diff --git a/src/parsers/instagram.rs b/src/parsers/instagram.rs
index f41cca38..8d2b1560 100644
--- a/src/parsers/instagram.rs
+++ b/src/parsers/instagram.rs
@@ -13,7 +13,7 @@ use crate::Message;
 use crate::config::InstagramConfig;
 use crate::error::ChatpackError;
 use crate::parser::{Parser, Platform};
-use crate::parsing::instagram::{InstagramExport, parse_instagram_message};
+use crate::parsing::instagram::{InstagramExport, parse_instagram_message_owned};
 
 #[cfg(feature = "streaming")]
 use crate::streaming::{InstagramStreamingParser, StreamingConfig, StreamingParser};
@@ -64,10 +64,11 @@ impl InstagramParser {
         let export: InstagramExport = serde_json::from_str(content)?;
 
         let fix = self.config.fix_encoding;
+        // Use into_iter() with owned version to avoid allocations
         let mut messages: Vec<Message> = export
             .messages
-            .iter()
-            .filter_map(|msg| parse_instagram_message(msg, fix))
+            .into_iter()
+            .filter_map(|msg| parse_instagram_message_owned(msg, fix))
             .collect();
 
         // Instagram stores messages newest-first, reverse for chronological order
diff --git a/src/parsing/instagram.rs b/src/parsing/instagram.rs
index 55cb1655..8c1435d9 100644
--- a/src/parsing/instagram.rs
+++ b/src/parsing/instagram.rs
@@ -64,6 +64,13 @@ pub struct InstagramExport {
 /// // (actual mojibake text would be converted back to proper UTF-8)
 /// ```
 pub fn fix_mojibake_encoding(s: &str) -> String {
+    // Fast path: if all chars are ASCII, no mojibake fix is needed
+    // Mojibake only occurs with non-ASCII chars (codepoints > 127)
+    if s.is_ascii() {
+        return s.to_string();
+    }
+
+    // Slow path: check if we have mojibake (chars in Latin-1 range that decode to UTF-8)
     let bytes: Vec<u8> = s.chars().map(|c| c as u8).collect();
     String::from_utf8(bytes).unwrap_or_else(|_| s.to_string())
 }
@@ -73,7 +80,58 @@ pub fn parse_ms_timestamp(timestamp_ms: i64) -> Option<DateTime<Utc>> {
     Utc.timestamp_millis_opt(timestamp_ms).single()
 }
 
-/// Parses a raw Instagram message into a `Message`.
+/// Parses a raw Instagram message into a `Message` (owned version).
+///
+/// Takes ownership of the message to avoid allocations. Use this version
+/// when you can consume the `InstagramRawMessage`.
+///
+/// Returns `None` if the message has no content.
+///
+/// If `fix_encoding` is true, applies Mojibake fix to sender and content.
+pub fn parse_instagram_message_owned(
+    msg: InstagramRawMessage,
+    fix_encoding: bool,
+) -> Option<Message> {
+    // Get content from various possible locations (move, no clone)
+    let content = msg
+        .content
+        .or_else(|| msg.share.and_then(|s| s.share_text));
+
+    // Apply encoding fix if needed, but skip for ASCII (no mojibake possible)
+    let content = content.map(|c| {
+        if fix_encoding && !c.is_ascii() {
+            fix_mojibake_encoding(&c)
+        } else {
+            c // move, no allocation
+        }
+    });
+
+    // Skip messages without content
+    let content = match content {
+        Some(c) if !c.trim().is_empty() => c,
+        _ => return None,
+    };
+
+    let timestamp = parse_ms_timestamp(msg.timestamp_ms);
+
+    // Move sender_name or apply fix (avoids allocation when ASCII or fix_encoding=false)
+    let sender = if fix_encoding && !msg.sender_name.is_ascii() {
+        fix_mojibake_encoding(&msg.sender_name)
+    } else {
+        msg.sender_name // move, no allocation
+    };
+
+    Some(Message::with_metadata(
+        sender, content, timestamp, None, // Instagram doesn't have message IDs in export
+        None, // No reply references
+        None, // No edit timestamps
+    ))
+}
+
+/// Parses a raw Instagram message into a `Message` (reference version).
+///
+/// Use this version when you need to borrow the message (e.g., streaming).
+/// For better performance when ownership is available, use [`parse_instagram_message_owned`].
 ///
 /// Returns `None` if the message has no content.
 ///
@@ -82,15 +140,15 @@ pub fn parse_instagram_message(msg: &InstagramRawMessage, fix_encoding: bool) ->
     // Get content from various possible locations
     let content = msg
         .content
-        .clone()
-        .or_else(|| msg.share.as_ref().and_then(|s| s.share_text.clone()));
+        .as_ref()
+        .or_else(|| msg.share.as_ref().and_then(|s| s.share_text.as_ref()));
 
-    // Apply encoding fix if needed
+    // Apply encoding fix if needed, but skip for ASCII (no mojibake possible)
     let content = content.map(|c| {
-        if fix_encoding {
-            fix_mojibake_encoding(&c)
+        if fix_encoding && !c.is_ascii() {
+            fix_mojibake_encoding(c)
         } else {
-            c
+            c.clone()
         }
     });
 
@@ -102,7 +160,8 @@ pub fn parse_instagram_message(msg: &InstagramRawMessage, fix_encoding: bool) ->
 
     let timestamp = parse_ms_timestamp(msg.timestamp_ms);
 
-    let sender = if fix_encoding {
+    // Apply encoding fix if needed, but skip for ASCII (no mojibake possible)
+    let sender = if fix_encoding && !msg.sender_name.is_ascii() {
         fix_mojibake_encoding(&msg.sender_name)
     } else {
         msg.sender_name.clone()
diff --git a/src/parsing/mod.rs b/src/parsing/mod.rs
index 799c0d4e..aff313f4 100644
--- a/src/parsing/mod.rs
+++ b/src/parsing/mod.rs
@@ -20,7 +20,10 @@ pub mod discord;
 pub use telegram::{TelegramRawMessage, extract_telegram_text, parse_telegram_message};
 
 #[cfg(feature = "instagram")]
-pub use instagram::{InstagramRawMessage, fix_mojibake_encoding, parse_instagram_message};
+pub use instagram::{
+    InstagramRawMessage, fix_mojibake_encoding, parse_instagram_message,
+    parse_instagram_message_owned,
+};
 
 #[cfg(feature = "whatsapp")]
 pub use whatsapp::{