[Bug] RVC index_ratio does nothing.

[maraan666]

as title, everything else works fab. RVC training works perfectly and is really fast.

RVC works great and inference is really fast.

RVC index_ratio seems to do nothing.

And it would be brilliant to be able to combine RVC models on the fly and so create hybrid voices.

Not complaining, just a heads up. Your work is entirely admirable. Indeed, stunning may be more accurate.

btw, chatterbox voice changer didn't work at all (not that I particularly care - but it might help other people).got this error...
❌ Voice conversion failed: Node processing failed: AttributeError: module 'onnx.onnx_cpp2py_export.shape_inference' has no attribute 'GraphInferencer'
Traceback:
Traceback (most recent call last):
File "C:\SD-Comfy\ComfyUI_windows_portable\ComfyUI\custom_nodes\tts_audio_suite\nodes\base\base_node.py", line 375, in process_with_error_handling
return process_func(*args, **kwargs)
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "C:\SD-Comfy\ComfyUI_windows_portable\ComfyUI\custom_nodes\tts_audio_suite\nodes\chatterbox\chatterbox_vc_node.py", line 155, in _process
self.load_vc_model(device, language=language)
File "C:\SD-Comfy\ComfyUI_windows_portable\ComfyUI\custom_nodes\tts_audio_suite\nodes\base\base_node.py", line 526, in load_vc_model
self.vc_model = self.model_manager.load_vc_model(device, force_reload, language)
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "C:\SD-Comfy\ComfyUI_windows_portable\ComfyUI\custom_nodes\tts_audio_suite\utils\models\manager.py", line 541, in load_vc_model
if not _get_chatterbox_vc_available():
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "C:\SD-Comfy\ComfyUI_windows_portable\ComfyUI\custom_nodes\tts_audio_suite\utils\models\manager.py", line 77, in get_chatterbox_vc_available
return lazy_import_chatterbox_vc()[0]
^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "C:\SD-Comfy\ComfyUI_windows_portable\ComfyUI\custom_nodes\tts_audio_suite\utils\models\manager.py", line 42, in lazy_import_chatterbox_vc
success, cls, source = import_manager.import_chatterbox_vc()
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "C:\SD-Comfy\ComfyUI_windows_portable\ComfyUI\custom_nodes\tts_audio_suite\utils\system\import_manager.py", line 125, in import_chatterbox_vc
from engines.chatterbox.vc import ChatterboxVC
File "C:\SD-Comfy\ComfyUI_windows_portable\ComfyUI\custom_nodes\tts_audio_suite\engines\chatterbox\vc.py", line 21, in
from .models.s3tokenizer import S3_SR
File "C:\SD-Comfy\ComfyUI_windows_portable\ComfyUI\custom_nodes\tts_audio_suite\engines\chatterbox\models\s3tokenizer_init.py", line 1, in
from .s3tokenizer import (
File "C:\SD-Comfy\ComfyUI_windows_portable\ComfyUI\custom_nodes\tts_audio_suite\engines\chatterbox\models\s3tokenizer\s3tokenizer.py", line 19, in
from s3tokenizer.utils import padding
File "C:\SD-Comfy\ComfyUI_windows_portable\python_embeded\Lib\site-packages\s3tokenizer_init.py", line 27, in
from s3tokenizer.model_v2 import S3TokenizerV2
File "C:\SD-Comfy\ComfyUI_windows_portable\python_embeded\Lib\site-packages\s3tokenizer\model_v2.py", line 21, in
from s3tokenizer.model import Conv1d, LayerNorm, Linear, MultiHeadAttention
File "C:\SD-Comfy\ComfyUI_windows_portable\python_embeded\Lib\site-packages\s3tokenizer\model.py", line 28, in
from .utils import make_non_pad_mask, mask_to_bias, onnx2torch, merge_tokenized_segments
File "C:\SD-Comfy\ComfyUI_windows_portable\python_embeded\Lib\site-packages\s3tokenizer\utils.py", line 25, in
import onnx
File "C:\SD-Comfy\ComfyUI_windows_portable\python_embeded\Lib\site-packages\onnx_init.py", line 142, in
from onnx import (
File "C:\SD-Comfy\ComfyUI_windows_portable\python_embeded\Lib\site-packages\onnx\compose.py", line 8, in
from onnx import (
File "C:\SD-Comfy\ComfyUI_windows_portable\python_embeded\Lib\site-packages\onnx\utils.py", line 13, in
import onnx.shape_inference
File "C:\SD-Comfy\ComfyUI_windows_portable\python_embeded\Lib\site-packages\onnx\shape_inference.py", line 28, in
GraphInferencer = C.GraphInferencer
^^^^^^^^^^^^^^^^^
AttributeError: module 'onnx.onnx_cpp2py_export.shape_inference' has no attribute 'GraphInferencer'

[diodiogod]

Hi, the ChatterBox VC error is real, but it is unrelated to the RVC index_ratio question.

The traceback indicates that your onnx Python package and compiled extension are mismatched, likely due to a partial upgrade or another custom node changing shared dependencies. Please run:

python_embeded\python.exe -m pip show onnx s3tokenizer onnxruntime
python_embeded\python.exe -m pip check

Please post the output and open a separate issue for the index_ratio behavior with a workflow and RVC model details.