支持V3omni模型

README.md

@@ -14,7 +14,7 @@ It bundles image generation, text-to-video, image-to-video, multi-image video, a
 
 ## Current model visibility
 
-The plugin keeps the visible model dropdowns aligned with models verified on the active endpoint.
+The plugin keeps the visible model dropdowns aligned with models confirmed for the active integration paths.
 
 Visible image generation models:
 
@@ -34,6 +34,7 @@ Visible text-to-video models:
 - `kling-v2-5-turbo`
 - `kling-v2-6`
 - `kling-v3`
+- `kling-v3-omni`
 
 Visible image-to-video models:
 
@@ -46,26 +47,29 @@ Visible image-to-video models:
 - `kling-v2-5-turbo`
 - `kling-v2-6`
 - `kling-v3`
+- `kling-v3-omni`
 
 Visible multi-image video models:
 
 - `kling-v2-1`
 - `kling-v2-5-turbo`
 - `kling-v3`
+- `kling-v3-omni`
 
 Visible video generation modes:
 
 - `std`
 - `pro`
-- `4k` (`kling-v3` text-to-video, image-to-video, and multi-image video only; the API uses `mode: "4k"`)
+- `4k` (`kling-v3` and `kling-v3-omni` text-to-video, image-to-video, and multi-image video; the API uses `mode: "4k"`)
 
-Documented newer models such as `kling-image-o1`, `kling-video-o1`, and `kling-v3-omni` still exist in the internal capability map, but are hidden from the UI until live endpoint support is confirmed for normal accounts.
+Documented newer models such as `kling-image-o1` and `kling-video-o1` still exist in the internal capability map, but are hidden from the UI until live endpoint support is confirmed for normal accounts.
 
 ## Important limitations
 
 - Native 4K video output is exposed through the video `mode` dropdown. Models that do not support `mode=4k` will be rejected by the node capability checks before submission.
 - Native preset and cloned voices are available for `kling-v2-6`; `Image2Video.custom_voice_id` sends cloned voice control through `voice_list`. The active API rejects this field for `kling-v3`.
-- `kling-v3-omni` and `kling-video-o1` remain hidden until live endpoint support is confirmed for normal accounts.
+- `kling-v3-omni` is visible for text-to-video, image-to-video, and multi-image video comparison. It can combine `voice_list`/`custom_voice_id` with `element_list`; reference-video requests remain guarded because native audio and `mode=4k` are rejected when `reference_video` is provided.
+- `kling-video-o1` remains hidden until live endpoint support is confirmed for normal accounts.
 - `Advanced Element Create` requires 1 frontal subject image plus 1-3 additional reference images of the same subject. Background images do not count as reference images.
 - `Advanced Element Create` also enforces `element_name <= 20` characters and `element_description <= 100` characters.
 - If you need both strong subject consistency and voiced output, the reliable flow on the current endpoint is usually: generate the bound subject video first, then add speech with `Lip Sync`, `TextToAudio`, or `Video2Audio`.
@@ -150,6 +154,7 @@ Python dependencies are listed in [requirements.txt](./requirements.txt).
 - `Comfyui-Kling-Wrapper TTS`
 - `Comfyui-Kling-Wrapper Custom Voice Create`
 - `Comfyui-Kling-Wrapper Custom Voice Query`
+- `Comfyui-Kling-Wrapper Custom Voice List Builder`
 - `Comfyui-Kling-Wrapper Video2Audio`
 - `Comfyui-Kling-Wrapper Advanced Element Create`
 - `Comfyui-Kling-Wrapper Advanced Element Query`
@@ -219,7 +224,7 @@ When `duration` is set to `auto`, the node tries to match the uploaded reference
 
 ## Voice presets
 
-`Text2Video` and `Image2Video` expose a `voice_preset` dropdown. `Image2Video` additionally accepts a connected `custom_voice_id`.
+`Text2Video` and `Image2Video` expose a `voice_preset` dropdown. `Image2Video` additionally accepts a connected `custom_voice_id` for one custom voice, or a connected `custom_voice_list` from `Custom Voice List Builder` for multiple custom voices.
 
 The selected preset is converted to:
 
@@ -229,6 +234,8 @@ The selected preset is converted to:
 
 These presets are mainly intended for models that support native audio generation on the active endpoint, especially `kling-v2-6`.
 
+For multiple custom voices, connect one or more `Custom Voice Create.voice_id` outputs to `Custom Voice List Builder`, then connect its `custom_voice_list` output to `Image2Video.custom_voice_list`. Set `sound=on` and reference each voice in order with prompt markers such as `<<<voice_1>>>` and `<<<voice_2>>>`. `custom_voice_list`, `custom_voice_id`, and `voice_preset` are mutually exclusive.
+
 For official or manually supplied preset voice IDs, use the dedicated `TTS` node. It calls `/v1/audio/tts` and returns an `audio_url` that can be saved or previewed with `Preview Audio`.
 
 ## Examples

examples/README.md

@@ -11,17 +11,18 @@ This directory contains importable ComfyUI workflow JSON examples and small Pyth
 - Replace placeholder filenames such as `example_portrait.png`, `example_subject_front.png`, `example_subject_ref_1.png`, `example_cloth.png`, `example_scene.png`, `example_lipsync_source.mp4`, `example_lipsync_audio.wav`, and `example_voice_reference.mp3` with files that exist in your ComfyUI input directory.
 - Replace placeholder URLs such as `https://example.com/reference-motion.mp4` with real URLs when needed.
 - Native preset and cloned voices are available for `kling-v2-6`; the active API rejects `voice_list` for `kling-v3`.
-- Newer documented models such as `kling-video-o1` and `kling-v3-omni` are intentionally hidden from the visible dropdowns until live endpoint support is confirmed.
+- `kling-v3-omni` is visible in the video generation dropdowns for comparison with `kling-v3`; `kling-video-o1` remains hidden until live endpoint support is confirmed.
 
 ## Important workflow notes
 
 - `05_comfyui_kling_wrapper_text2video_v26_sound.json` is the native audio example and is built around `kling-v2-6`.
 - `06_comfyui_kling_wrapper_advanced_element_subject_to_image2video.json` requires 1 frontal portrait plus 1-3 additional photos of the same subject. Background or scene images do not count as advanced-element reference images.
 - If you need both subject binding and speech, generate the bound video first and then add speech with the lip-sync or audio nodes.
+- `kling-v3-omni` can combine `custom_voice_id` or `custom_voice_list` with `element_list`; older models still keep stricter local guards.
 - `Motion Control` accepts either a direct http(s) reference-video URL or uploaded/local video workflow inputs. For local media, use a video loader node that provides either `VIDEO` or `reference_video_frames` plus `reference_video_info`. Set `duration` to `auto` if you want it to match the uploaded motion reference length.
 - `Lip Sync` now calls `identify-face` and `advanced-lip-sync`; leave `face_id` empty to use the first detected face.
 - `Avatar` accepts one and only one of `AUDIO`, `audio_id`, `audio_file`, or `audio_url`.
-- `Image2Video.custom_voice_id` is supported by `kling-v2-6`; it requires `sound=on` and the `<<<voice_1>>>` marker in the prompt.
+- `Image2Video.custom_voice_id` is the single-custom-voice path. For multiple voices, connect `Custom Voice List Builder.custom_voice_list` to `Image2Video.custom_voice_list`, set `sound=on`, and use ordered markers such as `<<<voice_1>>>` and `<<<voice_2>>>`.
 
 ## Included files
 

py/__init__.py

@@ -1,5 +1,5 @@
 from .api.capabilities import NODE_PREFIX
-from .nodes import AdvancedCustomElementCreateNode, AdvancedCustomElementDeleteNode, AdvancedCustomElementQueryNode, AvatarNode, EffectNode, ElementListBuilderNode, \
+from .nodes import AdvancedCustomElementCreateNode, AdvancedCustomElementDeleteNode, AdvancedCustomElementQueryNode, AvatarNode, CustomVoiceListBuilderNode, EffectNode, ElementListBuilderNode, \
     Image2VideoNode, ImageExpanderNode, ImageGeneratorNode, KolorsVirtualTryOnNode, LipSyncAudioInputNode, \
     LipSyncNode, LipSyncTextInputNode, MotionControlNode, MultiImagesToVideoNode, MultiModalVideoEditNode, PreviewAudio, PreviewVideo, Text2VideoNode, \
     TextToAudioNode, Video2AudioNode, VideoExtendNode, CustomVoiceCreateNode, CustomVoiceQueryNode, TTSNode
@@ -16,6 +16,7 @@ NODE_CLASS_MAPPINGS = {
     _node_name("TTS"): TTSNode,
     _node_name("Custom Voice Create"): CustomVoiceCreateNode,
     _node_name("Custom Voice Query"): CustomVoiceQueryNode,
+    _node_name("Custom Voice List Builder"): CustomVoiceListBuilderNode,
     _node_name("Avatar"): AvatarNode,
     _node_name("Video2Audio"): Video2AudioNode,
     _node_name("Text2Video"): Text2VideoNode,

py/api/capabilities.py

@@ -5,6 +5,7 @@ CLIENT_TYPE = "COMFYUI_KLING_WRAPPER_API_CLIENT"
 LIPSYNC_INPUT_TYPE = "COMFYUI_KLING_WRAPPER_LIPSYNC_INPUT"
 ELEMENT_TYPE = "COMFYUI_KLING_WRAPPER_ELEMENT"
 ELEMENT_LIST_TYPE = "COMFYUI_KLING_WRAPPER_ELEMENT_LIST"
+CUSTOM_VOICE_LIST_TYPE = "COMFYUI_KLING_WRAPPER_CUSTOM_VOICE_LIST"
 
 DEFAULT_VIDEO_ASPECT_RATIOS = ["16:9", "9:16", "1:1"]
 EXTENDED_IMAGE_ASPECT_RATIOS = ["16:9", "9:16", "1:1", "4:3", "3:4", "3:2", "2:3", "21:9", "auto"]
@@ -248,8 +249,8 @@ VIDEO_MODEL_CAPABILITIES = {
         "supports_image_list": True,
         "supports_element_list": True,
         "supports_reference_video": True,
-        "supports_sound": False,
-        "supports_voice_list": False,
+        "supports_sound": True,
+        "supports_voice_list": True,
         "supports_negative_prompt": False,
         "supports_cfg_scale": False,
         "shot_types": ["single", "intelligence"],
@@ -258,9 +259,9 @@ VIDEO_MODEL_CAPABILITIES = {
 }
 
 # ComfyUI input options are static, so we keep the visible dropdowns aligned
-# with models verified on the active endpoint. Newer documented models may
-# still exist in VIDEO_MODEL_CAPABILITIES for future use, but remain hidden
-# until live account support is confirmed.
+# with models confirmed for the active integration paths. Some documented
+# models may still exist in VIDEO_MODEL_CAPABILITIES for future use, but remain
+# hidden until endpoint support is confirmed.
 IMAGE_GENERATION_MODELS = [
     "kling-v1",
     "kling-v1-5",
@@ -278,6 +279,7 @@ TEXT_TO_VIDEO_MODELS = [
     "kling-v2-5-turbo",
     "kling-v2-6",
     "kling-v3",
+    "kling-v3-omni",
 ]
 
 IMAGE_TO_VIDEO_MODELS = [
@@ -290,6 +292,7 @@ IMAGE_TO_VIDEO_MODELS = [
     "kling-v2-5-turbo",
     "kling-v2-6",
     "kling-v3",
+    "kling-v3-omni",
 ]
 
 # Multi-image generation is routed through image2video with a primary `image`
@@ -298,6 +301,7 @@ MULTI_IMAGE_TO_VIDEO_MODELS = [
     "kling-v2-1",
     "kling-v2-5-turbo",
     "kling-v3",
+    "kling-v3-omni",
 ]
 
 

py/nodes.py

@@ -2,11 +2,12 @@ from .api import AdvancedLipSync, Avatar, Client, FaceIdentify, ImageGenerator,
     Video2Audio, Text2Audio, Text2Video, CameraControl, CameraControlConfig, KolorsVurtualTryOn, VideoExtend, \
     LipSyncInput, EffectInput, Effects, MultiImages2Video, MultiModelVideoEdit, AdvancedCustomElements, \
     MotionControl, CustomVoiceCreate, CustomVoiceQuery, TTS
-from .api.capabilities import DEFAULT_IMAGE_RESOLUTIONS, DEFAULT_MODES, DEFAULT_VIDEO_ASPECT_RATIOS, \
-    DEFAULT_VIDEO_DURATIONS, ELEMENT_LIST_TYPE, ELEMENT_TYPE, EXTENDED_IMAGE_ASPECT_RATIOS, IMAGE_GENERATION_MODELS, \
-    IMAGE_TO_VIDEO_MODELS, LIPSYNC_INPUT_TYPE, MULTI_IMAGE_TO_VIDEO_MODELS, NODE_CATEGORY, NODE_PREFIX, SHOT_TYPES, \
-    SOUND_OPTIONS, TEXT_TO_VIDEO_MODELS, VIDEO_GENERATION_MODES, get_image_capability, get_video_capability, \
-    validate_image_generation_request, validate_video_generation_request
+from .api.capabilities import CUSTOM_VOICE_LIST_TYPE, DEFAULT_IMAGE_RESOLUTIONS, DEFAULT_MODES, \
+    DEFAULT_VIDEO_ASPECT_RATIOS, DEFAULT_VIDEO_DURATIONS, ELEMENT_LIST_TYPE, ELEMENT_TYPE, \
+    EXTENDED_IMAGE_ASPECT_RATIOS, IMAGE_GENERATION_MODELS, IMAGE_TO_VIDEO_MODELS, LIPSYNC_INPUT_TYPE, \
+    MULTI_IMAGE_TO_VIDEO_MODELS, NODE_CATEGORY, NODE_PREFIX, SHOT_TYPES, SOUND_OPTIONS, TEXT_TO_VIDEO_MODELS, \
+    VIDEO_GENERATION_MODES, get_image_capability, get_video_capability, validate_image_generation_request, \
+    validate_video_generation_request
 from .api.exceptions import KLingAPIError
 from .api.providers import (
     OFFICIAL_PROVIDER,
@@ -616,6 +617,61 @@ def _build_voice_list_from_preset(voice_preset):
     return [{"voice_id": voice_id}]
 
 
+def _normalize_voice_reference(voice):
+    if voice is None:
+        return None
+    if isinstance(voice, dict):
+        voice_id = str(voice.get("voice_id") or voice.get("id") or "").strip()
+        if not voice_id:
+            raise ValueError("Voice payload is missing voice_id.")
+        return {"voice_id": voice_id}
+    if isinstance(voice, str):
+        voice_id = voice.strip()
+        if not voice_id:
+            return None
+        return {"voice_id": voice_id}
+    raise ValueError("Voice must be a dict or string.")
+
+
+def _normalize_voice_list(voices):
+    if voices is None:
+        return None
+    if isinstance(voices, str):
+        cleaned = voices.strip()
+        if not cleaned:
+            return None
+        if cleaned[0] in "[{":
+            decoded = _parse_json_input(cleaned, "custom_voice_list")
+            voices = decoded
+        else:
+            voices = [line.strip() for line in cleaned.replace(",", "\n").splitlines()]
+    if isinstance(voices, dict):
+        voices = [voices]
+    normalized = []
+    for voice in voices:
+        payload = _normalize_voice_reference(voice)
+        if payload:
+            normalized.append(payload)
+    return normalized or None
+
+
+def _validate_voice_prompt_markers(prompt, voice_list):
+    if not voice_list:
+        return
+    prompt_text = str(prompt or "")
+    missing_markers = [
+        f"<<<voice_{index}>>>"
+        for index in range(1, len(voice_list) + 1)
+        if f"<<<voice_{index}>>>" not in prompt_text
+    ]
+    if missing_markers:
+        raise ValueError(
+            "custom_voice_list requires prompt markers: "
+            + ", ".join(missing_markers)
+            + "."
+        )
+
+
 def _apply_prompt_controls(target, capability, prompt=None, negative_prompt=None, cfg_scale=None):
     cleaned_prompt = _clean_optional_text(prompt)
     cleaned_negative_prompt = _clean_optional_text(negative_prompt)
@@ -1960,6 +2016,7 @@ class Image2VideoNode:
                 "sound": (SOUND_OPTIONS,),
                 "voice_preset": (VOICE_PRESET_OPTIONS, {"default": DEFAULT_VOICE_PRESET}),
                 "custom_voice_id": ("STRING", {"forceInput": True}),
+                "custom_voice_list": (CUSTOM_VOICE_LIST_TYPE,),
                 "shot_type": (SHOT_TYPES,),
                 "image_list": ("IMAGE",),
                 "element_list": (ELEMENT_LIST_TYPE,),
@@ -1992,6 +2049,7 @@ class Image2VideoNode:
                  sound="off",
                  voice_preset=DEFAULT_VOICE_PRESET,
                  custom_voice_id="",
+                 custom_voice_list=None,
                  shot_type="single",
                  image_list=None,
                  element_list=None,
@@ -2001,16 +2059,28 @@ class Image2VideoNode:
         mode = _normalize_requested_mode(model, capability, mode)
         normalized_element_list = _normalize_element_list(element_list)
         preset_voice_list = _build_voice_list_from_preset(voice_preset)
+        normalized_custom_voice_list = _normalize_voice_list(custom_voice_list)
         custom_voice_id = str(custom_voice_id or "").strip()
         if custom_voice_id and preset_voice_list:
             raise ValueError("Use custom_voice_id or voice_preset, not both.")
+        if normalized_custom_voice_list and (custom_voice_id or preset_voice_list):
+            raise ValueError("Use custom_voice_list, custom_voice_id, or voice_preset, not more than one.")
         if custom_voice_id and sound != "on":
             raise ValueError("custom_voice_id requires sound='on'.")
+        if normalized_custom_voice_list and sound != "on":
+            raise ValueError("custom_voice_list requires sound='on'.")
         if custom_voice_id and "<<<voice_1>>>" not in str(prompt or ""):
             raise ValueError("custom_voice_id requires the prompt marker <<<voice_1>>>.")
-        if custom_voice_id and normalized_element_list:
+        _validate_voice_prompt_markers(prompt, normalized_custom_voice_list)
+        if custom_voice_id and normalized_element_list and model != "kling-v3-omni":
             raise ValueError("custom_voice_id cannot be used together with element_list.")
-        parsed_voice_list = [{"voice_id": custom_voice_id}] if custom_voice_id else preset_voice_list
+        if normalized_custom_voice_list and normalized_element_list and model != "kling-v3-omni":
+            raise ValueError("custom_voice_list cannot be used together with element_list.")
+        parsed_voice_list = (
+            normalized_custom_voice_list
+            or ([{"voice_id": custom_voice_id}] if custom_voice_id else None)
+            or preset_voice_list
+        )
         encoded_image_list = _image_batch_to_base64_list(image_list)
         has_reference_video = isinstance(reference_video, str) and bool(reference_video.strip())
         has_camera_control = camera_control_type not in (None, "None")
@@ -3439,6 +3509,48 @@ class ElementListBuilderNode:
         )
 
 
+class CustomVoiceListBuilderNode:
+    @classmethod
+    def INPUT_TYPES(cls):
+        return {
+            "required": {
+                "voice_1": ("STRING", {"forceInput": True}),
+            },
+            "optional": {
+                "voice_2": ("STRING", {"forceInput": True}),
+                "voice_3": ("STRING", {"forceInput": True}),
+                "voice_4": ("STRING", {"forceInput": True}),
+                "extra_voice_ids": ("STRING", {"multiline": True, "default": ""}),
+            }
+        }
+
+    RETURN_TYPES = (CUSTOM_VOICE_LIST_TYPE, "STRING")
+    RETURN_NAMES = ("custom_voice_list", "custom_voice_list_json")
+    FUNCTION = "build"
+    OUTPUT_NODE = False
+    CATEGORY = NODE_CATEGORY
+
+    def build(self, voice_1, voice_2=None, voice_3=None, voice_4=None, extra_voice_ids=""):
+        voice_refs = []
+        for voice_id in (voice_1, voice_2, voice_3, voice_4):
+            if voice_id is not None:
+                voice_refs.append(voice_id)
+
+        for raw_line in (extra_voice_ids or "").replace(",", "\n").splitlines():
+            voice_id = raw_line.strip()
+            if voice_id:
+                voice_refs.append(voice_id)
+
+        normalized = _normalize_voice_list(voice_refs)
+        if not normalized:
+            raise ValueError("At least one voice_id is required.")
+
+        return (
+            normalized,
+            json.dumps(normalized, ensure_ascii=False),
+        )
+
+
 class MotionControlNode:
     @classmethod
     def INPUT_TYPES(cls):

tests/test_v3_omni_visibility.py

@@ -0,0 +1,193 @@
+import importlib
+import sys
+import types
+import unittest
+from pathlib import Path
+from types import SimpleNamespace
+from unittest import mock
+
+
+REPO_ROOT = Path(__file__).resolve().parents[1]
+COMFYUI_ROOT = REPO_ROOT.parent.parent
+if str(COMFYUI_ROOT) not in sys.path:
+    sys.path.insert(0, str(COMFYUI_ROOT))
+if str(REPO_ROOT) not in sys.path:
+    sys.path.insert(0, str(REPO_ROOT))
+
+
+def _install_import_stubs():
+    if "torch" not in sys.modules:
+        sys.modules["torch"] = types.SimpleNamespace()
+    if "comfy_extras.nodes_audio" not in sys.modules:
+        audio_module = types.ModuleType("comfy_extras.nodes_audio")
+        audio_module.LoadAudio = object
+        sys.modules["comfy_extras.nodes_audio"] = audio_module
+
+
+class V3OmniVisibilityTests(unittest.TestCase):
+    @classmethod
+    def setUpClass(cls):
+        _install_import_stubs()
+        cls.capabilities = importlib.import_module("py.api.capabilities")
+        cls.nodes = importlib.import_module("py.nodes")
+        cls.tencent_video = importlib.import_module("py.api.providers.tencent_vod.video")
+
+    def test_v3_omni_is_visible_for_video_comparison_nodes(self):
+        for model_list_name in (
+                "TEXT_TO_VIDEO_MODELS",
+                "IMAGE_TO_VIDEO_MODELS",
+                "MULTI_IMAGE_TO_VIDEO_MODELS",
+        ):
+            with self.subTest(model_list_name=model_list_name):
+                self.assertIn(
+                    "kling-v3-omni",
+                    getattr(self.capabilities, model_list_name),
+                )
+
+    def test_tencent_vod_v3_omni_model_version_matches_docs(self):
+        self.assertEqual(
+            self.tencent_video.TencentVodVideoProvider._resolve_model_version("kling-v3-omni"),
+            "3.0-Omni",
+        )
+
+    def test_v3_omni_reference_video_capability_stays_guarded(self):
+        capability = self.capabilities.validate_video_generation_request(
+            task_name="image2video",
+            model_name="kling-v3-omni",
+            mode="pro",
+            duration="5",
+            has_reference_video=True,
+        )
+
+        self.assertTrue(capability["supports_reference_video"])
+        self.assertTrue(capability["supports_sound"])
+        self.assertTrue(capability["supports_voice_list"])
+
+    def test_v3_omni_allows_custom_voice_with_element_list(self):
+        captured_requests = []
+
+        class FakeProvider:
+            def generate_image2video(self, request):
+                captured_requests.append(request)
+                return SimpleNamespace(
+                    url="https://example.com/video.mp4",
+                    video_id="video-omni",
+                    final_unit_deduction=None,
+                )
+
+            def close(self):
+                pass
+
+        with mock.patch.object(self.nodes, "_create_video_provider", return_value=FakeProvider()):
+            url, video_id = self.nodes.Image2VideoNode().generate(
+                model="kling-v3-omni",
+                provider="official",
+                prompt='The subject <<<voice_1>>> says: "Hello."',
+                mode="pro",
+                duration="5",
+                sound="on",
+                custom_voice_id="voice-custom",
+                element_list=[{"element_id": "element-123"}],
+            )
+
+        self.assertEqual(url, "https://example.com/video.mp4")
+        self.assertEqual(video_id, "video-omni")
+        self.assertEqual(captured_requests[0].voice_list, [{"voice_id": "voice-custom"}])
+        self.assertEqual(captured_requests[0].element_list, [{"element_id": "element-123"}])
+
+    def test_custom_voice_list_builder_collects_multiple_voice_ids(self):
+        voice_list, voice_list_json = self.nodes.CustomVoiceListBuilderNode().build(
+            voice_1="voice-a",
+            voice_2="voice-b",
+            extra_voice_ids="voice-c\nvoice-d",
+        )
+
+        self.assertEqual(
+            voice_list,
+            [
+                {"voice_id": "voice-a"},
+                {"voice_id": "voice-b"},
+                {"voice_id": "voice-c"},
+                {"voice_id": "voice-d"},
+            ],
+        )
+        self.assertIn("voice-d", voice_list_json)
+
+    def test_custom_voice_list_builder_is_registered(self):
+        mappings = importlib.import_module("py").NODE_CLASS_MAPPINGS
+        self.assertIn("Comfyui-Kling-Wrapper Custom Voice List Builder", mappings)
+
+    def test_v3_omni_custom_voice_list_sends_multiple_voices(self):
+        captured_requests = []
+
+        class FakeProvider:
+            def generate_image2video(self, request):
+                captured_requests.append(request)
+                return SimpleNamespace(
+                    url="https://example.com/video.mp4",
+                    video_id="video-omni",
+                    final_unit_deduction=None,
+                )
+
+            def close(self):
+                pass
+
+        custom_voice_list = [
+            {"voice_id": "voice-a"},
+            {"voice_id": "voice-b"},
+        ]
+        with mock.patch.object(self.nodes, "_create_video_provider", return_value=FakeProvider()):
+            self.nodes.Image2VideoNode().generate(
+                model="kling-v3-omni",
+                provider="official",
+                prompt='<<<voice_1>>> says hello. <<<voice_2>>> replies.',
+                mode="pro",
+                duration="5",
+                sound="on",
+                custom_voice_list=custom_voice_list,
+                element_list=[{"element_id": "element-123"}],
+            )
+
+        self.assertEqual(captured_requests[0].voice_list, custom_voice_list)
+        self.assertEqual(captured_requests[0].element_list, [{"element_id": "element-123"}])
+
+    def test_custom_voice_list_requires_matching_prompt_markers(self):
+        with self.assertRaisesRegex(ValueError, "<<<voice_2>>>"):
+            self.nodes.Image2VideoNode().generate(
+                model="kling-v3-omni",
+                provider="official",
+                prompt="<<<voice_1>>> says hello.",
+                mode="pro",
+                duration="5",
+                sound="on",
+                custom_voice_list=[{"voice_id": "voice-a"}, {"voice_id": "voice-b"}],
+            )
+
+    def test_older_custom_voice_models_still_reject_element_list(self):
+        with self.assertRaisesRegex(ValueError, "cannot be used together with element_list"):
+            self.nodes.Image2VideoNode().generate(
+                model="kling-v2-6",
+                provider="official",
+                prompt='The subject <<<voice_1>>> says: "Hello."',
+                mode="pro",
+                duration="5",
+                sound="on",
+                custom_voice_id="voice-custom",
+                element_list=[{"element_id": "element-123"}],
+            )
+
+        with self.assertRaisesRegex(ValueError, "custom_voice_list cannot be used together with element_list"):
+            self.nodes.Image2VideoNode().generate(
+                model="kling-v2-6",
+                provider="official",
+                prompt='<<<voice_1>>> says hello. <<<voice_2>>> replies.',
+                mode="pro",
+                duration="5",
+                sound="on",
+                custom_voice_list=[{"voice_id": "voice-a"}, {"voice_id": "voice-b"}],
+                element_list=[{"element_id": "element-123"}],
+            )
+
+
+if __name__ == "__main__":
+    unittest.main()