ComfyUI Node: FL Qwen3 TTS Custom Voice

FL Qwen3 TTS Custom Voice:

The FL_Qwen3TTS_CustomVoice node is designed to generate speech using a custom voice model, specifically tailored for the Qwen3-TTS system. This node allows you to create speech outputs with a high degree of customization, leveraging a model that supports multiple predefined speakers. The primary goal of this node is to provide a flexible and powerful tool for generating speech that can be tailored to specific needs, such as creating unique voiceovers or enhancing multimedia projects with distinct vocal characteristics. By utilizing this node, you can take advantage of advanced text-to-speech capabilities, ensuring that the generated audio aligns with your creative vision. The node is particularly beneficial for projects requiring specific voice attributes or when working with predefined speaker profiles.

FL Qwen3 TTS Custom Voice Input Parameters:

model

This parameter requires a model object that contains the custom voice model. It is crucial for the node's operation as it defines the voice characteristics and capabilities. The model must be compatible with the Qwen3-TTS-12Hz-1.7B-CustomVoice type, which includes nine predefined speakers. If an incompatible model is provided, the node will not function correctly.

seed

The seed parameter is used to initialize the random number generator, ensuring that the speech generation process can be replicated. This is particularly useful for achieving consistent results across different runs. There is no specific minimum or maximum value, but using the same seed will produce the same output for identical inputs.

speaker

This parameter specifies which predefined speaker profile to use for generating speech. The available options include Ryan, Vivian, Serena, among others. Selecting a speaker determines the vocal characteristics of the output, allowing for a wide range of expressive possibilities.

top_k

The top_k parameter controls the number of highest probability vocabulary tokens to keep for sampling. A higher value allows for more diversity in the generated speech, while a lower value makes the output more deterministic. Adjusting this parameter can significantly impact the creativity and variability of the speech output.

top_p

This parameter, also known as nucleus sampling, limits the sampling pool to a subset of tokens with a cumulative probability above a certain threshold. It helps balance between diversity and coherence in the generated speech. A value close to 1.0 allows for more diverse outputs, while a lower value results in more focused speech.

temperature

The temperature parameter influences the randomness of the speech generation. A higher temperature results in more random outputs, while a lower temperature makes the output more deterministic. This parameter is essential for controlling the creativity of the generated speech.

repetition_penalty

This parameter penalizes the model for repeating the same phrases or words, encouraging more varied and interesting speech outputs. It is particularly useful for avoiding monotonous or repetitive speech patterns.

max_new_tokens

This parameter defines the maximum number of new tokens to generate. It directly impacts the length of the generated speech, allowing you to control how verbose or concise the output should be.

FL Qwen3 TTS Custom Voice Output Parameters:

generated_speech

The primary output of the node is the generated_speech, which is the audio representation of the input text processed through the custom voice model. This output is crucial for applications requiring high-quality, customized speech synthesis, and it can be used in various multimedia projects to enhance user engagement with unique vocal outputs.

FL Qwen3 TTS Custom Voice Usage Tips:

• Ensure that the model provided is compatible with the Qwen3-TTS-12Hz-1.7B-CustomVoice type to avoid compatibility issues.
• Experiment with different top_k and top_p values to find the right balance between diversity and coherence for your specific project needs.
• Use the seed parameter to replicate results when consistency is required across multiple runs.

FL Qwen3 TTS Custom Voice Common Errors and Solutions:

No model provided. Please connect a Model Loader node.

• Explanation: This error occurs when the node is not connected to a valid model object.
• Solution: Ensure that a compatible model is loaded and connected to the node before execution.

Wrong model type for Custom Voice node!

• Explanation: The model provided is not of the Qwen3-TTS-12Hz-1.7B-CustomVoice type.
• Solution: Change the model to Qwen3-TTS-12Hz-1.7B-CustomVoice or use the Voice Clone node with your current base model to clone a voice from reference audio.