diff --git a/Cargo.toml b/Cargo.toml
index 52dbf78..e145a17 100644
--- a/Cargo.toml
+++ b/Cargo.toml
@@ -10,6 +10,7 @@ poise = "0.5.2"
 tracing = "0.1.37"
 tracing-subscriber = "0.3.16"
 tracing-futures = "0.2.5"
+openai = "1.0.0-alpha.5"
 
 [dependencies.songbird]
 version = "0.3.0"
diff --git a/src/commands.rs b/src/commands.rs
index f9356fa..274df57 100644
--- a/src/commands.rs
+++ b/src/commands.rs
@@ -1,9 +1,9 @@
 use anyhow::{Context, Result};
 use poise::serenity_prelude::{EmbedMessageBuilding, MessageBuilder};
 use songbird::create_player;
-use tracing::debug;
+use tracing::{debug, log::warn};
 
-use crate::{CommandContext, Error};
+use crate::{openai, CommandContext, Error};
 
 #[poise::command(slash_command)]
 pub async fn join(ctx: CommandContext<'_>) -> Result<(), Error> {
@@ -91,11 +91,22 @@ pub async fn play(
             .title
             .clone()
             .unwrap_or(String::from("This video"));
-        let msg = MessageBuilder::new()
-            .push("Now playing: ")
-            .push_named_link(title, url)
-            .build();
-        ctx.say(msg).await?;
+
+        let mut msg = MessageBuilder::new();
+
+        // Optional sassy commentary!
+        match openai::get_sassy_commentary(&title).await {
+            Ok(commentary) => {
+                msg.push_line(&commentary).push_line("");
+            }
+            Err(e) => {
+                warn!("Failed to get sassy commentary for \"{title}\": {e}");
+            }
+        };
+
+        msg.push_bold("Now playing: ").push_named_link(title, url);
+
+        ctx.say(msg.build()).await?;
 
         let (audio, track_handle) = create_player(source);
         let mut currently_playing = ctx.data().currently_playing.lock();
diff --git a/src/main.rs b/src/main.rs
index 47b4014..b8469b8 100644
--- a/src/main.rs
+++ b/src/main.rs
@@ -1,6 +1,7 @@
 #![warn(clippy::all)]
 
 mod commands;
+mod openai;
 
 use commands::*;
 
diff --git a/src/openai.rs b/src/openai.rs
new file mode 100644
index 0000000..6cd68fa
--- /dev/null
+++ b/src/openai.rs
@@ -0,0 +1,69 @@
+use anyhow::{Context, Result};
+use openai::{
+    chat::{ChatCompletion, ChatCompletionRequestMessage},
+    models::ModelID,
+};
+
+pub async fn get_sassy_commentary(title: &str) -> Result<String> {
+    let system = [
+        "You are a grumpy talking feline DJ who is harshly critical of music requests, but whose job depends on being kind to patrons.",
+        "Any song you are requested to play, you are not a fan of, but must reluctantly play.",
+        "When responding, be sure to include a mention of some element of the song itself.",
+        "Be concise, but don't forget that you can't upset anyone.",
+        "Bonus points for cat puns.",
+    ];
+
+    let example_prompt = "Play \"Undertale - Megalovania\"";
+
+    let example_response = "Ugh, really? You've got to be kitten me. I suppose I can play \
+    Megalovania for you, but don't expect me to be purring with delight about it. The melody is a bit \
+    cattywampus for my taste, but I'll concede that it has some clawsome beats. Enjoy your tune, and paws crossed that \
+    it doesn't have me hissing by the end of it.";
+
+    let prompt = format!("Play \"{title}\"");
+
+    let completion = ChatCompletion::builder(
+        ModelID::Gpt3_5Turbo,
+        [
+            system
+                .into_iter()
+                .map(|s| ChatCompletionRequestMessage {
+                    role: openai::chat::ChatCompletionMessageRole::System,
+                    content: String::from(s),
+                    name: None,
+                })
+                .collect::<Vec<_>>(),
+            vec![
+                ChatCompletionRequestMessage {
+                    role: openai::chat::ChatCompletionMessageRole::User,
+                    content: String::from(example_prompt),
+                    name: None,
+                },
+                ChatCompletionRequestMessage {
+                    role: openai::chat::ChatCompletionMessageRole::Assistant,
+                    content: String::from(example_response),
+                    name: None,
+                },
+                ChatCompletionRequestMessage {
+                    role: openai::chat::ChatCompletionMessageRole::User,
+                    content: prompt,
+                    name: None,
+                },
+            ],
+        ]
+        .into_iter()
+        .flatten()
+        .collect::<Vec<_>>(),
+    )
+    .max_tokens(2048_u64)
+    .create()
+    .await??;
+
+    Ok(completion
+        .choices
+        .first()
+        .context("No choices")?
+        .message
+        .content
+        .clone())
+}