diff --git a/voicevox-remotion-template/src/avatar-animations.ts b/voicevox-remotion-template/src/avatar-animations.ts
new file mode 100644
index 0000000..3fcadf5
--- /dev/null
+++ b/voicevox-remotion-template/src/avatar-animations.ts
@@ -0,0 +1,33 @@
+export type AvatarAnimationContext = Readonly<{
+  frame: number;
+  fps: number;
+  focused: boolean;
+  hasMultipleCharacters: boolean;
+}>;
+
+export type AvatarAnimation = (context: AvatarAnimationContext) => number;
+
+const gentleBobAmplitude = ({focused}: AvatarAnimationContext) =>
+  focused ? 10 : 3.5;
+
+export const idleAvatarAnimations = {
+  none: () => 0,
+  gentleBob: (context) =>
+    Math.sin((context.frame / context.fps) * Math.PI * 2) *
+    gentleBobAmplitude(context),
+} satisfies Record<string, AvatarAnimation>;
+
+export type IdleAvatarAnimationType = keyof typeof idleAvatarAnimations;
+
+export const speakingAvatarAnimations = {
+  none: () => 0,
+  gentleBob: idleAvatarAnimations.gentleBob,
+  quickHop: ({frame, fps}) => {
+    const cycleFrames = Math.max(1, Math.round(fps * 0.25));
+    const progress = (frame % cycleFrames) / cycleFrames;
+
+    return -Math.sin(progress * Math.PI) * 7;
+  },
+} satisfies Record<string, AvatarAnimation>;
+
+export type SpeakingAvatarAnimationType = keyof typeof speakingAvatarAnimations;
diff --git a/voicevox-remotion-template/src/data/script.ts b/voicevox-remotion-template/src/data/script.ts
index 82a2767..fdc83fd 100644
--- a/voicevox-remotion-template/src/data/script.ts
+++ b/voicevox-remotion-template/src/data/script.ts
@@ -1,3 +1,8 @@
+import type {
+  IdleAvatarAnimationType,
+  SpeakingAvatarAnimationType,
+} from "../avatar-animations";
+
 export type VoicevoxVoice = Readonly<{
   speakerName: string;
   styleName: string;
@@ -14,6 +19,8 @@
     flipX?: boolean;
   }>;
   nameplatePosition?: "top" | "bottom" | "none";
+  idleAnimationType?: IdleAvatarAnimationType;
+  speakingAnimationType?: SpeakingAvatarAnimationType;
 }>;
 
 export type CharacterDefinition = Readonly<{
@@ -40,6 +47,8 @@
         flipX: true,
       },
       nameplatePosition: "none",
+      idleAnimationType: "none",
+      speakingAnimationType: "quickHop",
     },
   },
   sayo: {
@@ -58,6 +67,8 @@
         translateY: -60,
       },
       nameplatePosition: "none",
+      idleAnimationType: "none",
+      speakingAnimationType: "quickHop",
     },
   },
 } as const satisfies Record<string, CharacterDefinition>;
diff --git a/voicevox-remotion-template/src/yukkuri-composition.tsx b/voicevox-remotion-template/src/yukkuri-composition.tsx
index 0e6c7d3..47b2ea7 100644
--- a/voicevox-remotion-template/src/yukkuri-composition.tsx
+++ b/voicevox-remotion-template/src/yukkuri-composition.tsx
@@ -19,6 +19,10 @@
   type TimelineEvent,
 } from "./data/script";
 import {
+  idleAvatarAnimations,
+  speakingAvatarAnimations,
+} from "./avatar-animations";
+import {
   audioFileForSpeech,
   GAP_FRAMES,
   durationForTimelineEvent,
@@ -381,14 +385,30 @@
     characterId: CharacterId;
     focused: boolean;
     hasMultipleCharacters: boolean;
-    bounce: number;
+    frame: number;
+    fps: number;
+    isSpeaking: boolean;
   }>
-> = ({characterId, focused, hasMultipleCharacters, bounce}) => {
+> = ({characterId, focused, hasMultipleCharacters, frame, fps, isSpeaking}) => {
   const character = characters[characterId];
   const {avatar}: {avatar: AvatarDefinition} = character;
   const scale = hasMultipleCharacters ? 0.88 : focused ? 1.05 : 1;
   const opacity = focused || !hasMultipleCharacters ? 1 : 0.72;
-  const translateY = focused ? bounce : bounce * 0.35;
+  const idleAnimationType = avatar.idleAnimationType ?? "gentleBob";
+  const speakingAnimationType = avatar.speakingAnimationType ?? "gentleBob";
+  const translateY = isSpeaking
+    ? speakingAvatarAnimations[speakingAnimationType]({
+        frame,
+        fps,
+        focused,
+        hasMultipleCharacters,
+      })
+    : idleAvatarAnimations[idleAnimationType]({
+        frame,
+        fps,
+        focused,
+        hasMultipleCharacters,
+      });
   const imageTranslateY = avatar.imageLayout?.translateY ?? 0;
   const imageScaleX = avatar.imageLayout?.flipX ? -1 : 1;
   const imageTransform = `translateY(${imageTranslateY}px) scaleX(${imageScaleX})`;
@@ -462,9 +482,11 @@
   Readonly<{
     visibleCharacters: CharacterId[];
     focusedCharacter?: CharacterId;
-    bounce: number;
+    speakingCharacter?: CharacterId;
+    frame: number;
+    fps: number;
   }>
-> = ({visibleCharacters, focusedCharacter, bounce}) => {
+> = ({visibleCharacters, focusedCharacter, speakingCharacter, frame, fps}) => {
   const hasMultipleCharacters = visibleCharacters.length > 1;
 
   return (
@@ -485,7 +507,9 @@
           characterId={characterId}
           focused={focusedCharacter === characterId}
           hasMultipleCharacters={hasMultipleCharacters}
-          bounce={bounce}
+          frame={frame}
+          fps={fps}
+          isSpeaking={speakingCharacter === characterId}
         />
       ))}
     </div>
@@ -559,12 +583,10 @@
     fps,
     config: {damping: 18, mass: 0.6},
   });
-
-  const bounce = interpolate(
-    Math.sin((frame / fps) * Math.PI * 2),
-    [-1, 1],
-    [-10, 10]
-  );
+  const speakingCharacter =
+    isInsideActiveSegment && activeSegment.event.type === "say"
+      ? activeSegment.event.character
+      : undefined;
 
   const sequences = scheduledEvents.map((scheduledEvent, index) => (
     <Sequence
@@ -602,7 +624,9 @@
         focusedCharacter={
           isInsideActiveSegment ? activeSegment.focusedCharacter : undefined
         }
-        bounce={bounce}
+        speakingCharacter={speakingCharacter}
+        frame={frame}
+        fps={fps}
       />
       {sequences}
     </AbsoluteFill>