VieNeu TTS (Voice Clone):
VieNeuTTSClone generates Vietnamese speech in a voice guided by a reference recording. This lets you create new spoken lines that draw on the reference voice, which can be useful for narration, dialogue, and other audio work where a consistent voice matters. The node uses the VieNeu speech engine to generate the audio, saves a WAV file in ComfyUI’s output directory, and makes the audio available to other nodes in your workflow. The engine is initialized when first needed, although that first use may take longer if its models need to be loaded or downloaded.
VieNeu TTS (Voice Clone) Input Parameters:
text
The words you want the generated voice to speak. This is a required, multiline STRING field. Its default is Đây là giọng nói đã được clone từ mẫu. Changing this text changes the spoken content; it does not change which recording guides the voice. No minimum or maximum length is specified in the node interface.
ref_audio_path
The path to the reference audio recording used to guide the generated voice. This is a required STRING field with the default examples/audio_ref/example.wav. The node checks whether a file exists at this path before starting inference, so replace the example path with a path to a file that is accessible from the environment running ComfyUI. The interface does not specify an allowed-format list or a minimum or maximum path length; the VieNeu engine must also be able to process the file.
ref_text
The words spoken in the reference recording. This is a required, multiline STRING field with the default Văn bản tương ứng với audio mẫu. The node sends this text alongside the reference audio to the speech engine, so it should correspond to what is actually said in the recording. It is separate from text, which contains the new words you want generated. No minimum or maximum length is specified in the node interface.
filename_prefix
The name used to begin the saved WAV file’s name. This is a required STRING field with the default vieneu_cloned. For example, that default produces vieneu_cloned.wav in ComfyUI’s output directory when the name is available. If a file with that name already exists, the node tries a numbered name such as vieneu_cloned_1.wav. No minimum or maximum length or character restrictions are specified in the interface, but the resulting filename must be valid and writable on your system.
VieNeu TTS (Voice Clone) Output Parameters:
audio
An AUDIO output containing the generated speech for use by downstream ComfyUI audio nodes. The node returns a waveform as a float32 tensor arranged as batch, channels, and samples, together with a declared sample rate of 24,000 Hz. This is the output to connect when you want to continue processing or using the sound within a workflow.
filepath
A STRING containing the path of the WAV file saved in ComfyUI’s output directory. Use it when you need to locate the generated file or pass its path to another compatible node. The exact path can change between runs because the node adds a number to the filename when a file with the intended name already exists.
VieNeu TTS (Voice Clone) Usage Tips:
- Set
ref_audio_pathto a recording that exists in the environment running ComfyUI. The supplied default is an example path, not a guarantee that the file is present. - Enter the words actually spoken in that recording in
ref_text, then put the new line you want to hear intext. Keeping these two roles distinct makes the reference information clear. - Use a descriptive
filename_prefixwhen generating several lines or testing different references. Checkfilepathfor the final saved name rather than assuming the unnumbered filename was used. - Allow extra time for the first execution. The shared VieNeu engine is initialized on demand and may need to load or download models; later calls reuse the initialized engine.
- Check the generated
audioin your workflow and the WAV file atfilepathbefore relying on either for a final project. The provided source shows how the node prepares and returns audio, but does not establish how a particular reference will sound.
VieNeu TTS (Voice Clone) Common Errors and Solutions:
Reference audio not found: <path>
- Explanation: The node could not find a file at the value entered in
ref_audio_path. This check happens before the speech engine runs. - Solution: Confirm the spelling and location of the file from the environment running ComfyUI. Use a path that points to the actual recording rather than leaving the example path unchanged unless that example file exists.
Error cloning voice: <details>
- Explanation: This is the node’s general error log. It appears when a failure occurs during engine initialization, speech generation, saving, or audio conversion; the text after the colon identifies the underlying problem.
- Solution: Read the full error details in the ComfyUI console. Check the reference path and file, confirm that
ref_textdescribes its spoken content, and verify that the output directory is writable. Use the specific underlying message to narrow down the cause.
Permission denied or another file-write error
- Explanation: The node creates ComfyUI’s output directory if needed and asks VieNeu to save a WAV file there. Saving can fail if the process cannot write to that location.
- Solution: Check that ComfyUI has write permission to its output directory and that the destination is available. Also use a simple
filename_prefixwithout path separators or characters prohibited by your operating system.
Audio tensor MUST be 3D [batch, channels, samples] for ComfyUI, got <dimensions>D with shape <shape>
- Explanation: The node requires a three-dimensional audio waveform for its
AUDIOoutput. It attempts to reshape other results first, then raises this error if the final dimensionality check still fails. - Solution: Review the preceding console output for the reported audio type and shape. Check that the reference file can be processed by VieNeu and that the installed VieNeu version is compatible with this custom node; this message indicates an unexpected generation or conversion result rather than a filename problem.
