diff --git a/README.md b/README.md
new file mode 100644
index 0000000..0ce9ada
--- /dev/null
+++ b/README.md
@@ -0,0 +1,21 @@
+# Kokoro Japanese TTS
+
+Small CLI wrapper around the Kokoro TTS pipeline for Japanese speech synthesis. It chunks long text at punctuation so the model keeps a natural pace and writes 24 kHz audio with `soundfile`.
+
+## Quick start
+- Requires Python 3.11+ with `pip`.
+- Run `./run.sh "こんにちは、ココロです。"` to create a `.venv`, install deps, download UniDic on first run, and generate `out.wav`.
+
+## Direct usage
+```bash
+source .venv/bin/activate              # or your own environment
+python kokoro_ja.py "テストです。" \
+  --voice jf_alpha \                   # e.g. jf_alpha, jf_tebukuro, jm_kumo
+  --speed 1.0 \                        # >1 faster, <1 slower
+  --out test_ja.wav
+```
+
+## Notes
+- Output is mono 24 kHz WAV.
+- UniDic is only downloaded if not already present (handled inside `run.sh`).
+- `PYTORCH_ENABLE_MPS_FALLBACK=1` is set in `run.sh` for smoother macOS GPU fallback; adjust as needed on other platforms.