fix issue.

Signed-off-by: xipingya <xiping.yan@intel.com>

Author: xipingyan

samples/cpp/module_genai/config_yaml/Qwen3-Omni/config_prompt_tts_int4.yaml

@@ -0,0 +1,51 @@
+global_context:
+  model_type: "qwen3_omni"
+
+pipeline_modules:
+  pipeline_params:
+    type: "ParameterModule"
+    outputs:
+      - name: "prompts"
+        type: "String"
+
+  text_to_speech:
+    type: "TextToSpeechModule"
+    device: "CPU"
+    inputs:
+      - name: "text"
+        type: "String"
+        source: "pipeline_params.prompts"
+    outputs:
+      - name: "audios"
+        type: "VecOVTensor"
+      - name: "sample_rates"
+        type: "VecInt"
+      - name: "generated_texts"
+        type: "VecString"
+    params:
+      config_path: "./tests/module_genai/cpp/test_models/Qwen3-Omni-4B-Instruct-multilingual-int4/config.json"
+      tokenizer_path: "./tests/module_genai/cpp/test_models/Qwen3-Omni-4B-Instruct-multilingual-int4/"
+      embedding_model_path: "./tests/module_genai/cpp/test_models/Qwen3-Omni-4B-Instruct-multilingual-int4/qwen3_omni_talker_embedding_model.xml"
+      prefill_model_path: "./tests/module_genai/cpp/test_models/Qwen3-Omni-4B-Instruct-multilingual-int4/qwen3_omni_talker_prefill_model.xml"
+      decode_model_path: "./tests/module_genai/cpp/test_models/Qwen3-Omni-4B-Instruct-multilingual-int4/qwen3_omni_talker_decode_model.xml"
+      codec_embedding_model_path: "./tests/module_genai/cpp/test_models/Qwen3-Omni-4B-Instruct-multilingual-int4/qwen3_omni_talker_codec_embedding_model.xml"
+      code_predictor_ar_model_path: "./tests/module_genai/cpp/test_models/Qwen3-Omni-4B-Instruct-multilingual-int4"
+      code_predictor_single_codec_embed_model_path: "./tests/module_genai/cpp/test_models/Qwen3-Omni-4B-Instruct-multilingual-int4"
+      code_predictor_single_codec_embedding_model_path: "./tests/module_genai/cpp/test_models/Qwen3-Omni-4B-Instruct-multilingual-int4/qwen3_omni_code_predictor_codec_embed_model.xml"
+      speech_decoder_model_path: "./tests/module_genai/cpp/test_models/Qwen3-Omni-4B-Instruct-multilingual-int4/qwen3_omni_speech_decoder_model.xml"
+      sample_codec_token_greedy_search: true
+      merge_ar_and_sce_ov_models: true
+
+  pipeline_result:
+    type: "ResultModule"
+    description: "Collects final results and formats the output structure."
+    inputs:
+      - name: "audios"
+        type: "VecOVTensor"
+        source: "text_to_speech.audios"
+      - name: "sample_rates"
+        type: "VecInt"
+        source: "text_to_speech.sample_rates"
+      - name: "generated_texts"
+        type: "VecString"
+        source: "text_to_speech.generated_texts"

src/cpp/src/module_genai/modules/md_text_to_speech/models/qwen3_omni.cpp

@@ -203,7 +203,7 @@ void TextToSpeechImpl_Qwen3Omni::load_code_predictor_models(const ov::AnyMap& tt
     }
     OPENVINO_ASSERT(!m_code_predictor_single_codec_embed_infers.empty(), "No single-codec-embed step models found in " + sce_dir.string());
 
-    if (m_sample_codec_token_greedy_search) {
+    if (m_sample_codec_token_greedy_search && m_merge_ar_and_sce_ov_models) {
         GENAI_INFO("TextToSpeechModule[" + module_desc->name +
                    "]: sample_codec_token_greedy_search is enabled, will use greedy decoding in sample_codec_token");
         merge_code_predictor_ov_models(ar_models, sce_models);
@@ -361,7 +361,7 @@ std::shared_ptr<ov::Model> merge_neighbor_models(std::shared_ptr<ov::Model>& mod
     auto model_1_output_embeddings = model_1->get_results()[0]->input_value(0);
     auto model_1_output_layer_tokens = model_1->get_results()[1]->input_value(0);
 
-    // Append model_1's output Embeddings to model_1's input inputs_embeds, and then take it as model_2's inputs_embeds.
+    // Append model_1's output embeddings (all previous steps) to inputs_embeds for model_2.
     auto merged_inputs_embeds = std::make_shared<ov::op::v0::Concat>(
         ov::OutputVector{model_1_inputs_embeds->output(0), model_1_output_embeddings},
         1);
@@ -370,12 +370,13 @@ std::shared_ptr<ov::Model> merge_neighbor_models(std::shared_ptr<ov::Model>& mod
     model_2->inputs()[1].replace(model_1_output_layer_tokens);
 
     auto model_2_output_embeddings = model_2->get_results()[0]->input_value(0);
-    auto merged_2_outputs_embeddings =
-        std::make_shared<ov::op::v0::Concat>(ov::OutputVector{merged_inputs_embeds, model_2_output_embeddings}, 1);
-    
-    auto merged_2_outputs_embeddings_result = std::make_shared<ov::op::v0::Result>(merged_2_outputs_embeddings);
+    auto merged_output_embeddings = std::make_shared<ov::op::v0::Concat>(
+        ov::OutputVector{model_1_output_embeddings, model_2_output_embeddings},
+        1);
+
+    auto merged_output_embeddings_result = std::make_shared<ov::op::v0::Result>(merged_output_embeddings);
 
-    return std::make_shared<ov::Model>(ov::ResultVector{merged_2_outputs_embeddings_result, model_2->get_results()[1]},
+    return std::make_shared<ov::Model>(ov::ResultVector{merged_output_embeddings_result, model_2->get_results()[1]},
                                        ov::ParameterVector{model_1_inputs_embeds, model_1_input_current_layer_tokens},
                                        "merged_model");
 };