ComfyUI Node: 🌈 IndexTTS-2 Text Emotion

🌈 IndexTTS-2 Text Emotion:

The QwenEmotionNode is a specialized component within the IndexTTS-2 framework designed to perform text-based emotion analysis. Its primary function is to extract emotion vectors from text, allowing for dynamic emotion control in text-to-speech applications. By leveraging the QwenEmotion model, this node can analyze text to determine the emotional tone, which can then be used to modulate speech synthesis, making it more expressive and contextually appropriate. This capability is particularly beneficial for creating more engaging and lifelike audio outputs in various applications, such as virtual assistants, audiobooks, and interactive storytelling. The node supports both static and dynamic emotion analysis, where dynamic analysis allows for per-segment emotion adjustments using a template system. This flexibility makes it a powerful tool for developers and artists looking to enhance the emotional depth of their audio projects.

🌈 IndexTTS-2 Text Emotion Input Parameters:

qwen_model

The qwen_model parameter specifies the QwenEmotion model to be used for text emotion analysis. This parameter allows you to choose from available models, which can be either downloadable or local. The default model is qwen0.6bemo4-merge, but you can specify a local model by using the local: prefix followed by the model name. This choice impacts the accuracy and style of emotion detection, as different models may have varying capabilities and characteristics. Selecting the appropriate model is crucial for achieving the desired emotional output in your text-to-speech application.

emotion_text

The emotion_text parameter is a string that describes the desired emotion to be applied to the text. It supports dynamic per-segment analysis through the use of the {seg} placeholder, which allows for different emotions to be applied to different segments of text. For example, you might use a template like "Angry man shouting: {seg}" to apply an angry tone to specific segments. If the {seg} placeholder is not used, the same emotion is applied to all segments. This parameter is essential for customizing the emotional tone of the output, providing flexibility in how emotions are expressed in the synthesized speech.

🌈 IndexTTS-2 Text Emotion Output Parameters:

emotion_control

The emotion_control output parameter is a dictionary that contains the emotion control data generated by the QwenEmotion text analysis. This data is used by the IndexTTS-2 adapter to modulate the emotional tone of the synthesized speech. The dictionary includes information such as the type of emotion analysis (qwen_emotion), whether the emotion text is used, the specific emotion text provided, the model used, and whether a dynamic template was applied. This output is crucial for integrating emotion analysis results into the text-to-speech process, enabling more expressive and contextually appropriate audio outputs.

🌈 IndexTTS-2 Text Emotion Usage Tips:

• To achieve the most accurate emotion analysis, ensure that the emotion_text parameter is well-crafted and contextually relevant to the text segments being analyzed. Use the {seg} placeholder for dynamic emotion adjustments when needed.
• Experiment with different QwenEmotion models to find the one that best suits your project's needs. Local models can be specified using the local: prefix, which may offer performance benefits if they are optimized for your specific use case.

🌈 IndexTTS-2 Text Emotion Common Errors and Solutions:

ModelNotFoundError

• Explanation: This error occurs when the specified QwenEmotion model is not found in the available models list.
• Solution: Ensure that the model name is correctly specified and that it is either available for download or exists locally with the correct local: prefix.

InvalidEmotionTextError

• Explanation: This error arises when the emotion_text parameter is not properly formatted or contains invalid placeholders.
• Solution: Verify that the emotion_text is correctly formatted and that the {seg} placeholder is used appropriately for dynamic analysis. Ensure that the text is relevant to the intended emotional output.