diff --git a/main.gd b/main.gd
index 0219278..e410cf1 100644
--- a/main.gd
+++ b/main.gd
@@ -19,7 +19,13 @@ var levels: Array[PackedScene] = [
 	preload("res://level/level_03.tscn"),
 	preload("res://level/level_04.tscn"),
 ]
-var time := 0
+# TODO: screen transition to hide awkward animations lol
+var time := 0:
+	set(new_time):
+		var tween := get_tree().create_tween()
+		var anim_time = 1 if slowmo() else 0.1
+		tween.tween_property($Sun, "rotation_degrees:y", -20*new_time, anim_time)
+		time = new_time
 var advancing := false
 
 @onready var sound: AudioStreamPlayer = $Sound
@@ -35,10 +41,9 @@ func _process(delta: float) -> void:
 	$RedoButton.disabled = not hist.has_redo()
 	$RestartButton.disabled = not hist.has_undo()
 	$Clock.text = "T = %d" % time
-	$Sun.rotation_degrees.y = -10*time
-	if $TopLeft.is_on_screen() and $BottomRight.is_on_screen():
-		return
-	camera.position.y += 10*delta
+	$SlowmoIndicator.text = "slowmo" if slowmo() else ""
+	if not $TopLeft.is_on_screen() or not $BottomRight.is_on_screen():
+		camera.position.y += 10*delta
 
 func _input(event: InputEvent) -> void:
 	if event.is_action_pressed("u", true, true):
@@ -72,6 +77,11 @@ func redo():
 func restart():
 	while hist.has_undo():
 		hist.undo()
+	# HACK: avoid an inconsistent visual board state
+	for piece in board.pieces():
+		for tween in piece.tweens:
+			tween.kill()
+		piece.tween_to_target()
 	# i think it's unintuitive to be able to redo from restart
 	# scratch that i like that, don't clear the history
 	# you can like replay your steps up to a point
@@ -83,19 +93,12 @@ func board_step():
 	hist.add_undo_method(board.undo_step())
 	hist.add_undo_property(self, "time", time)
 
-# TODO: OH MY GOSH OG MY GOSH OH MY GOSH
-# CONSOLIDATE ALL MOVEMENT INTO APPLYING VELOCITY
-# INCLUDING PLAYER MOVEMENT
-# I REALLY REALLY REALLY REALLY LIKE ADDING VELOCITY WITH TIME
-# NOT CHANGING OH MY GOSH THAT IS *GOOD*
 func step(move: Vector2i):
 	if advancing:
 		return
 	if won():
 		advance_level()
 		return
-	# the ability to wait removes some parity stuff
-	# but that stuff kinda sucks
 	if move == Vector2i.ZERO:
 		hist.create_action("wait")
 		board_step()
@@ -139,6 +142,7 @@ func advance_level():
 		print("level won")
 		await get_tree().create_timer(1).timeout
 	advancing = false
+	$Sun.rotation.y = 0
 	time = 0
 	level_num += 1
 	if level_num >= levels.size():
@@ -167,3 +171,6 @@ func audio_stream_randomizer_from_dir(dir: String) -> AudioStreamRandomizer:
 			stream.add_stream(-1, load(dir+"/"+file_name))
 		file_name = hit_dir.get_next()
 	return stream
+
+func slowmo() -> bool:
+	return Input.is_action_pressed("slowmo")
diff --git a/main.tscn b/main.tscn
index c9ee173..30b1114 100644
--- a/main.tscn
+++ b/main.tscn
@@ -24,7 +24,7 @@ albedo_color = Color(0.250136, 0.493198, 0.284902, 1)
 
 [sub_resource type="PlaneMesh" id="PlaneMesh_1bvp3"]
 material = SubResource("StandardMaterial3D_1bvp3")
-size = Vector2(200, 200)
+size = Vector2(64, 64)
 
 [node name="Main" type="Node3D"]
 script = ExtResource("1_ig7tw")
@@ -90,6 +90,15 @@ theme_override_font_sizes/normal_font_size = 32
 text = "T = 0"
 fit_content = true
 
+[node name="SlowmoIndicator" type="Label" parent="."]
+offset_left = 1065.0
+offset_top = 615.0
+offset_right = 1150.0
+offset_bottom = 659.0
+text = "slowmo"
+horizontal_alignment = 1
+vertical_alignment = 1
+
 [node name="Ground" type="MeshInstance3D" parent="."]
 transform = Transform3D(1, 0, 0, 0, 1, 0, 0, 0, 1, 0, -0.1, 0)
 mesh = SubResource("PlaneMesh_1bvp3")
diff --git a/piece.gd b/piece.gd
index 6fae8f3..22418ab 100644
--- a/piece.gd
+++ b/piece.gd
@@ -16,37 +16,45 @@ enum Type {
 }
 
 ## squares per second
-@export_range(0.1, 50) var anim_speed := 10.0
+@export_range(0.1, 50) var anim_speed_normal := 10.0
 @export_range(0.1, 50) var anim_speed_slow := 1.0
 
+var tweens: Array[Tween] = []
+
+func anim_speed() -> float:
+	return anim_speed_slow if Input.is_action_pressed("slowmo") else anim_speed_normal
+
+func anim_time() -> float:
+	return 1/anim_speed()
+
 #TODO: {un,}do_bump animation method (good visual clarity)
 
+func pos_of_lpos(pos: Vector2i, y := 0.0) -> Vector3:
+	return Vector3(pos.x, y, pos.y)
+
+func target_pos() -> Vector3:
+	return pos_of_lpos(lpos, position.y) + Vector3(0.5, 0, 0.5)
+
+func tween_to_target(tween := create_tween()):
+	tween.tween_property(self, "position", target_pos(), anim_time())
+	tweens.push_back(tween)
+
+func lpos_of_pos(pos: Vector3) -> Vector2i:
+	return Vector2i(int(pos.x), int(pos.z))
+
 ## logical position
-var lpos: Vector2i:
-	get:
-		return Vector2i(int(target_pos.x), int(target_pos.z))
-	set(val):
-		position = target_pos
-		start_pos = position
-		# we offset by 0.5 to be in the middle of the square (for easier editing)
-		target_pos = Vector3(val.x + 0.5, position.y, val.y + 0.5)
-		anim_progress = 0
+@onready var lpos := lpos_of_pos(position)
 
 ## logical velocity
 @export var lvel := Vector2i.ZERO
 
 @export var type: Piece.Type
 
-@onready var start_pos := position
-@onready var target_pos := position
-@onready var anim_progress := 1.0
 var speedometer: Label3D
 
 # TODO: much, *much* better animation is needed. ESPECIALLY if we're doing slowmo
 # (we're doing slowmo because it is SICK)
 func do_step(board: Board):
-	if lvel == Vector2i.ZERO:
-		return
 	var move := lvel.clampi(-1, 1)
 	var new_pos := lpos + move
 	# ball being collided *with* gets the remainder of the momentum
@@ -65,6 +73,7 @@ func do_step(board: Board):
 	var on_ice := !!board.type_at(lpos, Piece.Type.FloorIce)
 	if not on_ice:
 		lvel -= move
+	tween_to_target()
 
 func undo_step() -> Callable:
 	var old_pos := lpos
@@ -72,35 +81,39 @@ func undo_step() -> Callable:
 	return func():
 		lpos = old_pos
 		lvel = old_vel
+		tween_to_target()
 
 func do_move(move: Vector2i) -> Callable:
 	return func():
+		for tween in tweens:
+			tween.kill()
+		position = target_pos()
 		lpos += move
+		tween_to_target()
 
 func undo_move() -> Callable:
 	var old_pos := lpos
 	return func():
 		lpos = old_pos
+		tween_to_target()
 
 func do_push(move: Vector2i) -> Callable:
 	return func():
 		lvel += move
-		anim_progress = 1
 
 func undo_push() -> Callable:
 	var old_vel := lvel
 	return func():
 		lvel = old_vel
-		anim_progress = 1
 
 # no logical effect, purely for aesthetics
 # (and communicating !!!! player yes)
 func do_bump(move: Vector2i) -> Callable:
-	var real_lpos := lpos
 	return func():
-		lpos += move
-		lpos = real_lpos
-		anim_progress = 0.5
+		var tween := create_tween()
+		tween.tween_property(self, "position", target_pos() + pos_of_lpos(move)/2, anim_time())
+		tween_to_target(tween)
+		tweens.push_back(tween)
 
 # TODO?: maybe fix? maybe the bump should be more complicated?
 func undo_bump(move: Vector2i) -> Callable:
@@ -122,7 +135,6 @@ static func wall(pos: Vector2i) -> Piece:
 	return WALL.instantiate().with_lpos(pos)
 
 func _ready() -> void:
-	anim_progress = 1
 	speedometer = Label3D.new()
 	speedometer.billboard = BaseMaterial3D.BILLBOARD_ENABLED
 	speedometer.no_depth_test = true
@@ -137,14 +149,4 @@ func format_vel(vel: Vector2) -> String:
 		return "%d,%d" % [vel.x,vel.y]
 
 func _process(delta: float) -> void:
-	var speed := anim_speed_slow if Input.is_action_pressed("slowmo") else anim_speed
-	anim_progress = min(1, anim_progress + speed*delta)
-	position = start_pos.lerp(target_pos, anim_progress)
 	speedometer.text = format_vel(lvel)
-
-# weird hack only necessary for the soon to be gone text parser
-func with_lpos(pos: Vector2i) -> Piece:
-	lpos = pos
-	lpos = pos
-	anim_progress = 1
-	return self