Added support for Fish Audio S2.1

2026-05-26 14:54:50 +05:30
parent ccf055ac4e
commit dbcb23e201
9 changed files with 331 additions and 5 deletions
@@ -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",
+}