diff --git a/src/frontends/pytorch/src/transforms/fftn_complex_replacer.cpp b/src/frontends/pytorch/src/transforms/fftn_complex_replacer.cpp
index d5fc5ce8f0f662..1ba4989e7a2d9a 100644
--- a/src/frontends/pytorch/src/transforms/fftn_complex_replacer.cpp
+++ b/src/frontends/pytorch/src/transforms/fftn_complex_replacer.cpp
@@ -11,12 +11,13 @@
 #include "openvino/op/equal.hpp"
 #include "openvino/op/gather.hpp"
 #include "openvino/op/range.hpp"
-#include "openvino/op/rdft.hpp"
+#include "openvino/op/dft.hpp"
 #include "openvino/op/reduce_prod.hpp"
 #include "openvino/op/reshape.hpp"
 #include "openvino/op/select.hpp"
 #include "openvino/op/shape_of.hpp"
 #include "openvino/op/slice.hpp"
+#include "openvino/op/split.hpp"
 #include "openvino/op/sqrt.hpp"
 #include "openvino/op/squeeze.hpp"
 #include "openvino/op/subtract.hpp"
@@ -102,22 +103,22 @@ FFTNComplexReplacer::FFTNComplexReplacer() {
             return false;
         }
 
-        auto rdft = std::make_shared<v9::RDFT>(input, dim, s);
+        auto dft = std::make_shared<v7::DFT>(input, dim, s);
 
         // Apply normalizations
         auto n_int = std::make_shared<v1::ReduceProd>(s, const_0);
-        auto n = std::make_shared<v1::ConvertLike>(n_int, rdft);
+        auto n = std::make_shared<v1::ConvertLike>(n_int, dft);
         std::shared_ptr<ov::Node> normalized_fftn;
         if (norm == "forward") {
             // Normalize by 1/n
-            normalized_fftn = std::make_shared<v1::Divide>(rdft, n);
+            normalized_fftn = std::make_shared<v1::Divide>(dft, n);
         } else if (norm == "backward") {
             // No normalization
-            normalized_fftn = rdft;
+            normalized_fftn = dft;
         } else if (norm == "ortho") {
             // Normalize by 1/sqrt(n)
             auto sqrt_n = std::make_shared<v0::Sqrt>(n);
-            normalized_fftn = std::make_shared<v1::Divide>(rdft, sqrt_n);
+            normalized_fftn = std::make_shared<v1::Divide>(dft, sqrt_n);
         } else {
             add_exception_to_fw_node(
                 fftn_op,
@@ -127,21 +128,12 @@ FFTNComplexReplacer::FFTNComplexReplacer() {
 
         // Replace outputs that are either torch operators aten::real or aten::imag. Apply squeeze to remove last
         // dimension used to concatenate.
+        auto normalized_rfftn_splitted = std::make_shared<v1::Split>(normalized_fftn, const_neg_1, 2);
         auto fftn_outs = fftn_op->get_users();
         bool rval = false;
         for (auto& out : fftn_outs) {
             if (auto real_op = cast_fw_node(out, "aten::real")) {
-                // Check if the last dimension of the tensor is 1
-                auto last_dim = normalized_fftn->output(0).get_shape().back();
-                if (last_dim != 1) {
-                    // Reshape the tensor to make the last dimension 1
-                    auto new_shape = normalized_fftn->output(0).get_shape();
-                    new_shape.back() = 1;
-                    auto new_shape_tensor = std::make_shared<v0::Constant>(element::i64, Shape{new_shape.size()}, new_shape);
-                    auto reshaped = std::make_shared<v1::Reshape>(normalized_fftn->output(0), new_shape_tensor, false);
-                    normalized_fftn = reshaped;
-                }
-                auto squeezed = std::make_shared<v0::Squeeze>(normalized_fftn->output(0), const_neg_1);
+                auto squeezed = std::make_shared<v0::Squeeze>(normalized_rfftn_splitted->output(0), const_neg_1);
                 copy_runtime_info({fftn_op, real_op}, squeezed);
                 squeezed->set_friendly_name(real_op->get_friendly_name());
                 replace_node(real_op, squeezed);