Added support for Fish Audio S2.1

__init__.py

@@ -60,6 +60,7 @@
 from .modules.textInferenceInputsVideos import RunwareTextInferenceInputsVideos
 from .modules.audioSections import RunwareAudioSections
 from .modules.audioInferenceInputs import audioInferenceInputs
+from .modules.audioInferenceReferenceVoices import RunwareAudioInferenceReferenceVoices
 from .modules.audioSettings import RunwareAudioSettings
 from .modules.audioSettingsVoiceModify import RunwareAudioSettingsVoiceModify
 from .modules.providerSettings.elevenlabsProviderSettings import RunwareElevenLabsProviderSettings
@@ -191,6 +192,7 @@
     "Runware Text Inference Inputs Videos": RunwareTextInferenceInputsVideos,
     "Runware Audio Sections": RunwareAudioSections,
     "Runware Audio Inference Inputs": audioInferenceInputs,
+    "Runware Audio Inference Inputs Reference Audio": RunwareAudioInferenceReferenceVoices,
     "Runware Audio Inference Settings": RunwareAudioSettings,
     "Runware Audio Inference Settings Voice Modify": RunwareAudioSettingsVoiceModify,
     "Runware ElevenLabs Provider Settings": RunwareElevenLabsProviderSettings,

clientlibs/main.js

@@ -1,6 +1,6 @@
 import { app } from "../../scripts/app.js";
 import { api } from "../../scripts/api.js";
-import { promptEnhanceHandler, syncDimensionsNodeHandler, searchNodeHandler, APIKeyHandler, captionNodeHandler, saveTextHandler, mediaUUIDHandler, save3DFilepathHandler, videoTranscriptionHandler, videoOutputsHandler, handleCustomErrors, videoInferenceDimensionsHandler, videoModelSearchFilterHandler, audioModelSearchFilterHandler, textModelSearchFilterHandler, vectorizeModelSearchFilterHandler, vectorizeToggleHandler, useParameterToggleHandler, imageInferenceToggleHandler, imageInferenceAdvancedFeaturesToggleHandler, watermarkAdvancedFeatureToggleHandler, videoInferenceSpeechInputToggleHandler, regionalPromptingRegionsToggleHandler, upscalerToggleHandler, imageUpscalerSettingsToggleHandler, videoUpscalerToggleHandler, audioInferenceToggleHandler, audioInferenceSpeechToggleHandler, audioSettingsToggleHandler, textInferenceSettingsToggleHandler, videoSettingsToggleHandler, videoInferenceSettingsTtsToggleHandler, videoInferenceSettingsActiveSpeakerDetectionToggleHandler, videoInferenceSettingsActiveSpeakerBoundingBoxesToggleHandler, videoInferenceSettingsSegmentsToggleHandler, acceleratorOptionsToggleHandler, bytedanceProviderSettingsToggleHandler, xaiProviderSettingsToggleHandler, viduProviderSettingsToggleHandler, sourcefulProviderSettingsToggleHandler, sourcefulProviderSettingsFontsToggleHandler, threeDInferenceToggleHandler, threeDInferenceSettingsToggleHandler, threeDInferenceSettingsLatToggleHandler, threeDInferenceSettingsMeshClusterToggleHandler, ultralyticsProviderSettingsToggleHandler, openaiProviderSettingsToggleHandler, lightricksProviderSettingsToggleHandler, klingProviderSettingsToggleHandler, lumaProviderSettingsToggleHandler, briaProviderSettingsToggleHandler, pixverseProviderSettingsToggleHandler, alibabaProviderSettingsToggleHandler, mireloProviderSettingsToggleHandler, googleProviderSettingsToggleHandler, syncProviderSettingsToggleHandler, syncSegmentToggleHandler, settingsToggleHandler, outpaintSettingsToggleHandler, safetyInputsToggleHandler, imageInferenceSettingsColorPaletteToggleHandler, imageInferenceSettingsMoodboardsToggleHandler, audioInputToggleHandler, speechInputToggleHandler, briaProviderMaskToggleHandler, wanAnimateAdvancedFeatureSettingsToggleHandler, videoAdvancedFeatureInputsToggleHandler, audioInferenceInputsToggleHandler, audioInferenceSpeechVoicesToggleHandler, referenceVideosToggleHandler } from "./utils.js";
+import { promptEnhanceHandler, syncDimensionsNodeHandler, searchNodeHandler, APIKeyHandler, captionNodeHandler, saveTextHandler, mediaUUIDHandler, save3DFilepathHandler, videoTranscriptionHandler, videoOutputsHandler, handleCustomErrors, videoInferenceDimensionsHandler, videoModelSearchFilterHandler, audioModelSearchFilterHandler, textModelSearchFilterHandler, vectorizeModelSearchFilterHandler, vectorizeToggleHandler, useParameterToggleHandler, imageInferenceToggleHandler, imageInferenceAdvancedFeaturesToggleHandler, watermarkAdvancedFeatureToggleHandler, videoInferenceSpeechInputToggleHandler, regionalPromptingRegionsToggleHandler, upscalerToggleHandler, imageUpscalerSettingsToggleHandler, videoUpscalerToggleHandler, audioInferenceToggleHandler, audioInferenceSpeechToggleHandler, audioSettingsToggleHandler, textInferenceSettingsToggleHandler, videoSettingsToggleHandler, videoInferenceSettingsTtsToggleHandler, videoInferenceSettingsActiveSpeakerDetectionToggleHandler, videoInferenceSettingsActiveSpeakerBoundingBoxesToggleHandler, videoInferenceSettingsSegmentsToggleHandler, acceleratorOptionsToggleHandler, bytedanceProviderSettingsToggleHandler, xaiProviderSettingsToggleHandler, viduProviderSettingsToggleHandler, sourcefulProviderSettingsToggleHandler, sourcefulProviderSettingsFontsToggleHandler, threeDInferenceToggleHandler, threeDInferenceSettingsToggleHandler, threeDInferenceSettingsLatToggleHandler, threeDInferenceSettingsMeshClusterToggleHandler, ultralyticsProviderSettingsToggleHandler, openaiProviderSettingsToggleHandler, lightricksProviderSettingsToggleHandler, klingProviderSettingsToggleHandler, lumaProviderSettingsToggleHandler, briaProviderSettingsToggleHandler, pixverseProviderSettingsToggleHandler, alibabaProviderSettingsToggleHandler, mireloProviderSettingsToggleHandler, googleProviderSettingsToggleHandler, syncProviderSettingsToggleHandler, syncSegmentToggleHandler, settingsToggleHandler, outpaintSettingsToggleHandler, safetyInputsToggleHandler, imageInferenceSettingsColorPaletteToggleHandler, imageInferenceSettingsMoodboardsToggleHandler, audioInputToggleHandler, speechInputToggleHandler, briaProviderMaskToggleHandler, wanAnimateAdvancedFeatureSettingsToggleHandler, videoAdvancedFeatureInputsToggleHandler, audioInferenceInputsToggleHandler, audioInferenceReferenceVoiceToggleHandler, audioInferenceSpeechVoicesToggleHandler, referenceVideosToggleHandler } from "./utils.js";
 import { RUNWARE_NODE_TYPES, RUNWARE_NODE_PROPS, SEARCH_TERMS } from "./types.js";
 
 const nodeInitList = [];
@@ -165,6 +165,8 @@ app.registerExtension({
             videoAdvancedFeatureInputsToggleHandler(node);
         } else if(nodeClass === RUNWARE_NODE_TYPES.AUDIOINFERENCEINPUTS) {
             audioInferenceInputsToggleHandler(node);
+        } else if(nodeClass === RUNWARE_NODE_TYPES.AUDIOINFERENCEINPUTSREFERENCEAUDIO) {
+            audioInferenceReferenceVoiceToggleHandler(node);
         } else if(nodeClass === RUNWARE_NODE_TYPES.REFERENCEVIDEOS ||
                   nodeClass === RUNWARE_NODE_TYPES.REFERENCEVIDEOS_LEGACY) {
             referenceVideosToggleHandler(node);

clientlibs/types.js

@@ -96,6 +96,7 @@ const RUNWARE_NODE_TYPES = {
     TEXTINFERENCEINPUTSVIDEOS: "Runware Text Inference Inputs Videos",
     AUDIOSECTIONS: "Runware Audio Sections",
     AUDIOINFERENCEINPUTS: "Runware Audio Inference Inputs",
+    AUDIOINFERENCEINPUTSREFERENCEAUDIO: "Runware Audio Inference Inputs Reference Audio",
     AUDIOSETTINGS: "Runware Audio Inference Settings",
     AUDIOSETTINGSVOICEMODIFY: "Runware Audio Inference Settings Voice Modify",
     PIXVERSEPROVIDERSETTINGS: "Runware Pixverse Provider Settings",
@@ -574,6 +575,10 @@ const RUNWARE_NODE_PROPS = {
         bgColor: DEFAULT_BGCOLOR,
         colorModeOnly: true,
     },
+    [RUNWARE_NODE_TYPES.AUDIOINFERENCEINPUTSREFERENCEAUDIO]: {
+        bgColor: DEFAULT_BGCOLOR,
+        colorModeOnly: true,
+    },
     [RUNWARE_NODE_TYPES.AUDIOSETTINGS]: {
         bgColor: DEFAULT_BGCOLOR,
     },

clientlibs/utils.js

@@ -1298,6 +1298,26 @@ function audioSettingsToggleHandler(settingsNode) {
     const cfgIntervalStartWidget = settingsNode.widgets.find(w => w && w.name === "cfgIntervalStart");
     const useCfgIntervalEndWidget = settingsNode.widgets.find(w => w && w.name === "useCfgIntervalEnd");
     const cfgIntervalEndWidget = settingsNode.widgets.find(w => w && w.name === "cfgIntervalEnd");
+    const useNormalizeLoudnessWidget = settingsNode.widgets.find(w => w && w.name === "useNormalizeLoudness");
+    const normalizeLoudnessWidget = settingsNode.widgets.find(w => w && w.name === "normalizeLoudness");
+    const useTopPWidget = settingsNode.widgets.find(w => w && w.name === "useTopP");
+    const topPWidget = settingsNode.widgets.find(w => w && w.name === "topP");
+    const useChunkLengthWidget = settingsNode.widgets.find(w => w && w.name === "useChunkLength");
+    const chunkLengthWidget = settingsNode.widgets.find(w => w && w.name === "chunkLength");
+    const useMinChunkLengthWidget = settingsNode.widgets.find(w => w && w.name === "useMinChunkLength");
+    const minChunkLengthWidget = settingsNode.widgets.find(w => w && w.name === "minChunkLength");
+    const useNormalizeWidget = settingsNode.widgets.find(w => w && w.name === "useNormalize");
+    const normalizeWidget = settingsNode.widgets.find(w => w && w.name === "normalize");
+    const useLatencyWidget = settingsNode.widgets.find(w => w && w.name === "useLatency");
+    const latencyWidget = settingsNode.widgets.find(w => w && w.name === "latency");
+    const useMaxTokensWidget = settingsNode.widgets.find(w => w && w.name === "useMaxTokens");
+    const maxTokensWidget = settingsNode.widgets.find(w => w && w.name === "maxTokens");
+    const useRepetitionPenaltyWidget = settingsNode.widgets.find(w => w && w.name === "useRepetitionPenalty");
+    const repetitionPenaltyWidget = settingsNode.widgets.find(w => w && w.name === "repetitionPenalty");
+    const useConditionOnPreviousChunksWidget = settingsNode.widgets.find(w => w && w.name === "useConditionOnPreviousChunks");
+    const conditionOnPreviousChunksWidget = settingsNode.widgets.find(w => w && w.name === "conditionOnPreviousChunks");
+    const useEarlyStopThresholdWidget = settingsNode.widgets.find(w => w && w.name === "useEarlyStopThreshold");
+    const earlyStopThresholdWidget = settingsNode.widgets.find(w => w && w.name === "earlyStopThreshold");
 
     function toggleWidgetState(useWidget, paramWidget, paramName) {
         if (!useWidget || !paramWidget) return;
@@ -1338,6 +1358,16 @@ function audioSettingsToggleHandler(settingsNode) {
     if (useTranscriptWidget && transcriptWidget) toggleWidgetState(useTranscriptWidget, transcriptWidget, "transcript");
     if (useCfgIntervalStartWidget && cfgIntervalStartWidget) toggleWidgetState(useCfgIntervalStartWidget, cfgIntervalStartWidget, "cfgIntervalStart");
     if (useCfgIntervalEndWidget && cfgIntervalEndWidget) toggleWidgetState(useCfgIntervalEndWidget, cfgIntervalEndWidget, "cfgIntervalEnd");
+    if (useNormalizeLoudnessWidget && normalizeLoudnessWidget) toggleWidgetState(useNormalizeLoudnessWidget, normalizeLoudnessWidget, "normalizeLoudness");
+    if (useTopPWidget && topPWidget) toggleWidgetState(useTopPWidget, topPWidget, "topP");
+    if (useChunkLengthWidget && chunkLengthWidget) toggleWidgetState(useChunkLengthWidget, chunkLengthWidget, "chunkLength");
+    if (useMinChunkLengthWidget && minChunkLengthWidget) toggleWidgetState(useMinChunkLengthWidget, minChunkLengthWidget, "minChunkLength");
+    if (useNormalizeWidget && normalizeWidget) toggleWidgetState(useNormalizeWidget, normalizeWidget, "normalize");
+    if (useLatencyWidget && latencyWidget) toggleWidgetState(useLatencyWidget, latencyWidget, "latency");
+    if (useMaxTokensWidget && maxTokensWidget) toggleWidgetState(useMaxTokensWidget, maxTokensWidget, "maxTokens");
+    if (useRepetitionPenaltyWidget && repetitionPenaltyWidget) toggleWidgetState(useRepetitionPenaltyWidget, repetitionPenaltyWidget, "repetitionPenalty");
+    if (useConditionOnPreviousChunksWidget && conditionOnPreviousChunksWidget) toggleWidgetState(useConditionOnPreviousChunksWidget, conditionOnPreviousChunksWidget, "conditionOnPreviousChunks");
+    if (useEarlyStopThresholdWidget && earlyStopThresholdWidget) toggleWidgetState(useEarlyStopThresholdWidget, earlyStopThresholdWidget, "earlyStopThreshold");
 }
 
 function textInferenceSettingsToggleHandler(settingsNode) {
@@ -3423,6 +3453,9 @@ function audioModelSearchFilterHandler(audioModelSearchNode) {
         "Google": [
             "google:gemini@3.1-flash-tts (Gemini 3.1 Flash TTS)",
         ],
+        "Fish": [
+            "fishaudio:s2.1@pro (Fish Audio S2.1 Pro)",
+        ],
     };
 
     function filterModelList() {
@@ -4727,6 +4760,43 @@ function regionalPromptingRegionsToggleHandler(regionsNode) {
     }
 }
 
+function audioInferenceReferenceVoiceToggleHandler(referenceVoiceNode) {
+    if (!referenceVoiceNode?.widgets) return;
+    if (referenceVoiceNode._audioInferenceReferenceVoiceToggleHandlerRegistered) return;
+    referenceVoiceNode._audioInferenceReferenceVoiceToggleHandlerRegistered = true;
+
+    function toggleWidgetState(useWidget, paramWidget, paramName) {
+        if (!useWidget || !paramWidget) return;
+
+        function applyState() {
+            const enabled = useWidget.value === true;
+            toggleWidgetEnabled(paramWidget, enabled, referenceVoiceNode);
+            if (paramWidget.options && paramWidget.options.element) {
+                paramWidget.options.element.disabled = !enabled;
+                paramWidget.options.element.style.opacity = enabled ? "1" : "0.5";
+                paramWidget.options.element.style.pointerEvents = enabled ? "auto" : "none";
+            }
+            referenceVoiceNode.setDirtyCanvas(true);
+        }
+
+        appendWidgetCB(useWidget, () => setTimeout(applyState, 50));
+        setTimeout(applyState, 100);
+    }
+
+    for (let i = 1; i <= 4; i++) {
+        const useReferenceVoiceWidget = referenceVoiceNode.widgets.find((w) => w && w.name === `useReferenceVoice${i}`);
+        const audioWidget = referenceVoiceNode.widgets.find((w) => w && w.name === `audio${i}`);
+        const textWidget = referenceVoiceNode.widgets.find((w) => w && w.name === `text${i}`);
+
+        if (useReferenceVoiceWidget && audioWidget) {
+            toggleWidgetState(useReferenceVoiceWidget, audioWidget, `audio${i}`);
+        }
+        if (useReferenceVoiceWidget && textWidget) {
+            toggleWidgetState(useReferenceVoiceWidget, textWidget, `text${i}`);
+        }
+    }
+}
+
 function audioInferenceInputsToggleHandler(audioInputsNode) {
     if (!audioInputsNode?.widgets) return;
 
@@ -5096,6 +5166,7 @@ export {
     wanAnimateAdvancedFeatureSettingsToggleHandler,
     videoAdvancedFeatureInputsToggleHandler,
     audioInferenceInputsToggleHandler,
+    audioInferenceReferenceVoiceToggleHandler,
     audioInferenceSpeechVoicesToggleHandler,
     referenceVideosToggleHandler,
 };

modules/audioInferenceInputs.py

@@ -52,13 +52,21 @@ def INPUT_TYPES(cls):
                 "tooltip": f"Audio URL or mediaUUID for the {ordinal} audio. Only used when 'Use Audios' is enabled.",
                 "default": "",
             })
+
+        optionalInputs["Reference Voice"] = ("RUNWAREAUDIOINFERENCEREFERENCEVOICES", {
+            "tooltip": "Connect Runware Audio Inference Inputs Reference Audio for zero-shot voice cloning (inputs.referenceVoices).",
+        })
 
         return {
             "required": {},
             "optional": optionalInputs
         }
 
-    DESCRIPTION = "Configure custom inputs for Runware Audio Inference, including optional single or multiple audio URL/mediaUUID (inputs.audio or inputs.audios), and single or multiple video inputs for audio extraction or generation."
+    DESCRIPTION = (
+        "Configure custom inputs for Runware Audio Inference, including optional single or multiple audio URL/mediaUUID "
+        "(inputs.audio or inputs.audios), reference voice for cloning (inputs.referenceVoices), "
+        "and single or multiple video inputs for audio extraction or generation."
+    )
     FUNCTION = "createInputs"
     RETURN_TYPES = ("RUNWAREAUDIOINFERENCEINPUTS",)
     RETURN_NAMES = ("Audio Inference Inputs",)
@@ -72,6 +80,7 @@ def createInputs(self, **kwargs) -> tuple[Dict[str, Any]]:
         useVideo = kwargs.get("useVideo", False)
         video = kwargs.get("Video", None)
         useVideos = kwargs.get("useVideos", False)
+        referenceVoices = kwargs.get("Reference Voice", None)
 
         inputs = {}
 
@@ -106,6 +115,9 @@ def createInputs(self, **kwargs) -> tuple[Dict[str, Any]]:
 
             if len(videoList) > 0:
                 inputs["videos"] = videoList
+
+        if referenceVoices is not None and isinstance(referenceVoices, list) and len(referenceVoices) > 0:
+            inputs["referenceVoices"] = referenceVoices
 
         return (inputs,)
 

modules/audioInferenceReferenceVoices.py

@@ -0,0 +1,74 @@
+"""
+Runware Audio Inference Inputs Reference Audio node.
+Builds inputs.referenceVoices for zero-shot voice cloning (up to 4 entries).
+"""
+
+from typing import Any, Dict, List
+
+from .utils import runwareUtils as rwUtils
+
+
+class RunwareAudioInferenceReferenceVoices:
+    """Build inputs.referenceVoices[] for Fish Audio and other TTS models."""
+
+    MAX_REFERENCE_VOICES = 4
+
+    @classmethod
+    def INPUT_TYPES(cls):
+        optional_inputs = {}
+        for i in range(1, cls.MAX_REFERENCE_VOICES + 1):
+            ordinal = rwUtils.getOrdinal(i)
+            optional_inputs[f"useReferenceVoice{i}"] = ("BOOLEAN", {
+                "default": False,
+                "tooltip": f"Enable to include the {ordinal} reference voice in inputs.referenceVoices.",
+            })
+            optional_inputs[f"audio{i}"] = ("STRING", {
+                "default": "",
+                "tooltip": f"Reference audio clip ({ordinal}) as media UUID, URL, or base64. Required when enabled.",
+            })
+            optional_inputs[f"text{i}"] = ("STRING", {
+                "multiline": True,
+                "default": "",
+                "tooltip": f"Transcript of the {ordinal} reference audio clip (1–1000 characters). Required when enabled.",
+            })
+
+        return {
+            "required": {},
+            "optional": optional_inputs,
+        }
+
+    RETURN_TYPES = ("RUNWAREAUDIOINFERENCEREFERENCEVOICES",)
+    RETURN_NAMES = ("referenceVoices",)
+    FUNCTION = "createReferenceVoices"
+    CATEGORY = "Runware/Audio"
+    DESCRIPTION = (
+        "Configure inputs.referenceVoices for zero-shot voice cloning (up to 4 entries). "
+        "Each entry: { \"audio\": \"<UUID/URL/base64>\", \"text\": \"<transcript>\" }. "
+        "Connect to Runware Audio Inference Inputs."
+    )
+
+    def createReferenceVoices(self, **kwargs) -> tuple[List[Dict[str, Any]]]:
+        reference_voices: List[Dict[str, Any]] = []
+
+        for i in range(1, self.MAX_REFERENCE_VOICES + 1):
+            if not kwargs.get(f"useReferenceVoice{i}", False):
+                continue
+
+            audio = (kwargs.get(f"audio{i}") or "").strip()
+            text = (kwargs.get(f"text{i}") or "").strip()
+
+            if not audio or not text:
+                continue
+
+            reference_voices.append({"audio": audio, "text": text})
+
+        return (reference_voices,)
+
+
+NODE_CLASS_MAPPINGS = {
+    "RunwareAudioInferenceReferenceVoices": RunwareAudioInferenceReferenceVoices,
+}
+
+NODE_DISPLAY_NAME_MAPPINGS = {
+    "RunwareAudioInferenceReferenceVoices": "Runware Audio Inference Inputs Reference Audio",
+}

modules/audioModelSearch.py

@@ -47,6 +47,9 @@ class RunwareAudioModelSearch:
         "Google": [
             "google:gemini@3.1-flash-tts (Gemini 3.1 Flash TTS)",
         ],
+        "Fish": [
+            "fishaudio:s2.1@pro (Fish Audio S2.1 Pro)",
+        ],
     }
 
     MODEL_PROVIDERS = [
@@ -61,6 +64,7 @@ class RunwareAudioModelSearch:
         "MiniMax",
         "Inworld",
         "Google",
+        "Fish",
     ]
 
     @classmethod