diff --git a/src/termidi/__init__.py b/src/termidi/__init__.py
new file mode 100644
index 0000000..e69de29
diff --git a/src/termidi/asyncmidi.py b/src/termidi/asyncmidi.py
new file mode 100644
index 0000000..2479d03
--- /dev/null
+++ b/src/termidi/asyncmidi.py
@@ -0,0 +1,47 @@
+import asyncio
+import time
+from collections.abc import AsyncGenerator, Callable
+from typing import Never
+
+from mido import Message, MetaMessage, MidiFile, UnknownMetaMessage
+
+
+class AsyncMidiFile(MidiFile):
+    async def play_async(
+        self, meta_messages: bool = False, now: Callable[[], float] = time.monotonic
+    ) -> AsyncGenerator[Message | MetaMessage | UnknownMetaMessage, Never]:
+        """Play back all tracks.
+
+        The generator will sleep between each message by
+        default. Messages are yielded with correct timing. The time
+        attribute is set to the number of seconds slept since the
+        previous message.
+
+        By default you will only get normal MIDI messages. Pass
+        `meta_messages=True` if you also want meta messages.
+
+        You will receive copies of the original messages, so you can
+        safely modify them without ruining the tracks.
+
+        By default the system clock is used for the timing of yielded
+        MIDI events. To use a different clock (e.g. to synchronize to
+        an audio stream), pass `now=time_fn` where `time_fn` is a zero
+        argument function that yields the current time in seconds.
+        """
+        start_time = now()
+        input_time = 0.0
+
+        for msg in self:
+            input_time += msg.time
+
+            playback_time = now() - start_time
+            duration_to_next_event = input_time - playback_time
+
+            if duration_to_next_event > 0.001:
+                await asyncio.sleep(duration_to_next_event - 0.001)
+                # await asyncio.to_thread(time.sleep, duration_to_next_event)
+
+            if msg.is_meta and not meta_messages:
+                continue
+            else:
+                yield msg
diff --git a/src/termidi/const.py b/src/termidi/const.py
new file mode 100644
index 0000000..f1accfd
--- /dev/null
+++ b/src/termidi/const.py
@@ -0,0 +1,16 @@
+import enum
+
+
+class NoteNames12TET(enum.IntEnum):
+    C = enum.auto()
+    Db = enum.auto()
+    D = enum.auto()
+    Eb = enum.auto()
+    E = enum.auto()
+    F = enum.auto()
+    Gb = enum.auto()
+    G = enum.auto()
+    Ab = enum.auto()
+    A = enum.auto()
+    Bb = enum.auto()
+    B = enum.auto()
diff --git a/src/termidi/filters.py b/src/termidi/filters.py
new file mode 100644
index 0000000..ff4169b
--- /dev/null
+++ b/src/termidi/filters.py
@@ -0,0 +1,10 @@
+from typing import cast
+
+from mido import Message
+
+
+def translate_note_on_velocity_0(msg: Message) -> Message:
+    mdi = msg.dict()
+    if mdi["type"] == "note_on" and cast(int, getattr(msg, "velocity")) == 0:
+        return Message("note_off", channel=mdi["channel"], note=mdi["note"], velocity=64, time=msg.time)
+    return msg
diff --git a/src/termidi/main.py b/src/termidi/main.py
new file mode 100644
index 0000000..047cb83
--- /dev/null
+++ b/src/termidi/main.py
@@ -0,0 +1,260 @@
+import asyncio
+from asyncio import Event
+from pathlib import Path
+from typing import TYPE_CHECKING, ClassVar, cast, override
+
+import mido
+from mido.ports import BaseInput, BaseOutput
+from textual import work
+from textual.app import App, ComposeResult
+from textual.binding import BindingType
+from textual.containers import Horizontal, HorizontalScroll, Vertical
+from textual.driver import Driver
+from textual.types import CSSPathType
+from textual.widgets import Footer, Header
+
+from termidi.asyncmidi import AsyncMidiFile
+from termidi.const import NoteNames12TET
+from termidi.filters import translate_note_on_velocity_0
+from termidi.ui.components import PianoKeyboard12Keys
+from termidi.ui.fileselector import FileSelector
+from termidi.ui.streamselector import StreamSelector
+from termidi.ui.waterfall import WaterfallDisplay
+
+if TYPE_CHECKING:
+    from mido.messages.specs import NoteMsgDict
+
+
+def midi_note_to_keyname_octave(note: int, octave_offset: int = 0) -> tuple[NoteNames12TET, int]:
+    octave, semitone = divmod(note, 12)
+    return NoteNames12TET(semitone + 1), octave + octave_offset
+
+
+class PlayerApp(App[None]):
+    TITLE: str | None = "MIDI Player"
+    DEFAULT_CSS: ClassVar[str] = """
+    #scroll {
+        height: 1fr;
+    }
+    #inner {
+        height: 100%;
+        width: auto;
+    }
+    WaterfallDisplay {
+        height: 1fr;
+        width: 100%;
+    }
+    #kbd {
+        height: 5;
+        width: auto;
+
+        PianoKeyboard12Keys {
+            height: 5;
+        }
+    }
+    """
+    BINDINGS: ClassVar[list[BindingType]] = [
+        ("q", "quit", "Quit"),
+        ("i", "set_input", "Set MIDI input"),
+        ("o", "set_output", "Set MIDI output"),
+        ("f", "open_file", "Open MIDI file"),
+        (".", "stop_playing", "Stop playing"),
+    ]
+    input_stream: BaseInput
+    output_stream: BaseOutput
+    _midi_play_interrupt: Event
+    _midi_play_free: Event
+    _midi_input_interrupt: Event
+
+    def __init__(
+        self,
+        driver_class: type[Driver] | None = None,
+        css_path: CSSPathType | None = None,
+        watch_css: bool = False,
+        ansi_color: bool = False,
+    ) -> None:
+        super().__init__(driver_class, css_path, watch_css, ansi_color)
+        self.input_stream = mido.open_input()
+        self.output_stream = mido.open_output()
+        self._midi_play_interrupt = Event()
+        self._midi_play_free = Event()
+        self._midi_play_free.set()
+        self._midi_input_interrupt = Event()
+
+    @override
+    def compose(self) -> ComposeResult:
+        yield Header()
+        yield Footer()
+        wkwidth = 2
+        bkwidth = 2
+        with Vertical():
+            with HorizontalScroll(id="scroll"):
+                with Vertical(id="inner"):
+                    yield WaterfallDisplay(wkwidth, bkwidth, id="waterfall")
+                    with Horizontal(id="kbd"):
+                        yield PianoKeyboard12Keys(wkwidth, bkwidth)
+                        yield PianoKeyboard12Keys(wkwidth, bkwidth)
+                        yield PianoKeyboard12Keys(wkwidth, bkwidth)
+                        yield PianoKeyboard12Keys(wkwidth, bkwidth)
+                        yield PianoKeyboard12Keys(wkwidth, bkwidth)
+                        yield PianoKeyboard12Keys(wkwidth, bkwidth)
+                        yield PianoKeyboard12Keys(wkwidth, bkwidth)
+                        yield PianoKeyboard12Keys(wkwidth, bkwidth)
+                        yield PianoKeyboard12Keys(wkwidth, bkwidth)
+                        yield PianoKeyboard12Keys(wkwidth, bkwidth)
+                        yield PianoKeyboard12Keys(wkwidth, bkwidth)
+
+    def on_mount(self) -> None:
+        _ = self.set_interval(1 / 45, self._waterfall_tick)
+
+    def _waterfall_tick(self) -> None:
+        _ = self.query_one(WaterfallDisplay).tick()
+
+    def reset_keys(self) -> None:
+        for kbd in self.query(PianoKeyboard12Keys):
+            kbd.key_c_on = 0
+            kbd.key_db_on = 0
+            kbd.key_d_on = 0
+            kbd.key_eb_on = 0
+            kbd.key_e_on = 0
+            kbd.key_f_on = 0
+            kbd.key_gb_on = 0
+            kbd.key_g_on = 0
+            kbd.key_ab_on = 0
+            kbd.key_a_on = 0
+            kbd.key_bb_on = 0
+            kbd.key_b_on = 0
+
+    def set_key_on(
+        self, key: NoteNames12TET, on: bool = True, octave: int = 4, midi_note: int = -1, velocity: int = 64
+    ) -> None:
+        wf = self.query_one(WaterfallDisplay)
+        if midi_note >= 0:
+            if on:
+                wf.note_on(midi_note, velocity)
+            else:
+                wf.note_off(midi_note)
+        for idx, kbd in enumerate(self.query(PianoKeyboard12Keys)):
+            if idx != octave:
+                continue
+            match key:
+                case NoteNames12TET.C:
+                    kbd.key_c_on += on if on else -1
+                case NoteNames12TET.Db:
+                    kbd.key_db_on += on if on else -1
+                case NoteNames12TET.D:
+                    kbd.key_d_on += on if on else -1
+                case NoteNames12TET.Eb:
+                    kbd.key_eb_on += on if on else -1
+                case NoteNames12TET.E:
+                    kbd.key_e_on += on if on else -1
+                case NoteNames12TET.F:
+                    kbd.key_f_on += on if on else -1
+                case NoteNames12TET.Gb:
+                    kbd.key_gb_on += on if on else -1
+                case NoteNames12TET.G:
+                    kbd.key_g_on += on if on else -1
+                case NoteNames12TET.Ab:
+                    kbd.key_ab_on += on if on else -1
+                case NoteNames12TET.A:
+                    kbd.key_a_on += on if on else -1
+                case NoteNames12TET.Bb:
+                    kbd.key_bb_on += on if on else -1
+                case NoteNames12TET.B:
+                    kbd.key_b_on += on if on else -1
+            return
+
+    @work
+    async def play_file(self, file: Path) -> None:
+        self._midi_play_free.clear()
+        self._midi_play_interrupt.clear()
+        self.output_stream.reset()
+        self.reset_keys()
+        try:
+            async for m in AsyncMidiFile(str(file), clip=True).play_async():
+                if self._midi_play_interrupt.is_set():
+                    break
+                if isinstance(m, mido.Message):
+                    m = translate_note_on_velocity_0(m)
+                    mdi = m.dict()
+                    self.output_stream.send(m)
+                    if mdi["type"] in {"note_on", "note_off"}:
+                        nmdi = cast("NoteMsgDict", mdi)
+                        note, octave = midi_note_to_keyname_octave(nmdi["note"], octave_offset=0)
+                        is_on = mdi["type"] == "note_on" and nmdi["velocity"] != 0
+                        self.set_key_on(note, is_on, octave=octave, midi_note=nmdi["note"], velocity=nmdi["velocity"])
+        except OSError as e:
+            self.notify(repr(e), title="Error occured while playing:", severity="error")
+        finally:
+            self.output_stream.reset()
+            self.reset_keys()
+            self.query_one(WaterfallDisplay).reset()
+            self._midi_play_free.set()
+
+    @work
+    async def receive_input(self) -> None:
+        while True:
+            msg = self.input_stream.poll()
+            if self._midi_input_interrupt.is_set():
+                break
+            if msg is None:
+                await asyncio.sleep(0.0005)
+                continue
+            self.output_stream.send(msg)
+            mdi = msg.dict()
+            if mdi["type"] in {"note_on", "note_off"}:
+                nmdi = cast("NoteMsgDict", mdi)
+                note, octave = midi_note_to_keyname_octave(nmdi["note"], octave_offset=0)
+                is_on = mdi["type"] == "note_on" and nmdi["velocity"] != 0
+                self.set_key_on(note, is_on, octave=octave, midi_note=nmdi["note"], velocity=nmdi["velocity"])
+
+    async def playback_stop(self):
+        self._midi_play_interrupt.set()
+        _ = await self._midi_play_free.wait()
+
+    async def action_set_input(self) -> None:
+        stream = StreamSelector()
+        stream.options = mido.get_input_names()
+        await self.push_screen(stream, self._input_selected)
+
+    async def _input_selected(self, input: str | None):
+        if input is None:
+            return
+        _old_stream = self.input_stream
+        self._midi_input_interrupt.set()
+        self.input_stream = mido.open_input(input)
+        _old_stream.close()
+        self._midi_input_interrupt.clear()
+        _ = self.receive_input()
+
+    async def action_set_output(self) -> None:
+        stream = StreamSelector()
+        stream.options = mido.get_output_names()
+        await self.push_screen(stream, self._output_selected)
+
+    async def _output_selected(self, output: str | None):
+        if output is None:
+            return
+        _old_stream = self.output_stream
+        self.output_stream = mido.open_output(output)
+        _old_stream.reset()
+        _old_stream.close()
+
+    async def action_open_file(self) -> None:
+        fileselector = FileSelector("~/")
+        await self.push_screen(fileselector, self._file_selected)
+
+    async def _file_selected(self, file: Path | None) -> None:
+        if file is None:
+            return
+        self._midi_play_interrupt.set()
+        _ = await self._midi_play_free.wait()
+        _ = self.play_file(file)
+
+    async def action_stop_playing(self) -> None:
+        await self.playback_stop()
+
+
+def main():
+    app = PlayerApp()
+    app.run()
diff --git a/src/termidi/ui/components.py b/src/termidi/ui/components.py
new file mode 100644
index 0000000..2fd7b36
--- /dev/null
+++ b/src/termidi/ui/components.py
@@ -0,0 +1,133 @@
+from collections.abc import Generator
+from typing import Any, ClassVar, override
+
+from rich.segment import Segment
+from textual.geometry import Size
+from textual.reactive import reactive
+from textual.strip import Strip
+from textual.widget import Widget
+
+type KbState12Keys = tuple[int, int, int, int, int, int, int, int, int, int, int, int]
+_KB_BLACKCOLOR_12_KEYS: KbState12Keys = False, True, False, True, False, False, True, False, True, False, True, False
+
+
+class PianoKeyboard12Keys(Widget):
+    COMPONENT_CLASSES: ClassVar[set[str]] = {
+        "pianokb--white-key",
+        "pianokb--white-key-on",
+        "pianokb--black-key",
+        "pianokb--black-key-on",
+    }
+    DEFAULT_CSS: str = """
+    PianoKeyboard12Keys {
+        width: auto;
+    }
+    PianoKeyboard12Keys .pianokb--white-key {
+        color: auto;
+        background: white;
+    }
+    PianoKeyboard12Keys .pianokb--white-key-on {
+        color: auto;
+        background: darkgrey;
+    }
+    PianoKeyboard12Keys .pianokb--black-key {
+        color: auto;
+        background: black;
+    }
+    PianoKeyboard12Keys .pianokb--black-key-on {
+        color: auto;
+        background: grey;
+    }
+    """
+
+    key_c_on: reactive[int] = reactive(0)
+    key_db_on: reactive[int] = reactive(0)
+    key_d_on: reactive[int] = reactive(0)
+    key_eb_on: reactive[int] = reactive(0)
+    key_e_on: reactive[int] = reactive(0)
+    key_f_on: reactive[int] = reactive(0)
+    key_gb_on: reactive[int] = reactive(0)
+    key_g_on: reactive[int] = reactive(0)
+    key_ab_on: reactive[int] = reactive(0)
+    key_a_on: reactive[int] = reactive(0)
+    key_bb_on: reactive[int] = reactive(0)
+    key_b_on: reactive[int] = reactive(0)
+
+    keyboard_height: reactive[int] = reactive(5)
+    whitekey_width: reactive[int] = reactive(1)
+    blackkey_width: reactive[int] = reactive(1)
+    blackkey_height: reactive[int] = reactive(3)
+
+    def __init__(
+        self,
+        whitekey_width: int | None = None,
+        blackkey_width: int | None = None,
+        blackkey_height: int | None = None,
+        *,
+        name: str | None = None,
+        id: str | None = None,
+        classes: str | None = None,
+        disabled: bool = False,
+        markup: bool = True,
+    ):
+        super().__init__(name=name, id=id, classes=classes, disabled=disabled, markup=markup)
+        if whitekey_width is not None:
+            self.whitekey_width = whitekey_width
+        if blackkey_width is not None:
+            self.blackkey_width = blackkey_width
+        if blackkey_height is not None:
+            self.blackkey_height = blackkey_height
+
+    @override
+    def get_content_width(self, container: Size, viewport: Size) -> int:
+        return self.whitekey_width * 7 + self.blackkey_width * 5
+
+    def _line(self, y: int, kbstate: KbState12Keys) -> Generator[Segment, Any, None]:
+        whitekey = self.get_component_rich_style("pianokb--white-key")
+        whitekey_on = self.get_component_rich_style("pianokb--white-key-on")
+        blackkey = self.get_component_rich_style("pianokb--black-key")
+        blackkey_on = self.get_component_rich_style("pianokb--black-key-on")
+        whitesep = y >= self.blackkey_height
+        whitesep_sym = "\N{BOX DRAWINGS LIGHT VERTICAL}"
+
+        seg_whitekey = Segment(" " * self.whitekey_width, whitekey), Segment(" " * self.whitekey_width, whitekey_on)
+        seg_blackkey = Segment(" " * self.blackkey_width, blackkey), Segment(" " * self.blackkey_width, blackkey_on)
+        seg_blackkey_whitesep = (
+            Segment(" " * (self.blackkey_width // 2), whitekey),
+            Segment(" " * (self.blackkey_width // 2), whitekey_on),
+        )
+        seg_blackkey_whitesep_bar = Segment(whitesep_sym, style=whitekey)
+        for on, black, left_on, right_on in zip(kbstate, _KB_BLACKCOLOR_12_KEYS, [0, *kbstate[:-1]], [*kbstate[1:], 0]):
+            if black:
+                if not whitesep:
+                    yield seg_blackkey[not not on]
+                else:
+                    yield seg_blackkey_whitesep[not not left_on]
+                    if self.blackkey_width & 1:
+                        yield seg_blackkey_whitesep_bar
+                    yield seg_blackkey_whitesep[not not right_on]
+            else:
+                yield seg_whitekey[not not on]
+
+    @override
+    def render_line(self, y: int) -> Strip:
+        return Strip(
+            self._line(
+                y,
+                (
+                    self.key_c_on,
+                    self.key_db_on,
+                    self.key_d_on,
+                    self.key_eb_on,
+                    self.key_e_on,
+                    self.key_f_on,
+                    self.key_gb_on,
+                    self.key_g_on,
+                    self.key_ab_on,
+                    self.key_a_on,
+                    self.key_bb_on,
+                    self.key_b_on,
+                ),
+            ),
+            self.whitekey_width * 7 + self.blackkey_width * 5,
+        )
diff --git a/src/termidi/ui/fileselector.py b/src/termidi/ui/fileselector.py
new file mode 100644
index 0000000..b61a936
--- /dev/null
+++ b/src/termidi/ui/fileselector.py
@@ -0,0 +1,56 @@
+from collections.abc import Iterable
+from pathlib import Path
+from typing import ClassVar, override
+
+from textual.app import ComposeResult
+from textual.binding import BindingType
+from textual.reactive import reactive
+from textual.screen import ModalScreen
+from textual.widgets import DirectoryTree
+
+
+class FilteredDirectoryTree(DirectoryTree):
+    show_hidden: reactive[bool] = reactive(False)
+    suffixes: tuple[str, ...] = (".mid", ".smf")
+
+    @override
+    def filter_paths(self, paths: Iterable[Path]) -> Iterable[Path]:
+        for path in paths:
+            if not self.show_hidden and path.name.startswith("."):
+                continue
+            if path.is_file() and path.suffix not in self.suffixes:
+                continue
+            yield path
+
+
+class FileSelector(ModalScreen[Path]):
+    DEFAULT_CSS: str = """
+    FileSelector {
+        align: center middle;
+    }
+
+    FilteredDirectoryTree {
+        min-width: 40;
+        max-width: 120;
+        max-height: 40;
+    }
+    """
+    BINDINGS: ClassVar[list[BindingType]] = [("esc", "dismiss", "Cancel")]
+
+    show_hidden: reactive[bool] = reactive(False)
+
+    start: str | Path
+
+    def __init__(
+        self, start: str | Path = "./", name: str | None = None, id: str | None = None, classes: str | None = None
+    ) -> None:
+        super().__init__(name, id, classes)
+        self.start = start
+
+    @override
+    def compose(self) -> ComposeResult:
+        yield FilteredDirectoryTree(self.start).data_bind(self.__class__.show_hidden)
+
+    def on_directory_tree_file_selected(self, event: FilteredDirectoryTree.FileSelected):
+        if event.path.is_file() and event.path.suffix in FilteredDirectoryTree.suffixes:
+            _ = self.dismiss(event.path)
diff --git a/src/termidi/ui/streamselector.py b/src/termidi/ui/streamselector.py
new file mode 100644
index 0000000..1cf6d4f
--- /dev/null
+++ b/src/termidi/ui/streamselector.py
@@ -0,0 +1,35 @@
+from collections.abc import Sequence
+from typing import ClassVar, override
+
+from textual.app import ComposeResult
+from textual.binding import BindingType
+from textual.reactive import var
+from textual.screen import ModalScreen
+from textual.widgets import OptionList
+
+
+class StreamSelector(ModalScreen[str]):
+    DEFAULT_CSS: str = """
+    StreamSelector {
+        align: center middle;
+    }
+
+    OptionList {
+        min-width: 40;
+        max-width: 80;
+        max-height: 40;
+    }
+    """
+    options: var[Sequence[str]] = var(list)
+    BINDINGS: ClassVar[list[BindingType]] = [("esc", "dismiss", "Cancel")]
+
+    @override
+    def compose(self) -> ComposeResult:
+        yield OptionList()
+
+    def on_mount(self) -> None:
+        option_list = self.query_one(OptionList)
+        _ = option_list.clear_options().add_options(self.options)
+
+    def on_option_list_option_selected(self, event: OptionList.OptionSelected) -> None:
+        _ = self.dismiss(self.options[event.option_index])
diff --git a/src/termidi/ui/waterfall.py b/src/termidi/ui/waterfall.py
new file mode 100644
index 0000000..0b78f2b
--- /dev/null
+++ b/src/termidi/ui/waterfall.py
@@ -0,0 +1,178 @@
+"""Strip-based note waterfall widget for termidi."""
+
+from collections.abc import Generator
+from typing import Any, ClassVar, override
+
+from rich.color import Color
+from rich.segment import Segment
+from rich.style import Style
+from textual import work
+from textual.geometry import Size
+from textual.strip import Strip
+from textual.widget import Widget
+
+# Note index within an octave (0-indexed) → is black key
+_IS_BLACK: tuple[bool, ...] = (False, True, False, True, False, False, True, False, True, False, True, False)
+
+# Number of cols occupied by each note slot given key widths
+# White keys → whitekey_width, black keys → blackkey_width
+# This must match PianoKeyboard12Keys._line() column layout exactly.
+
+
+def _note_col_offsets(whitekey_width: int, blackkey_width: int) -> tuple[int, ...]:
+    """Return the starting column of each of the 12 notes within one octave."""
+    offsets: list[int] = []
+    col = 0
+    for is_black in _IS_BLACK:
+        offsets.append(col)
+        col += blackkey_width if is_black else whitekey_width
+    return tuple(offsets)
+
+
+def _octave_width(whitekey_width: int, blackkey_width: int) -> int:
+    return whitekey_width * 7 + blackkey_width * 5
+
+
+# midi note → (octave_idx, note_idx_in_octave)
+def _note_to_octave_semitone(midi_note: int) -> tuple[int, int]:
+    octave, semitone = divmod(midi_note, 12)
+    return octave, semitone
+
+
+# Velocity → RGB color: low velocity = dim blue, high = bright cyan/white
+def _velocity_color(velocity: int) -> Color:
+    t = max(0, min(127, velocity)) / 127.0
+    r = int(80 + 175 * t)
+    g = int(100 + 155 * t)
+    b = int(200 + 55 * t)
+    return Color.from_rgb(r, g, b)
+
+
+class WaterfallDisplay(Widget):
+    """Scrolling note waterfall rendered with Strip/Segment, aligned to PianoKeyboard12Keys."""
+
+    COMPONENT_CLASSES: ClassVar[set[str]] = {
+        "waterfall--background",
+    }
+    DEFAULT_CSS: ClassVar[str] = """
+    WaterfallDisplay {
+        background: $surface-darken-2;
+    }
+    """
+
+    def __init__(
+        self,
+        whitekey_width: int = 2,
+        blackkey_width: int = 2,
+        num_octaves: int = 11,
+        *,
+        name: str | None = None,
+        id: str | None = None,
+        classes: str | None = None,
+        disabled: bool = False,
+    ) -> None:
+        super().__init__(name=name, id=id, classes=classes, disabled=disabled)
+        self._whitekey_width: int = whitekey_width
+        self._blackkey_width: int = blackkey_width
+        self._num_octaves: int = num_octaves
+        self._oct_width: int = _octave_width(whitekey_width, blackkey_width)
+        self._note_offsets: tuple[int, ...] = _note_col_offsets(whitekey_width, blackkey_width)
+        self._total_width: int = self._oct_width * num_octaves
+        # buffer: list of rows (bottom = index 0, top = last index)
+        # Each row is a dict mapping col_start → (width, Style)
+        self._rows: list[dict[int, tuple[int, Style]]] = []
+        # Currently held notes: midi_note → velocity (for painting while key is held)
+        self._held: dict[int, int] = {}
+
+    @override
+    def get_content_width(self, container: Size, viewport: Size) -> int:
+        return self._total_width
+
+    def _render_width(self) -> int:
+        w = self.size.width
+        return w if w > 0 else self._total_width
+
+    def _ensure_rows(self, height: int) -> None:
+        """Grow the row buffer to match current widget height."""
+        while len(self._rows) < height:
+            self._rows.append({})
+
+    def _col_for_note(self, midi_note: int) -> int:
+        """Column offset (within the full waterfall width) for a given MIDI note."""
+        octave_idx, semitone = _note_to_octave_semitone(midi_note)
+        if octave_idx >= self._num_octaves:
+            return -1
+        return octave_idx * self._oct_width + self._note_offsets[semitone]
+
+    def _note_width(self, midi_note: int) -> int:
+        _, semitone = _note_to_octave_semitone(midi_note)
+        return self._blackkey_width if _IS_BLACK[semitone] else self._whitekey_width
+
+    def note_on(self, midi_note: int, velocity: int) -> None:
+        """Call when a note_on event is received."""
+        if velocity == 0:
+            self.note_off(midi_note)
+            return
+        self._held[midi_note] = velocity
+
+    def note_off(self, midi_note: int) -> None:
+        """Call when a note_off event is received."""
+        _ = self._held.pop(midi_note, None)
+
+    def reset(self) -> None:
+        """Clear all held notes and the buffer."""
+        self._held.clear()
+        self._rows.clear()
+        _ = self.refresh()
+
+    @work(thread=True)
+    def tick(self) -> None:
+        """Advance the waterfall by one row. Call on a regular interval."""
+        height = self.size.height
+        if height <= 0:
+            return
+        self._ensure_rows(height)
+        # Build the new top row from currently held notes
+        new_row: dict[int, tuple[int, Style]] = {}
+        for midi_note, velocity in self._held.items():
+            col = self._col_for_note(midi_note)
+            if col < 0:
+                continue
+            w = self._note_width(midi_note)
+            color = _velocity_color(velocity)
+            new_row[col] = (w, Style(bgcolor=color))
+        # Shift buffer: discard oldest (index 0), append new row at end
+        if len(self._rows) >= height:
+            _ = self._rows.pop(0)
+        self._rows.append(new_row)
+        _ = self.refresh()
+
+    def _render_row(self, row: dict[int, tuple[int, Style]], width: int) -> Generator[Segment, Any, None]:
+        """Yield Segments for one waterfall row."""
+        bg = Style.null()
+        pos = 0
+        for col in sorted(row):
+            if col >= width:
+                break
+            if col > pos:
+                yield Segment(" " * (col - pos), bg)
+            w, style = row[col]
+            w = min(w, width - col)
+            yield Segment(" " * w, style)
+            pos = col + w
+        if pos < width:
+            yield Segment(" " * (width - pos), bg)
+
+    @override
+    def render_line(self, y: int) -> Strip:
+        height = self.size.height
+        width = self._render_width()
+        self._ensure_rows(height)
+        # y=0 is top (oldest notes), y=height-1 is bottom (newest, adjacent to keyboard).
+        # _rows is oldest-first; offset so that empty space sits at the top when buffer
+        # hasn't filled yet.
+        row_idx = y - (height - len(self._rows))
+        if row_idx < 0 or row_idx >= len(self._rows):
+            return Strip([Segment(" " * width)], width)
+        row = self._rows[row_idx]
+        return Strip(list(self._render_row(row, width)), width)
diff --git a/src/termidi/utils.py b/src/termidi/utils.py
new file mode 100644
index 0000000..e69de29
diff --git a/tests/__init__.py b/tests/__init__.py
new file mode 100644
index 0000000..e69de29