diff --git a/README.md b/README.md
index 7521b85..e3e89a0 100644
--- a/README.md
+++ b/README.md
@@ -43,4 +43,11 @@ Progress auto-saves on every completed set and on Ctrl-C. Re-run the same workou
 
 ## Voice
 
-Uses macOS `say` for coaching cues. Silently skipped on other platforms.
+Uses macOS `say` or Linux `espeak`/`spd-say` for coaching cues. Silently skipped if no TTS is available.
+
+On Linux, install a TTS engine:
+
+```
+sudo apt install espeak      # Debian/Ubuntu
+sudo dnf install espeak-ng   # Fedora
+```
diff --git a/coach.py b/coach.py
index 21e652d..46cf959 100755
--- a/coach.py
+++ b/coach.py
@@ -144,34 +144,70 @@ def exercises_match(a: list[Exercise], b: list[Exercise]) -> bool:
 
 
 # ---------------------------------------------------------------------------
-# Voice coaching (non-blocking macOS `say`)
+# Voice coaching (macOS `say`, Linux `espeak`/`spd-say`)
 # ---------------------------------------------------------------------------
 
 _say_proc: subprocess.Popen | None = None
 
 
+def _tts_cmd(text: str) -> list[str]:
+    """Return the TTS command + args for the current platform."""
+    if sys.platform == "darwin":
+        return ["say", text]
+    # Linux: prefer espeak, fall back to spd-say
+    for cmd in ("espeak", "spd-say"):
+        if _which(cmd):
+            return [cmd, text]
+    return []
+
+
+def _which(cmd: str) -> bool:
+    """Check whether *cmd* is on PATH (cached)."""
+    if cmd not in _which_cache:
+        _which_cache[cmd] = (
+            subprocess.call(
+                ["which", cmd],
+                stdout=subprocess.DEVNULL,
+                stderr=subprocess.DEVNULL,
+            )
+            == 0
+        )
+    return _which_cache[cmd]
+
+
+_which_cache: dict[str, bool] = {}
+
+
 def say(text: str) -> None:
     global _say_proc
+    cmd = _tts_cmd(text)
+    if not cmd:
+        return
     try:
         if _say_proc and _say_proc.poll() is None:
             _say_proc.terminate()
         _say_proc = subprocess.Popen(
-            ["say", text],
+            cmd,
             stdout=subprocess.DEVNULL,
             stderr=subprocess.DEVNULL,
         )
     except FileNotFoundError:
-        pass  # not on macOS
+        pass
 
 
 def say_sync(text: str, wait: float = 0) -> None:
     """Say something and optionally wait after it finishes."""
     global _say_proc
+    cmd = _tts_cmd(text)
+    if not cmd:
+        if wait > 0:
+            time.sleep(wait)
+        return
     try:
         if _say_proc and _say_proc.poll() is None:
             _say_proc.terminate()
         _say_proc = subprocess.Popen(
-            ["say", text],
+            cmd,
             stdout=subprocess.DEVNULL,
             stderr=subprocess.DEVNULL,
         )
diff --git a/plank.sh b/plank.sh
index 2566b4e..ab62036 100644
--- a/plank.sh
+++ b/plank.sh
@@ -1,13 +1,25 @@
 #!/bin/bash
+
+# Cross-platform TTS: macOS `say`, Linux `espeak` or `spd-say`
+speak() {
+  if command -v say &>/dev/null; then
+    say "$1"
+  elif command -v espeak &>/dev/null; then
+    espeak "$1"
+  elif command -v spd-say &>/dev/null; then
+    spd-say "$1"
+  fi
+}
+
 for set in 1 2 3; do
-  say "Set $set. Get in position."
+  speak "Set $set. Get in position."
   sleep 3
-  say "Go"
+  speak "Go"
   sleep 40
-  say "Done. Set $set complete."
+  speak "Done. Set $set complete."
   if [ $set -lt 3 ]; then
-    say "Rest 60 seconds"
+    speak "Rest 60 seconds"
     sleep 60
   fi
 done
-say "RKC Planks done. 3 sets of 40 seconds."
+speak "RKC Planks done. 3 sets of 40 seconds."