Supertonic Text-to-Speech 🗣️:
SupertonicTTS turns written dialogue or narration into speech using a loaded Supertonic text-to-speech model. You can use it to create voiceovers, character lines, and spoken elements for an audio workflow. It supports preset or custom voice styles, language selection, pacing and quality controls, and expression tags written into the text. The node produces audio you can pass to another compatible node; if synthesis fails, it returns silence instead of a generated voice clip.
Supertonic Text-to-Speech 🗣️ Input Parameters:
model
Required. Connect a SUPERTONIC_MODEL produced by a compatible model-loading node. This supplies the speech synthesizer and its audio sample rate. There is no default or selectable range for this connection. Without a suitable model, the node cannot generate speech.
text
Required. The words you want spoken. The default is Hello! <laugh> This is Supertonic-3 speaking. The field accepts multiple lines and has no stated length limit. You can include expression tags such as <laugh> in the script; the node passes the text to the synthesizer and reports any recognized tags when verbose logging is on. Empty or whitespace-only text produces one second of silence rather than spoken audio.
language
Required. Select the language code used for synthesis. The default is en. The dropdown draws its choices from SUPPORTED_LANGUAGES, but the provided source does not show that list, so its complete options cannot be stated here. Choose the option that matches the language of your script; the selected code is passed to the synthesizer.
speed
Required. Controls speech pacing. The default is 1.0, with a minimum of 0.5, a maximum of 2.0, and increments of 0.05. Lower values request slower speech, while higher values request faster speech. Start at 1.0 and adjust in small increments when timing a line to a scene.
steps
Required. Sets the number of diffusion steps used during synthesis. The default is 8, with a minimum of 5, a maximum of 12, and increments of 1. The node’s tooltip describes higher values as smoother but slower: 5 favors speed, while 12 is its maximum-quality setting. Use 8 as a starting point when balancing generation time and results.
voice_style
Required. Selects a preset voice style. The default is M1. The dropdown draws its choices from BUILTIN_VOICES, but the provided source does not show the complete list. The selected preset is used unless custom_style_path points to an existing file, in which case that custom style takes precedence.
custom_style_path
Optional. Enter an absolute path to a custom Supertonic voice-style .json file, such as one made with Voice Builder. The default is an empty string, which uses the selected voice_style preset. If the path points to an existing file, the node loads that style instead of the preset. If the path does not point to a file, it prints a warning and falls back to the preset.
verbose
Optional. Controls extra diagnostic messages. The default is False; the available values are True and False. When enabled, the node prints the selected settings, detected expression tags, a preview of the text, and details about the synthesized audio. It does not change the requested voice settings, but its messages can help you diagnose an unexpected result.
Supertonic Text-to-Speech 🗣️ Output Parameters:
audio
An AUDIO value containing the generated speech for use by downstream audio nodes. It carries the waveform and the sample rate supplied by the synthesizer. The node converts the returned waveform into a mono ComfyUI-compatible audio value. If the text is empty or whitespace-only, it returns one second of silence. If an exception occurs during synthesis, it prints an error and traceback and also returns one second of silence, so a silent result does not always mean the script was intentionally blank.
Supertonic Text-to-Speech 🗣️ Usage Tips:
- Start with
speedat1.0andstepsat8. For quick previews, trystepsat5; for a smoother result when generation time matters less, try a higher value up to12. - Choose
languageto match the script, then audition the availablevoice_stylepresets before changing speed. This makes it easier to judge the voice separately from its pacing. - To use a custom voice, provide the absolute path to an existing style
.jsonfile. Clearcustom_style_pathwhen you want thevoice_styledropdown to control the result again. - Place expression tags directly in the script where you want them, for example
Hello! <laugh> That was funny!. The text field also identifies<breath>,<surprise>,<sigh>,<scream>,<throatclear>,<sad>,<angry>,<cough>, and<yawn>as available tag examples. - Turn on
verbosewhile troubleshooting a silent or unexpected clip. Check its text preview and synthesis details, then turn it off when you no longer need the extra messages.
Supertonic Text-to-Speech 🗣️ Common Errors and Solutions:
[ Supertonic Text-to-Speech 🗣️] Empty text — returning silence
- Explanation: This message appears when
verboseis on andtextis empty or contains only whitespace. The node returns one second of silence. - Solution: Enter words in
textand run the node again. Check that an upstream text connection is supplying nonblank content, if applicable.
[Warn] custom_style_path not found: ... — falling back to preset '...'
- Explanation: The supplied
custom_style_pathdoes not point to an existing file. The node continues with the selected presetvoice_style. - Solution: Check the full path and filename of your custom style file, or clear the field if you intend to use the preset voice.
[ Supertonic Text-to-Speech 🗣️] ERROR — ...
- Explanation: Synthesis raised an exception. The node prints the underlying error and traceback to the console and returns one second of silence. The exact text after the dash depends on what failed.
- Solution: Read the message after the dash for the specific cause. Confirm that the connected model works, the selected language and style are appropriate for it, and any custom style file is valid; then retry with a short, plain-text script.
KeyError: 'sample_rate'
- Explanation: An empty-text request needs the model’s
sample_rateto create silence, but the connected model value does not provide it. - Solution: Use a compatible
SUPERTONIC_MODELconnection from the intended model loader, then retry.
KeyError: 'tts'
- Explanation: For nonempty text, the node expects the connected model value to contain a speech synthesizer under
tts. - Solution: Reconnect or reload a compatible
SUPERTONIC_MODEL. Check the model-loading step before changing the script or voice settings.
