Breeze TTS 2 · T8 桌面音色包:
This node brings a voice exported from the desktop voice library into a Breeze TTS workflow. It reads the local voice bundle offline, prepares a synthesis request using the voice’s saved settings, and preserves inline sound-event markers in the dialogue for later synthesis. You can keep the voice’s existing delivery, give an individual line new direction, or choose a neutral delivery without changing the exported bundle. The node prepares the request; it does not synthesize the finished speech itself.
Breeze TTS 2 · T8 桌面音色包 Input Parameters:
bundle_path
The local path to a desktop-exported .t8voice.zip voice bundle. The default is voice.t8voice.zip. There is no stated minimum or maximum path length. The node reads the bundle offline and does not extract it to disk. This path determines which saved voice profile and, when present, reference recording are used. Enter a path that points to an accessible, valid bundle; the default filename works only if that file can be found from the node’s environment.
text
The dialogue to prepare for synthesis. The default is [咳嗽] 欢迎使用桌面版与 ComfyUI 共用的音色。. This is a multiline text field with no stated length limit, but it cannot be empty after surrounding whitespace is removed. Inline sound-event markers in your text are retained in the request. The node does not alter the wording beyond trimming whitespace at the beginning and end.
line_direction_mode
Controls how the delivery instruction for this line is chosen. The options are inherit (default), override, and neutral; no numeric minimum or maximum applies. inherit uses the bundle’s saved instruction where appropriate, override uses your line_direction and requires it to be nonempty, and neutral selects a neutral route. The effective request mode also depends on whether the bundle contains reference audio: an override becomes a direction request when reference audio exists, or a design request when it does not. Neutral becomes a clone request with no instruction when reference audio exists, or a design request using the built-in default instruction when it does not.
line_direction
An optional natural-language instruction for this line’s emotion, pace, or expression. Its default is an empty string, and no length limit is stated. It is required only when line_direction_mode is override; in that mode, it replaces the saved instruction. For example, you might enter Speak gently and at a slower pace.. In inherit and neutral modes, this field does not determine the request instruction.
cfg_scale
An optional per-line CFG setting. The default is 0.0, the displayed minimum is 0.0, the maximum is 10.0, and the step is 0.1. Set it to 0 to use the value saved in the voice bundle; a nonzero value overrides that value for this request and must be from 0.1 through 10.0. This node passes the selected value into the request rather than producing audio itself, so judge its audible effect when you run the request through your synthesis workflow.
Breeze TTS 2 · T8 桌面音色包 Output Parameters:
request
A BREEZE_T8_REQUEST containing the prepared synthesis settings. It includes the effective mode, trimmed dialogue, selected CFG value, language, voice ID and name, and chosen line-direction mode. It includes an instruction only when one applies; when the bundle has reference audio, it also includes that audio and the bundle’s reference text. Connect this output to a compatible downstream node that accepts a Breeze T8 request. This output is a request, not finished speech.
reference_audio
An AUDIO output carrying the reference recording decoded from the bundle when one is present. It lets you access that recording separately from the prepared request. If the bundle has no reference-audio payload, the node’s code passes through the result of the decoder without creating a recording; do not assume this socket always contains usable audio.
voice_info
A STRING containing readable, indented JSON information from the loaded voice profile. It excludes bundle_path and adds effective_mode and line_direction_mode, so you can inspect which route the node selected for this line. This is informational metadata, not synthesized audio or a substitute for the request output.
Breeze TTS 2 · T8 桌面音色包 Usage Tips:
- Start with
line_direction_modeset toinheritandcfg_scaleset to0to use the bundle’s saved settings. Checkvoice_infoto see the effective mode. - For a one-line change in delivery, select
overrideand write a specific instruction inline_direction, such asSpeak calmly, with a brief pause before the final sentence.. You do not need to change the bundle’s saved instruction. - Select
neutralwhen you do not want the bundle’s saved delivery instruction applied to this line. A bundle with reference audio takes the clone route; one without it takes the design route with a built-in default instruction. - Keep intended inline sound-event markers in
text. The node preserves them in the prepared request, though the final rendition depends on the downstream synthesis workflow. - If you change or replace the file at
bundle_path, rerun the node. Its change check uses a bundle fingerprint so the workflow can detect changes to the bundle.
Breeze TTS 2 · T8 桌面音色包 Common Errors and Solutions:
text 不能为空。
- Explanation: The dialogue is empty or contains only whitespace after trimming.
- Solution: Enter dialogue in
text. Check that a connected text source supplies visible characters before running the node.
line_direction_mode=override 时 line_direction 不能为空。
- Explanation: You selected
override, butline_directionis empty or contains only whitespace. - Solution: Enter a delivery instruction in
line_direction, or switch toinheritorneutralif you do not need a per-line instruction.
line_direction_mode 必须是 inherit、override 或 neutral。
- Explanation: The supplied direction mode is not one of the three accepted choices. This is most likely when a value comes from another part of a workflow rather than the node’s dropdown.
- Solution: Set
line_direction_modeto exactlyinherit,override, orneutral.
cfg_scale 必须为 0(继承)或 0.1 到 10.0。
- Explanation: The effective CFG value is outside the accepted range. This can come from an invalid override or from the bundle’s saved CFG value when
cfg_scaleis0. - Solution: Set
cfg_scaleto a value from0.1through10.0. If you want to inherit the bundle setting, use0and verify that the bundle contains a valid saved CFG value.
Voice bundle cannot be loaded or reference audio cannot be decoded
- Explanation: The node could not read the file, load its voice profile, or decode its reference recording. The exact error text depends on the bundle-loading and audio-decoding implementations, which are not included in the provided source.
- Solution: Check that
bundle_pathpoints to an accessible desktop-exported.t8voice.zipfile. Re-export the bundle if it is damaged or incomplete, then rerun the node.
