From aed691ab4e0ea9fbdccbb8bd62feb306ae272b7c Mon Sep 17 00:00:00 2001
From: SpenserCai <spensercai@gmail.com>
Date: Fri, 4 Oct 2024 21:15:11 +0800
Subject: [PATCH] add InstructNode support

---
 README.md                         |  5 +++--
 __init__.py                       |  6 +++++-
 funaudio_utils/download_models.py |  9 +++++++-
 nodes/cosyvoice_nodes.py          | 36 +++++++++++++++++++++++++++++--
 4 files changed, 50 insertions(+), 6 deletions(-)

diff --git a/README.md b/README.md
index 416ac3a..daf52fc 100644
--- a/README.md
+++ b/README.md
@@ -3,7 +3,7 @@
  * @Date: 2024-10-04 11:28:28
  * @version: 
  * @LastEditors: SpenserCai
- * @LastEditTime: 2024-10-04 20:48:17
+ * @LastEditTime: 2024-10-04 21:14:03
  * @Description: file content
 -->
 # ComfyUI-FunAudioLLM
@@ -12,7 +12,8 @@ Comfyui custom node for [FunAudioLLM](https://funaudiollm.github.io/) include [C
 ## Features
 
 ### CosyVoice
-  - Support SFT,Zero-shot,Cross-lingual
+  - CosyVoice Version: 2024-10-04
+  - Support SFT,Zero-shot,Cross-lingual,Instruct
   - Support CosyVoice-300M-25Hz in zero-shot and cross-lingual
   - Support SFT's 25Hz(unoffical)
   - <details>
diff --git a/__init__.py b/__init__.py
index 3d29db6..3e22c39 100644
--- a/__init__.py
+++ b/__init__.py
@@ -3,7 +3,7 @@ Author: SpenserCai
 Date: 2024-10-04 12:14:22
 version: 
 LastEditors: SpenserCai
-LastEditTime: 2024-10-04 20:14:52
+LastEditTime: 2024-10-04 21:12:47
 Description: file content
 '''
 from .nodes.cosyvoice_nodes import *
@@ -22,6 +22,10 @@ NODE_CONFIG = {
         "class": CosyVoiceCrossLingualNode,
         "name": "CosyVoice 跨语言克隆"
     },
+    "CosyVoiceInstructNode": {
+        "class": CosyVoiceInstructNode,
+        "name": "CosyVoice 自然语言控制"
+    },
     "CosyVoiceSaveSpeakerModelNode": {
         "class": CosyVoiceSaveSpeakerModelNode,
         "name": "CosyVoice 保存说话人模型"
diff --git a/funaudio_utils/download_models.py b/funaudio_utils/download_models.py
index baeecbc..74c14d6 100644
--- a/funaudio_utils/download_models.py
+++ b/funaudio_utils/download_models.py
@@ -3,7 +3,7 @@ Author: SpenserCai
 Date: 2024-10-04 13:54:01
 version: 
 LastEditors: SpenserCai
-LastEditTime: 2024-10-04 18:42:14
+LastEditTime: 2024-10-04 21:10:55
 Description: file content
 '''
 import modelscope
@@ -34,5 +34,12 @@ def download_cosyvoice_300m_sft(is_25hz=False):
     snapshot_download(model_id=model_id, local_dir=model_dir)
     return model_name, model_dir
 
+def download_cosyvoice_300m_instruct():
+    model_name = "CosyVoice-300M-Instruct"
+    model_id = "iic/CosyVoice-300M-Instruct"
+    model_dir = os.path.join(base_cosyvoice_model_path, model_name)
+    snapshot_download(model_id=model_id, local_dir=model_dir)
+    return model_name, model_dir
+
 def get_speaker_default_path():
     return os.path.join(base_cosyvoice_model_path, "Speaker")
\ No newline at end of file
diff --git a/nodes/cosyvoice_nodes.py b/nodes/cosyvoice_nodes.py
index 1b25590..2c8c79d 100644
--- a/nodes/cosyvoice_nodes.py
+++ b/nodes/cosyvoice_nodes.py
@@ -3,7 +3,7 @@ Author: SpenserCai
 Date: 2024-10-04 12:13:28
 version: 
 LastEditors: SpenserCai
-LastEditTime: 2024-10-04 20:13:31
+LastEditTime: 2024-10-04 21:11:53
 Description: file content
 '''
 import os
@@ -12,7 +12,7 @@ import random
 import numpy as np
 import torch
 from funaudio_utils.pre import FunAudioLLMTool
-from funaudio_utils.download_models import download_cosyvoice_300m, get_speaker_default_path, download_cosyvoice_300m_sft
+from funaudio_utils.download_models import download_cosyvoice_300m, get_speaker_default_path, download_cosyvoice_300m_sft,download_cosyvoice_300m_instruct
 from funaudio_utils.cosyvoice_plus import CosyVoicePlus
 from cosyvoice.utils.common import set_all_random_seed
 
@@ -158,6 +158,38 @@ class CosyVoiceCrossLingualNode:
         output = cosyvoice.inference_cross_lingual(tts_text, prompt_speech_16k, False, speed)
         return return_audio(output,t0,None)
 
+class CosyVoiceInstructNode:
+    sft_spk_list = ['中文女', '中文男', '日语男', '粤语女', '英文女', '英文男', '韩语女']
+    @classmethod
+    def INPUT_TYPES(s):
+        return {
+            "required":{
+                "tts_text":("STRING",),
+                "speaker_name":(s.sft_spk_list,{
+                    "default":"中文女"
+                }),
+                "instruct_text":("STRING",),
+                "speed":("FLOAT",{
+                    "default": 1.0
+                }),
+                "seed":("INT",{
+                    "default": 42
+                }),
+            }
+        }
+    
+    CATEGORY = CATEGORY_NAME
+    RETURN_TYPES = ("AUDIO",)
+    FUNCTION="generate"
+
+    def generate(self, tts_text, speaker_name, instruct_text, speed, seed, use_25hz):
+        t0 = ttime()
+        _, model_dir = download_cosyvoice_300m_instruct()
+        cosyvoice = CosyVoicePlus(model_dir)
+        set_all_random_seed(seed)
+        output = cosyvoice.inference_instruct(tts_text, speaker_name, instruct_text, False, speed)
+        return return_audio(output,t0,None)
+
 class CosyVoiceLoadSpeakerModelNode:
     @classmethod
     def INPUT_TYPES(s):