From ef66d759515fcb7306b808961997ddec6f78871a Mon Sep 17 00:00:00 2001
From: mateo-berri <277851410+mateo-berri@users.noreply.github.com>
Date: Mon, 9 Nov 2026 20:31:54 -0800
Subject: [PATCH] fix(speech): call Gemini once per TTS request

aspeech re-ran speech whenever the first result was not a coroutine, so the
speech-to-completion bridge sent every async Gemini and Vertex AI TTS request
upstream twice. It now returns the sync result

Prerequisite for the backport of #50803 to lts/1.108.x, whose tests count one
upstream request per speech call.
Hand-ported from f4347d4f50 (#44551, main), the aspeech hunk only.
---
 litellm/main.py | 8 +-------
 1 file changed, 1 insertion(+), 7 deletions(-)

diff --git a/litellm/main.py b/litellm/main.py
index 035e0628d8c..0b35892824b 100644
--- a/litellm/main.py
+++ b/litellm/main.py
@@ -8430,14 +8430,8 @@ async def aspeech(*args, **kwargs) -> HttpxBinaryResponseContent:
             api_base=api_base_hint,
         )
 
-        # Await normally
         init_response: Final = await loop.run_in_executor(None, func_with_context)
-        if asyncio.iscoroutine(init_response):
-            response = await init_response
-        else:
-            # Call the synchronous function using run_in_executor
-            response = await loop.run_in_executor(None, func_with_context)
-        return response
+        return await init_response if asyncio.iscoroutine(init_response) else init_response
     except Exception as e:
         custom_llm_provider = custom_llm_provider or "openai"
         raise exception_type(