Fish Audio Text to Speech - ComfyUI Built-in Node Documentation

This node converts written text into spoken audio using Fish Audio text-to-speech models. It supports cues embedded in the text: [happy] and [whispering] on s2.1-pro, (happy) on s1, and plain-language [directions] plus <whisper>...</whisper> word tags on drama-3. Multi-speaker dialogue uses @Voice1/@Voice2 tags when multiple voices are connected. Three models are available: drama-3 and s2.1-pro, which both support up to five voices and multi-speaker dialogue, and s1, which uses a single optional voice.

Inputs

Common Inputs

Parameter	Description	Data Type	Required	Range
`text`	The text to convert to speech. With two or more voices connected, mark speaker changes with @Voice1, @Voice2, etc. (default: empty)	STRING	Yes	Any non-empty text
`model`	Model to use for text-to-speech.	DYNAMIC_COMBO	Yes	”drama-3" "s2.1-pro" "s1”
`seed`	Seed controls whether the node should re-run; results are non-deterministic regardless of seed. (default: 42)	INT	Yes	0 to 2147483647

drama-3 and s2.1-pro Inputs

These inputs appear when the drama-3 or s2.1-pro model is selected. The parameter sets are identical.

Parameter	Description	Data Type	Required	Range
`voices`	Growable slot: connect 0 to 5 voice items (`voice_1`, `voice_2`, …). Voices for synthesis. Leave empty for a random voice. With two or more voices, mark speaker changes in the text with @Voice1, @Voice2, etc.	FISHAUDIO_VOICE	No	0 to 5 voices
`temperature`	Expressiveness. Higher values are more varied, lower values are more consistent. (default: 0.7)	FLOAT	Yes	0.0 to 1.0
`top_p`	Diversity via nucleus sampling. (default: 0.7)	FLOAT	Yes	0.01 to 1.0
`speed`	Speaking rate. 1.0 is normal, <1.0 slower, >1.0 faster. (default: 1.0)	FLOAT	Yes	0.5 to 2.0
`volume`	Volume adjustment in decibels. 0 is no change. (default: 0.0)	FLOAT	Yes	-10.0 to 10.0
`normalize`	Normalize numbers and text for English and Chinese, improving stability for numbers and dates. (default: true)	BOOLEAN	Yes	true / false

s1 Inputs

These inputs appear when the s1 model is selected.

Parameter	Description	Data Type	Required	Range
`voice`	Voice for synthesis. Leave unconnected for a random voice.	FISHAUDIO_VOICE	No	Optional single voice
`temperature`	Expressiveness. Higher values are more varied, lower values are more consistent. (default: 0.7)	FLOAT	Yes	0.0 to 1.0
`top_p`	Diversity via nucleus sampling. (default: 0.7)	FLOAT	Yes	0.01 to 1.0
`speed`	Speaking rate. 1.0 is normal, <1.0 slower, >1.0 faster. (default: 1.0)	FLOAT	Yes	0.5 to 2.0
`volume`	Volume adjustment in decibels. 0 is no change. (default: 0.0)	FLOAT	Yes	-10.0 to 10.0
`normalize`	Normalize numbers and text for English and Chinese, improving stability for numbers and dates. (default: true)	BOOLEAN	Yes	true / false

Note: The text input must not be empty. Speaker tags (@Voice1, @Voice2, etc.) are case-insensitive and must refer to a connected voice; tagging a voice that is not connected raises an error. When two or more voices are connected, the text must reference every connected voice at least once, or the node reports the missing tags. On drama-3 and s2.1-pro, connecting 0 voices uses a random voice, 1 voice uses that voice alone, and 2 or more voices enable multi-speaker dialogue. On s1, a single optional voice is used and leaving it unconnected uses a random voice. Cues can be placed in the text: [happy] and [whispering] on s2.1-pro, (happy) on s1, and plain-language [directions] and <whisper>...</whisper> word tags on drama-3.

Outputs

Output Name	Description	Data Type
`audio`	The generated speech as an audio file.	AUDIO

This documentation was AI-generated. If you find any errors or have suggestions for improvement, please feel free to contribute! Edit on GitHub

Source fingerprint (SHA-256): adcfe8f994b8fbe07efb3dd3f86c8955ea3cec62cc3f0321efd2b63c0470b7a3

​Inputs

​Common Inputs

​drama-3 and s2.1-pro Inputs

​s1 Inputs

​Outputs

Inputs

Common Inputs

drama-3 and s2.1-pro Inputs

s1 Inputs

Outputs