diff --git a/tests/librispeech/Makefile b/tests/librispeech/Makefile
index 79fdd07a..0afa2465 100644
--- a/tests/librispeech/Makefile
+++ b/tests/librispeech/Makefile
@@ -8,13 +8,6 @@ eval:
 clean:
 	$(MAKE) -f eval.mk clean
 
-setup-venv:
-	python3 -m venv venv
-	./venv/bin/pip install -r requirements.txt
-
-clean-venv:
-	rm -r venv
-
 get-audio:
 	wget -c $(TAR_URL)
 	tar -xf test-clean.tar.gz
diff --git a/tests/librispeech/README.md b/tests/librispeech/README.md
index 026605ce..85478a0f 100644
--- a/tests/librispeech/README.md
+++ b/tests/librispeech/README.md
@@ -29,7 +29,15 @@ performance of whisper.cpp on LibriSpeech corpus.
 3. Set up the environment to compute WER score.
 
    ```
-   $ make setup-venv
+   $ pip install -r requirements.txt
+   ```
+
+   For example, if you use `virtualenv`, you can set up it as follows:
+
+   ```
+   $ python3 -m venv venv
+   $ . venv/bin/activate
+   $ pip install -r requirements.txt
    ```
 
 4. Run the benchmark test.
diff --git a/tests/librispeech/eval.mk b/tests/librispeech/eval.mk
index 67dd33d4..6d504c8b 100644
--- a/tests/librispeech/eval.mk
+++ b/tests/librispeech/eval.mk
@@ -1,4 +1,4 @@
-PYTHON = ./venv/bin/python3
+PYTHON = python
 
 WHISPER_PREFIX = ../../
 WHISPER_MODEL = tiny