception as err:
+            return backend_pb2.Result(success=False, message=f"Unexpected {err=}, {type(err)=}")
+        return backend_pb2.Result(success=True)
+
+def serve(address):
+    server = grpc.server(futures.ThreadPoolExecutor(max_workers=MAX_WORKERS))
+    backend_pb2_grpc.add_BackendServicer_to_server(BackendServicer(), server)
+    server.add_insecure_port(address)
+    server.start()
+    print("[OpenVoice] Server started. Listening on: " + address, file=sys.stderr)
+
+    # Define the signal handler function
+    def signal_handler(sig, frame):
+        print("[OpenVoice] Received termination signal. Shutting down...")
+        server.stop(0)
+        sys.exit(0)
+
+    # Set the signal handlers for SIGINT and SIGTERM
+    signal.signal(signal.SIGINT, signal_handler)
+    signal.signal(signal.SIGTERM, signal_handler)
+
+    try:
+        while True:
+            time.sleep(_ONE_DAY_IN_SECONDS)
+    except KeyboardInterrupt:
+        server.stop(0)
+
+if __name__ == "__main__":
+    parser = argparse.ArgumentParser(description="Run the gRPC server.")
+    parser.add_argument(
+        "--addr", default="localhost:50051", help="The address to bind the server to."
+    )
+    args = parser.parse_args()
+    print(f"[OpenVoice] startup: {args}", file=sys.stderr)
+    serve(args.addr)

backend/python/openvoice/install.sh

@@ -0,0 +1,16 @@
+#!/bin/bash
+set -e
+
+source $(dirname $0)/../common/libbackend.sh
+
+# This is here because the Intel pip index is broken and returns 200 status codes for every package name, it just doesn't return any package links.
+# This makes uv think that the package exists in the Intel pip index, and by default it stops looking at other pip indexes once it finds a match.
+# We need uv to continue falling through to the pypi default index to find optimum[openvino] in the pypi index
+# the --upgrade actually allows us to *downgrade* torch to the version provided in the Intel pip index
+if [ "x${BUILD_PROFILE}" == "xintel" ]; then
+    EXTRA_PIP_INSTALL_FLAGS+=" --upgrade --index-strategy=unsafe-first-match"
+fi
+
+installRequirements
+
+python -m unidic download

backend/python/openvoice/requirements-intel.txt

@@ -0,0 +1,23 @@
+--extra-index-url https://pytorch-extension.intel.com/release-whl/stable/xpu/us/
+intel-extension-for-pytorch
+torch
+optimum[openvino]
+grpcio==1.63.0
+protobuf
+librosa==0.9.1
+faster-whisper==0.9.0
+pydub==0.25.1
+wavmark==0.0.3
+numpy==1.22.0
+eng_to_ipa==0.0.2
+inflect==7.0.0
+unidecode==1.3.7
+whisper-timestamped==1.14.2
+openai
+python-dotenv
+pypinyin==0.50.0
+cn2an==0.5.22
+jieba==0.42.1
+gradio==3.48.0
+langid==1.1.6
+git+https://github.com/myshell-ai/MeloTTS.git

backend/python/openvoice/requirements.txt

@@ -0,0 +1,20 @@
+grpcio==1.63.0
+protobuf
+librosa==0.9.1
+faster-whisper==0.9.0
+pydub==0.25.1
+wavmark==0.0.3
+numpy==1.22.0
+eng_to_ipa==0.0.2
+inflect==7.0.0
+unidecode==1.3.7
+whisper-timestamped==1.14.2
+openai
+python-dotenv
+pypinyin==0.50.0
+cn2an==0.5.22
+jieba==0.42.1
+gradio==3.48.0
+langid==1.1.6
+git+https://github.com/myshell-ai/MeloTTS.git
+git+https://github.com/myshell-ai/OpenVoice.git

backend/python/openvoice/run.sh

@@ -0,0 +1,4 @@
+#!/bin/bash
+source $(dirname $0)/../common/libbackend.sh
+
+startBackend $@

backend/python/openvoice/test.py

@@ -0,0 +1,82 @@
+"""
+A test script to test the gRPC service
+"""
+import unittest
+import subprocess
+import time
+import backend_pb2
+import backend_pb2_grpc
+
+import grpc
+
+
+class TestBackendServicer(unittest.TestCase):
+    """
+    TestBackendServicer is the class that tests the gRPC service
+    """
+    def setUp(self):
+        """
+        This method sets up the gRPC service by starting the server
+        """
+        self.service = subprocess.Popen(["python3", "backend.py", "--addr", "localhost:50051"])
+        time.sleep(10)
+
+    def tearDown(self) -> None:
+        """
+        This method tears down the gRPC service by terminating the server
+        """
+        self.service.terminate()
+        self.service.wait()
+
+    def test_server_startup(self):
+        """
+        This method tests if the server starts up successfully
+        """
+        try:
+            self.setUp()
+            with grpc.insecure_channel("localhost:50051") as channel:
+                stub = backend_pb2_grpc.BackendStub(channel)
+                response = stub.Health(backend_pb2.HealthMessage())
+                self.assertEqual(response.message, b'OK')
+        except Exception as err:
+            print(err)
+            self.fail("Server failed to start")
+        finally:
+            self.tearDown()
+
+    def test_load_model(self):
+        """
+        This method tests if the model is loaded successfully
+        """
+        try:
+            self.setUp()
+            with grpc.insecure_channel("localhost:50051") as channel:
+                stub = backend_pb2_grpc.BackendStub(channel)
+                response = stub.LoadModel(backend_pb2.ModelOptions(Model="checkpoints_v2", 
+                                                                    Type="en-us"))
+                self.assertTrue(response.success)
+                self.assertEqual(response.message, "Model loaded successfully")
+        except Exception as err:
+            print(err)
+            self.fail("LoadModel service failed")
+        finally:
+            self.tearDown()
+
+    def test_tts(self):
+        """
+        This method tests if the embeddings are generated successfully
+        """
+        try:
+            self.setUp()
+            with grpc.insecure_channel("localhost:50051") as channel:
+                stub = backend_pb2_grpc.BackendStub(channel)
+                response = stub.LoadModel(backend_pb2.ModelOptions(Model="dingzhen"))
+                self.assertTrue(response.success)
+                tts_request = backend_pb2.TTSRequest(text="80s TV news production music hit for tonight's biggest story", voice="EN")
+                tts_response = stub.TTS(tts_request)
+                self.assertIsNotNone(tts_response)
+        except Exception as err:
+            print(err)
+            self.fail("TTS service failed")
+        finally:
+            self.tearDown()

backend/python/openvoice/test.sh

@@ -0,0 +1,12 @@
+#!/bin/bash
+set -e
+
+source $(dirname $0)/../common/libbackend.sh
+
+# Download checkpoints if not present
+if [ ! -d "checkpoints_v2" ]; then
+    wget https://myshell-public-repo-hosting.s3.amazonaws.com/openvoice/checkpoints_v2_0417.zip -O checkpoints_v2.zip
+    unzip checkpoints_v2.zip
+fi
+
+runUnittests