diff --git a/Userland/Applets/Audio/main.cpp b/Userland/Applets/Audio/main.cpp
index be6247078f..0e1e736942 100644
--- a/Userland/Applets/Audio/main.cpp
+++ b/Userland/Applets/Audio/main.cpp
@@ -38,8 +38,8 @@ public:
             update();
         };
 
-        m_audio_client->on_main_mix_volume_change = [this](int volume) {
-            m_audio_volume = volume;
+        m_audio_client->on_main_mix_volume_change = [this](double volume) {
+            m_audio_volume = static_cast<int>(volume * 100);
             if (!m_audio_muted)
                 update();
         };
@@ -87,14 +87,14 @@ public:
         };
 
         m_slider = m_root_container->add<GUI::VerticalSlider>();
-        m_slider->set_max(20);
-        int non_log_volume = sqrt(100 * m_audio_volume);
-        m_slider->set_value(-(non_log_volume / 5.0f) + 20);
+        m_slider->set_max(100);
+        m_slider->set_page_step(5);
+        m_slider->set_step(5);
+        m_slider->set_value(m_slider->max() - m_audio_volume);
         m_slider->set_knob_size_mode(GUI::Slider::KnobSizeMode::Proportional);
         m_slider->on_change = [&](int value) {
-            int volume = clamp((20 - value) * 5, 0, 100);
-            double volume_log = ((volume / 100.0) * (volume / 100.0)) * 100.0;
-            m_audio_client->set_main_mix_volume(static_cast<i32>(volume_log));
+            double volume = clamp(static_cast<double>(m_slider->max() - value) / m_slider->max(), 0.0, 1.0);
+            m_audio_client->set_main_mix_volume(volume);
             update();
         };
 
@@ -131,8 +131,7 @@ private:
     {
         if (m_audio_muted)
             return;
-        int new_slider_value = m_slider->value() + event.wheel_delta() / 4;
-        m_slider->set_value(new_slider_value);
+        m_slider->dispatch_event(event);
         update();
     }
 
diff --git a/Userland/Applications/Piano/Music.h b/Userland/Applications/Piano/Music.h
index a1080155e9..ce2c814735 100644
--- a/Userland/Applications/Piano/Music.h
+++ b/Userland/Applications/Piano/Music.h
@@ -30,7 +30,8 @@ constexpr int buffer_size = sample_count * sizeof(Sample);
 
 constexpr double sample_rate = 44100;
 
-constexpr double volume_factor = 1800;
+// Headroom for the synth
+constexpr double volume_factor = 0.1;
 
 enum Switch {
     Off,
diff --git a/Userland/Applications/Piano/Track.cpp b/Userland/Applications/Piano/Track.cpp
index f077efc994..5727a945f7 100644
--- a/Userland/Applications/Piano/Track.cpp
+++ b/Userland/Applications/Piano/Track.cpp
@@ -95,8 +95,8 @@ void Track::fill_sample(Sample& sample)
         default:
             VERIFY_NOT_REACHED();
         }
-        new_sample.left += note_sample.left * m_power[note] * volume_factor * (static_cast<double>(volume()) / volume_max);
-        new_sample.right += note_sample.right * m_power[note] * volume_factor * (static_cast<double>(volume()) / volume_max);
+        new_sample.left += note_sample.left * m_power[note] * NumericLimits<i16>::max() * volume_factor * (static_cast<double>(volume()) / volume_max);
+        new_sample.right += note_sample.right * m_power[note] * NumericLimits<i16>::max() * volume_factor * (static_cast<double>(volume()) / volume_max);
     }
 
     auto new_sample_dsp = LibDSP::Signal(LibDSP::Sample { new_sample.left / NumericLimits<i16>::max(), new_sample.right / NumericLimits<i16>::max() });
@@ -105,6 +105,9 @@ void Track::fill_sample(Sample& sample)
     new_sample.left = delayed_sample.left * NumericLimits<i16>::max();
     new_sample.right = delayed_sample.right * NumericLimits<i16>::max();
 
+    new_sample.left = clamp(new_sample.left, NumericLimits<i16>::min(), NumericLimits<i16>::max());
+    new_sample.right = clamp(new_sample.right, NumericLimits<i16>::min(), NumericLimits<i16>::max());
+
     sample.left += new_sample.left;
     sample.right += new_sample.right;
 }
diff --git a/Userland/Applications/SoundPlayer/Player.h b/Userland/Applications/SoundPlayer/Player.h
index bce057c60d..23b4fdfebe 100644
--- a/Userland/Applications/SoundPlayer/Player.h
+++ b/Userland/Applications/SoundPlayer/Player.h
@@ -47,7 +47,7 @@ public:
     virtual void set_volume(double volume)
     {
         m_player_state.volume = volume;
-        client_connection().set_main_mix_volume((double)(volume * 100));
+        client_connection().set_self_volume(volume);
     }
     virtual void set_loaded_file_samplerate(int samplerate) { m_player_state.loaded_file_samplerate = samplerate; }
     virtual void set_looping_file(bool loop)
diff --git a/Userland/Libraries/LibAudio/Buffer.h b/Userland/Libraries/LibAudio/Buffer.h
index ebf40af0bd..1a518784e7 100644
--- a/Userland/Libraries/LibAudio/Buffer.h
+++ b/Userland/Libraries/LibAudio/Buffer.h
@@ -8,6 +8,7 @@
 #pragma once
 
 #include <AK/ByteBuffer.h>
+#include <AK/Math.h>
 #include <AK/MemoryStream.h>
 #include <AK/String.h>
 #include <AK/Types.h>
@@ -16,25 +17,32 @@
 #include <string.h>
 
 namespace Audio {
+using namespace AK::Exponentials;
+
+// Constants for logarithmic volume. See Frame::operator*
+// Corresponds to 60dB
+constexpr double DYNAMIC_RANGE = 1000;
+constexpr double VOLUME_A = 1 / DYNAMIC_RANGE;
+double const VOLUME_B = log(DYNAMIC_RANGE);
 
 // A single sample in an audio buffer.
 // Values are floating point, and should range from -1.0 to +1.0
 struct Frame {
-    Frame()
+    constexpr Frame()
         : left(0)
         , right(0)
     {
     }
 
     // For mono
-    Frame(double left)
+    constexpr Frame(double left)
         : left(left)
         , right(left)
     {
     }
 
     // For stereo
-    Frame(double left, double right)
+    constexpr Frame(double left, double right)
         : left(left)
         , right(right)
     {
@@ -53,26 +61,54 @@ struct Frame {
             right = -1;
     }
 
-    void scale(int percent)
+    // Logarithmic scaling, as audio should ALWAYS do.
+    // Reference: https://www.dr-lex.be/info-stuff/volumecontrols.html
+    // We use the curve `factor = a * exp(b * change)`,
+    // where change is the input fraction we want to change by,
+    // a = 1/1000, b = ln(1000) = 6.908 and factor is the multiplier used.
+    // The value 1000 represents the dynamic range in sound pressure, which corresponds to 60 dB(A).
+    // This is a good dynamic range because it can represent all loudness values from
+    // 30 dB(A) (barely hearable with background noise)
+    // to 90 dB(A) (almost too loud to hear and about the reasonable limit of actual sound equipment).
+    ALWAYS_INLINE Frame& log_multiply(double const change)
     {
-        double pct = (double)percent / 100.0;
-        left *= pct;
-        right *= pct;
+        double factor = VOLUME_A * exp(VOLUME_B * change);
+        left *= factor;
+        right *= factor;
+        return *this;
     }
 
-    // FIXME: This is temporary until we have log scaling
-    Frame scaled(double fraction) const
+    ALWAYS_INLINE Frame log_multiplied(double const volume_change)
     {
-        return Frame { left * fraction, right * fraction };
+        Frame new_frame { left, right };
+        new_frame.log_multiply(volume_change);
+        return new_frame;
     }
 
-    Frame& operator+=(const Frame& other)
+    constexpr Frame& operator*=(double const mult)
+    {
+        left *= mult;
+        right *= mult;
+        return *this;
+    }
+
+    constexpr Frame operator*(double const mult)
+    {
+        return { left * mult, right * mult };
+    }
+
+    constexpr Frame& operator+=(Frame const& other)
     {
         left += other.left;
         right += other.right;
         return *this;
     }
 
+    constexpr Frame operator+(Frame const& other)
+    {
+        return { left + other.left, right + other.right };
+    }
+
     double left;
     double right;
 };
diff --git a/Userland/Libraries/LibAudio/ClientConnection.cpp b/Userland/Libraries/LibAudio/ClientConnection.cpp
index 81e69e14df..949683e42a 100644
--- a/Userland/Libraries/LibAudio/ClientConnection.cpp
+++ b/Userland/Libraries/LibAudio/ClientConnection.cpp
@@ -48,7 +48,7 @@ void ClientConnection::muted_state_changed(bool muted)
         on_muted_state_change(muted);
 }
 
-void ClientConnection::main_mix_volume_changed(i32 volume)
+void ClientConnection::main_mix_volume_changed(double volume)
 {
     if (on_main_mix_volume_change)
         on_main_mix_volume_change(volume);
diff --git a/Userland/Libraries/LibAudio/ClientConnection.h b/Userland/Libraries/LibAudio/ClientConnection.h
index b69696642b..226b65215f 100644
--- a/Userland/Libraries/LibAudio/ClientConnection.h
+++ b/Userland/Libraries/LibAudio/ClientConnection.h
@@ -27,12 +27,12 @@ public:
 
     Function<void(i32 buffer_id)> on_finish_playing_buffer;
     Function<void(bool muted)> on_muted_state_change;
-    Function<void(int volume)> on_main_mix_volume_change;
+    Function<void(double volume)> on_main_mix_volume_change;
 
 private:
     virtual void finished_playing_buffer(i32) override;
     virtual void muted_state_changed(bool) override;
-    virtual void main_mix_volume_changed(i32) override;
+    virtual void main_mix_volume_changed(double) override;
 };
 
 }
diff --git a/Userland/Services/AudioServer/AudioClient.ipc b/Userland/Services/AudioServer/AudioClient.ipc
index 56b4f89d1e..ee8922f944 100644
--- a/Userland/Services/AudioServer/AudioClient.ipc
+++ b/Userland/Services/AudioServer/AudioClient.ipc
@@ -4,5 +4,5 @@ endpoint AudioClient
 {
     finished_playing_buffer(i32 buffer_id) =|
     muted_state_changed(bool muted) =|
-    main_mix_volume_changed(i32 volume) =|
+    main_mix_volume_changed(double volume) =|
 }
diff --git a/Userland/Services/AudioServer/AudioServer.ipc b/Userland/Services/AudioServer/AudioServer.ipc
index d48a50c4cf..9b4973130b 100644
--- a/Userland/Services/AudioServer/AudioServer.ipc
+++ b/Userland/Services/AudioServer/AudioServer.ipc
@@ -5,8 +5,8 @@ endpoint AudioServer
     // Mixer functions
     set_muted(bool muted) => ()
     get_muted() => (bool muted)
-    get_main_mix_volume() => (i32 volume)
-    set_main_mix_volume(i32 volume) => ()
+    get_main_mix_volume() => (double volume)
+    set_main_mix_volume(double volume) => ()
 
     // Audio device
     set_sample_rate(u16 sample_rate) => ()
diff --git a/Userland/Services/AudioServer/ClientConnection.cpp b/Userland/Services/AudioServer/ClientConnection.cpp
index dc151b8ce5..a267b7126b 100644
--- a/Userland/Services/AudioServer/ClientConnection.cpp
+++ b/Userland/Services/AudioServer/ClientConnection.cpp
@@ -48,7 +48,7 @@ void ClientConnection::did_change_muted_state(Badge<Mixer>, bool muted)
     async_muted_state_changed(muted);
 }
 
-void ClientConnection::did_change_main_mix_volume(Badge<Mixer>, int volume)
+void ClientConnection::did_change_main_mix_volume(Badge<Mixer>, double volume)
 {
     async_main_mix_volume_changed(volume);
 }
@@ -58,7 +58,7 @@ Messages::AudioServer::GetMainMixVolumeResponse ClientConnection::get_main_mix_v
     return m_mixer.main_volume();
 }
 
-void ClientConnection::set_main_mix_volume(i32 volume)
+void ClientConnection::set_main_mix_volume(double volume)
 {
     m_mixer.set_main_volume(volume);
 }
diff --git a/Userland/Services/AudioServer/ClientConnection.h b/Userland/Services/AudioServer/ClientConnection.h
index 96b4d3fdbf..c5276e1097 100644
--- a/Userland/Services/AudioServer/ClientConnection.h
+++ b/Userland/Services/AudioServer/ClientConnection.h
@@ -28,7 +28,7 @@ public:
 
     void did_finish_playing_buffer(Badge<BufferQueue>, int buffer_id);
     void did_change_muted_state(Badge<Mixer>, bool muted);
-    void did_change_main_mix_volume(Badge<Mixer>, int volume);
+    void did_change_main_mix_volume(Badge<Mixer>, double volume);
 
     virtual void die() override;
 
@@ -36,7 +36,7 @@ public:
 
 private:
     virtual Messages::AudioServer::GetMainMixVolumeResponse get_main_mix_volume() override;
-    virtual void set_main_mix_volume(i32) override;
+    virtual void set_main_mix_volume(double) override;
     virtual Messages::AudioServer::EnqueueBufferResponse enqueue_buffer(Core::AnonymousBuffer const&, i32, int) override;
     virtual Messages::AudioServer::GetRemainingSamplesResponse get_remaining_samples() override;
     virtual Messages::AudioServer::GetPlayedSamplesResponse get_played_samples() override;
diff --git a/Userland/Services/AudioServer/FadingProperty.h b/Userland/Services/AudioServer/FadingProperty.h
new file mode 100644
index 0000000000..5df78fe79c
--- /dev/null
+++ b/Userland/Services/AudioServer/FadingProperty.h
@@ -0,0 +1,85 @@
+/*
+ * Copyright (c) 2021, kleines Filmröllchen <malu.bertsch@gmail.com>.
+ *
+ * SPDX-License-Identifier: BSD-2-Clause
+ */
+
+#pragma once
+
+#include "Mixer.h"
+#include <compare>
+
+namespace AudioServer {
+
+// This is in buffer counts.
+// As each buffer is approx 1/40 of a second, this means about 1/4 of a second of fade time.
+constexpr int DEFAULT_FADE_TIME = 10;
+
+// A property of an audio system that needs to fade briefly whenever changed.
+template<typename T>
+class FadingProperty {
+public:
+    FadingProperty(T const value)
+        : FadingProperty(value, DEFAULT_FADE_TIME)
+    {
+    }
+    FadingProperty(T const value, int const fade_time)
+        : m_old_value(value)
+        , m_new_value(move(value))
+        , m_fade_time(fade_time)
+    {
+    }
+    virtual ~FadingProperty()
+    {
+        m_old_value.~T();
+        m_new_value.~T();
+    }
+
+    FadingProperty<T>& operator=(T const& new_value)
+    {
+        // The origin of the fade is wherever we're right now.
+        m_old_value = static_cast<T>(*this);
+        m_new_value = new_value;
+        m_current_fade = 0;
+        return *this;
+    }
+    FadingProperty<T>& operator=(FadingProperty<T> const&) = delete;
+
+    operator T() const
+    {
+        if (!is_fading())
+            return m_new_value;
+        return m_old_value * (1 - m_current_fade) + m_new_value * (m_current_fade);
+    }
+
+    auto operator<=>(FadingProperty<T> const& other) const
+    {
+        return static_cast<T>(this) <=> static_cast<T>(other);
+    }
+
+    auto operator<=>(T const& other) const
+    {
+        return static_cast<T>(*this) <=> other;
+    }
+
+    void advance_time()
+    {
+        m_current_fade += 1.0 / static_cast<double>(m_fade_time);
+        m_current_fade = clamp(m_current_fade, 0.0, 1.0);
+    }
+
+    bool is_fading() const
+    {
+        return m_current_fade < 1;
+    }
+
+    T target() const { return m_new_value; }
+
+private:
+    T m_old_value {};
+    T m_new_value {};
+    double m_current_fade { 0 };
+    int const m_fade_time;
+};
+
+}
diff --git a/Userland/Services/AudioServer/Mixer.cpp b/Userland/Services/AudioServer/Mixer.cpp
index 8911018748..386b79857a 100644
--- a/Userland/Services/AudioServer/Mixer.cpp
+++ b/Userland/Services/AudioServer/Mixer.cpp
@@ -40,7 +40,7 @@ Mixer::Mixer(NonnullRefPtr<Core::ConfigFile> config)
     pthread_cond_init(&m_pending_cond, nullptr);
 
     m_muted = m_config->read_bool_entry("Master", "Mute", false);
-    m_main_volume = m_config->read_num_entry("Master", "Volume", 100);
+    m_main_volume = static_cast<double>(m_config->read_num_entry("Master", "Volume", 100)) / 100.0;
 
     m_sound_thread->start();
 }
@@ -78,18 +78,23 @@ void Mixer::mix()
         Audio::Frame mixed_buffer[1024];
         auto mixed_buffer_length = (int)(sizeof(mixed_buffer) / sizeof(Audio::Frame));
 
+        m_main_volume.advance_time();
+
+        int active_queues = 0;
         // Mix the buffers together into the output
         for (auto& queue : active_mix_queues) {
             if (!queue->client()) {
                 queue->clear();
                 continue;
             }
+            ++active_queues;
 
             for (int i = 0; i < mixed_buffer_length; ++i) {
                 auto& mixed_sample = mixed_buffer[i];
                 Audio::Frame sample;
                 if (!queue->get_next_sample(sample))
                     break;
+                sample.log_multiply(SAMPLE_HEADROOM);
                 mixed_sample += sample;
             }
         }
@@ -103,7 +108,11 @@ void Mixer::mix()
             for (int i = 0; i < mixed_buffer_length; ++i) {
                 auto& mixed_sample = mixed_buffer[i];
 
-                mixed_sample.scale(m_main_volume);
+                // Even though it's not realistic, the user expects no sound at 0%.
+                if (m_main_volume < 0.01)
+                    mixed_sample = { 0 };
+                else
+                    mixed_sample.log_multiply(m_main_volume);
                 mixed_sample.clip();
 
                 LittleEndian<i16> out_sample;
@@ -121,20 +130,20 @@ void Mixer::mix()
     }
 }
 
-void Mixer::set_main_volume(int volume)
+void Mixer::set_main_volume(double volume)
 {
     if (volume < 0)
         m_main_volume = 0;
-    else if (volume > 200)
-        m_main_volume = 200;
+    else if (volume > 2)
+        m_main_volume = 2;
     else
         m_main_volume = volume;
 
-    m_config->write_num_entry("Master", "Volume", volume);
+    m_config->write_num_entry("Master", "Volume", static_cast<int>(volume * 100));
     request_setting_sync();
 
     ClientConnection::for_each([&](ClientConnection& client) {
-        client.did_change_main_mix_volume({}, m_main_volume);
+        client.did_change_main_mix_volume({}, main_volume());
     });
 }
 
diff --git a/Userland/Services/AudioServer/Mixer.h b/Userland/Services/AudioServer/Mixer.h
index 33549fca3c..1b4c09a516 100644
--- a/Userland/Services/AudioServer/Mixer.h
+++ b/Userland/Services/AudioServer/Mixer.h
@@ -8,6 +8,7 @@
 #pragma once
 
 #include "ClientConnection.h"
+#include "FadingProperty.h"
 #include <AK/Atomic.h>
 #include <AK/Badge.h>
 #include <AK/ByteBuffer.h>
@@ -23,6 +24,10 @@
 
 namespace AudioServer {
 
+// Headroom, i.e. fixed attenuation for all audio streams.
+// This is to prevent clipping when two streams with low headroom (e.g. normalized & compressed) are playing.
+constexpr double SAMPLE_HEADROOM = 0.7;
+
 class ClientConnection;
 
 class BufferQueue : public RefCounted<BufferQueue> {
@@ -82,6 +87,10 @@ public:
         return -1;
     }
 
+    FadingProperty<double>& volume() { return m_volume; }
+    double volume() const { return m_volume; }
+    void set_volume(double const volume) { m_volume = volume; }
+
 private:
     RefPtr<Audio::Buffer> m_current;
     Queue<NonnullRefPtr<Audio::Buffer>> m_queue;
@@ -89,7 +98,9 @@ private:
     int m_remaining_samples { 0 };
     int m_played_samples { 0 };
     bool m_paused { false };
+
     WeakPtr<ClientConnection> m_client;
+    FadingProperty<double> m_volume { 1 };
 };
 
 class Mixer : public Core::Object {
@@ -100,8 +111,9 @@ public:
 
     NonnullRefPtr<BufferQueue> create_queue(ClientConnection&);
 
-    int main_volume() const { return m_main_volume; }
-    void set_main_volume(int volume);
+    // To the outside world, we pretend that the target volume is already reached, even though it may be still fading.
+    double main_volume() const { return m_main_volume.target(); }
+    void set_main_volume(double volume);
 
     bool is_muted() const { return m_muted; }
     void set_muted(bool);
@@ -122,7 +134,7 @@ private:
     NonnullRefPtr<Threading::Thread> m_sound_thread;
 
     bool m_muted { false };
-    int m_main_volume { 100 };
+    FadingProperty<double> m_main_volume { 1 };
 
     NonnullRefPtr<Core::ConfigFile> m_config;
     RefPtr<Core::Timer> m_config_write_timer;