diff --git a/ggml/src/ggml-cpu/ops.cpp b/ggml/src/ggml-cpu/ops.cpp
index 212e52ef6a1c8..c4824d145a54d 100644
--- a/ggml/src/ggml-cpu/ops.cpp
+++ b/ggml/src/ggml-cpu/ops.cpp
@@ -8599,7 +8599,6 @@ static void ggml_compute_forward_timestep_embedding_f32(
         }
         if (dim % 2 != 0 && ith == 0) {
             embed_data[2 * half] = 0.f;
-            embed_data[dim] = 0.f;
         }
     }
 }
diff --git a/ggml/src/ggml-cuda/tsembd.cu b/ggml/src/ggml-cuda/tsembd.cu
index 153ddbcda92dc..b91a26fc80e61 100644
--- a/ggml/src/ggml-cuda/tsembd.cu
+++ b/ggml/src/ggml-cuda/tsembd.cu
@@ -7,11 +7,11 @@ static __global__ void timestep_embedding_f32(const float * timesteps, float * d
     int j = threadIdx.x + blockIdx.x * blockDim.x;
     float * embed_data = (float *)((char *)dst +  i*nb1);
 
-    if (dim % 2 != 0 && j == ((dim + 1) / 2)) {
-        embed_data[dim] = 0.f;
+    int half = dim / 2;
+    if (dim % 2 != 0 && j == half) {
+        embed_data[2 * half] = 0.f;
     }
 
-    int half = dim / 2;
     if (j >= half) {
         return;
     }
diff --git a/ggml/src/ggml-metal/ggml-metal.metal b/ggml/src/ggml-metal/ggml-metal.metal
index 4314c9cc9388a..5057e264f6090 100644
--- a/ggml/src/ggml-metal/ggml-metal.metal
+++ b/ggml/src/ggml-metal/ggml-metal.metal
@@ -4167,7 +4167,7 @@ kernel void kernel_timestep_embedding_f32(
     }
 
     if (args.dim % 2 != 0 && tpitg.x == 0) {
-        embed_data[args.dim] = 0.f;
+        embed_data[2 * half_] = 0.f;
     }
 }
 
diff --git a/ggml/src/ggml-opencl/kernels/tsembd.cl b/ggml/src/ggml-opencl/kernels/tsembd.cl
index 4b1107f70ba7a..21444bd958298 100644
--- a/ggml/src/ggml-opencl/kernels/tsembd.cl
+++ b/ggml/src/ggml-opencl/kernels/tsembd.cl
@@ -26,8 +26,8 @@ kernel void kernel_timestep_embedding(
     local_half_dim = logical_dim / 2;
     local_embed_data_ptr = (global float *)((global char *)local_dst_output_base_ptr + local_i * dst_nb1_bytes);
 
-    if (logical_dim % 2 != 0 && local_j == ((logical_dim + 1) / 2)) {
-        local_embed_data_ptr[logical_dim] = 0.0f;
+    if (logical_dim % 2 != 0 && local_j == local_half_dim) {
+        local_embed_data_ptr[2 * local_half_dim] = 0.0f;
     }
 
     if (local_j >= local_half_dim) {
diff --git a/ggml/src/ggml-sycl/tsembd.cpp b/ggml/src/ggml-sycl/tsembd.cpp
index f6ca626ea7a53..f2003794d3f55 100644
--- a/ggml/src/ggml-sycl/tsembd.cpp
+++ b/ggml/src/ggml-sycl/tsembd.cpp
@@ -21,11 +21,12 @@ static void timestep_embedding_f32(
     int j = item_ct1.get_local_id(2) + item_ct1.get_group(2) * item_ct1.get_local_range(2);
     float * embed_data = (float *)((char *)dst +  i*nb1);
 
-    if (dim % 2 != 0 && j == ((dim + 1) / 2)) {
-        embed_data[dim] = 0.f;
+    int half = dim / 2;
+
+    if (dim % 2 != 0 && j == half) {
+        embed_data[2 * half] = 0.f;
     }
 
-    int half = dim / 2;
     if (j >= half) {
         return;
     }
diff --git a/ggml/src/ggml-vulkan/vulkan-shaders/timestep_embedding.comp b/ggml/src/ggml-vulkan/vulkan-shaders/timestep_embedding.comp
index 79e065a9313aa..ce8e09442d9b6 100644
--- a/ggml/src/ggml-vulkan/vulkan-shaders/timestep_embedding.comp
+++ b/ggml/src/ggml-vulkan/vulkan-shaders/timestep_embedding.comp
@@ -24,11 +24,12 @@ void main() {
     const uint j = gl_GlobalInvocationID.x;
     const uint d_offset = i * p.nb1;
 
-    if (p.dim % 2 != 0 && j == ((p.dim + 1) / 2)) {
-        data_d[d_offset + p.dim] = 0.f;
+    const uint half_dim = p.dim / 2;
+
+    if (p.dim % 2 != 0 && j == half_dim) {
+        data_d[d_offset + 2 * half_dim] = 0.f;
     }
 
-    const uint half_dim = p.dim / 2;
     if (j >= half_dim) {
         return;
     }
diff --git a/ggml/src/ggml.c b/ggml/src/ggml.c
index 50dc1aa24fff5..3584827dca7fc 100644
--- a/ggml/src/ggml.c
+++ b/ggml/src/ggml.c
@@ -4923,12 +4923,8 @@ struct ggml_tensor * ggml_timestep_embedding(
         struct ggml_tensor  * timesteps,
         int                   dim,
         int                   max_period) {
-    int actual_dim = dim;
-    if (dim % 2 != 0) {
-        actual_dim = dim + 1;
-    }
 
-    struct ggml_tensor * result = ggml_new_tensor_2d(ctx, GGML_TYPE_F32, actual_dim, timesteps->ne[0]);
+    struct ggml_tensor * result = ggml_new_tensor_2d(ctx, GGML_TYPE_F32, dim, timesteps->ne[0]);
 
     ggml_set_op_params_i32(result, 0, dim);
     ggml_set_op_params_i32(result, 1, max_period);