diff --git a/src/app/browser/window/tab/item/page/content/text/markdown.rs b/src/app/browser/window/tab/item/page/content/text/markdown.rs
index 9e4d211b..3ffef743 100644
--- a/src/app/browser/window/tab/item/page/content/text/markdown.rs
+++ b/src/app/browser/window/tab/item/page/content/text/markdown.rs
@@ -19,8 +19,6 @@ use gtk::{
 };
 use gutter::Gutter;
 use icon::Icon;
-use reference::Reference;
-use regex::Regex;
 use sourceview::prelude::{ActionExt, ActionMapExt, DisplayExt, ToVariant};
 use std::{cell::Cell, collections::HashMap, rc::Rc};
 use syntax::Syntax;
@@ -53,7 +51,7 @@ impl Markdown {
         let hover: Rc<Cell<Option<TextTag>>> = Rc::new(Cell::new(None));
 
         // Init code features
-        let mut code = None;
+        //let mut code = None;
 
         // Init quote icon feature
         let mut is_line_after_quote = false;
@@ -76,6 +74,7 @@ impl Markdown {
 
         // Init new text buffer
         let buffer = TextBuffer::new(Some(&tag.text_tag_table));
+        buffer.set_text(markdown);
 
         // Init main widget
         let text_view = {
@@ -109,8 +108,15 @@ impl Markdown {
             t == 0 || t.is_multiple_of(2)
         };
 
+        // Parse in-line markdown tags
+        // * keep order!
+
+        reference::image_link(&buffer, &tag, base, &link_color.0, &mut links);
+        reference::image(&buffer, &tag, base, &link_color.0, &mut links);
+        reference::link(&buffer, &tag, base, &link_color.0, &mut links);
+
         // Parse single-line markdown tags
-        'l: for line in markdown.lines() {
+        /*'l: for line in markdown.lines() {
             if is_code_enabled {
                 use ggemtext::line::Code;
                 match code {
@@ -254,12 +260,7 @@ impl Markdown {
             // just append plain text covered in empty tag (to handle controller events properly)
             buffer.insert_with_tags(&mut buffer.end_iter(), line, &[&tag.plain]);
             buffer.insert(&mut buffer.end_iter(), NEW_LINE);
-        }
-
-        // Parse in-line markdown tags
-
-        image_link(&buffer, &tag, base, &link_color.0, &mut links);
-        link(&buffer, &tag, base, &link_color.0, &mut links);
+        }*/
 
         // Context menu
         let action_link_tab =
@@ -557,91 +558,6 @@ fn link_prefix(request: String, prefix: &str) -> String {
     format!("{prefix}{}", request.trim_start_matches(prefix))
 }
 
-/// Link
-fn image_link(
-    buffer: &TextBuffer,
-    tag: &Tag,
-    base: &Uri,
-    link_color: &RGBA,
-    links: &mut HashMap<TextTag, Uri>,
-) {
-    let start_iter = buffer.start_iter();
-    let end_iter = buffer.end_iter();
-    let full_content = buffer.text(&start_iter, &end_iter, true).to_string();
-
-    buffer.set_text("");
-
-    let mut last_pos = 0;
-    for cap in Regex::new(r"(?P<full_match>\[(?P<is_img>!|)?\[(?P<alt>[^\]]+)\]\((?P<img_url>[^\)]+)\)\]\((?P<link_url>[^\)]+)\))")
-        .unwrap()
-        .captures_iter(&full_content)
-    {
-        let full_match = cap.get(0).unwrap();
-        let before = &full_content[last_pos..full_match.start()];
-        if !before.is_empty() {
-            buffer.insert(&mut buffer.end_iter(), before);
-        }
-        if let Some(link) = Reference::parse(
-            &cap["link_url"],
-            None,
-            base,
-        ) {
-                link.into_buffer(buffer,
-                    link_color,
-                    tag,
-                    links)
-            }
-        last_pos = full_match.end();
-    }
-    let after = &full_content[last_pos..];
-    if !after.is_empty() {
-        buffer.insert(&mut buffer.end_iter(), after);
-    }
-}
-
-/// Link
-fn link(
-    buffer: &TextBuffer,
-    tag: &Tag,
-    base: &Uri,
-    link_color: &RGBA,
-    links: &mut HashMap<TextTag, Uri>,
-) {
-    let start_iter = buffer.start_iter();
-    let end_iter = buffer.end_iter();
-    let full_content = buffer.text(&start_iter, &end_iter, true).to_string();
-
-    buffer.set_text("");
-
-    let mut last_pos = 0;
-    for cap in Regex::new(r"(?P<is_img>!)\[(?P<text>[^\]]+)\]\((?P<url>[^\)]+)\)")
-        .unwrap()
-        .captures_iter(&full_content)
-    {
-        let full_match = cap.get(0).unwrap();
-        let before = &full_content[last_pos..full_match.start()];
-        if !before.is_empty() {
-            buffer.insert(&mut buffer.end_iter(), before);
-        }
-        if let Some(link) = Reference::parse(
-            &cap["url"],
-            if cap["text"].is_empty() {
-                None
-            } else {
-                Some(&cap["text"])
-            },
-            base,
-        ) {
-            link.into_buffer(buffer, link_color, tag, links)
-        }
-        last_pos = full_match.end();
-    }
-    let after = &full_content[last_pos..];
-    if !after.is_empty() {
-        buffer.insert(&mut buffer.end_iter(), after);
-    }
-}
-
 /// Header tag
 fn header(buffer: &TextBuffer, tag: &TextTag, line: &str, pattern: &str) -> Option<String> {
     if let Some(h) = line.trim_start().strip_prefix(pattern)
diff --git a/src/app/browser/window/tab/item/page/content/text/markdown/reference.rs b/src/app/browser/window/tab/item/page/content/text/markdown/reference.rs
index d9c5583d..b1b027b5 100644
--- a/src/app/browser/window/tab/item/page/content/text/markdown/reference.rs
+++ b/src/app/browser/window/tab/item/page/content/text/markdown/reference.rs
@@ -1,8 +1,16 @@
-use gtk::glib::{Uri, UriFlags};
+use super::Tag;
+use gtk::{
+    TextBuffer, TextIter, TextTag, WrapMode,
+    gdk::RGBA,
+    glib::{Uri, UriFlags},
+    prelude::{TextBufferExt, TextBufferExtManual},
+};
+use regex::Regex;
+use std::collections::HashMap;
 
-pub const REGEX_LINK: &str = r"\[(?P<text>[^\]]+)\]\((?P<url>[^\)]+)\)";
-
-pub const REGEX_IMAGE_LINK: &str =
+const REGEX_LINK: &str = r"\[(?P<text>[^\]]+)\]\((?P<url>[^\)]+)\)";
+const REGEX_IMAGE: &str = r"!\[(?P<alt>[^\]]+)\]\((?P<url>[^\)]+)\)";
+const REGEX_IMAGE_LINK: &str =
     r"\[(?P<is_img>!)\[(?P<alt>[^\]]+)\]\((?P<img_url>[^\)]+)\)\]\((?P<link_url>[^\)]+)\)";
 
 pub struct Reference {
@@ -11,7 +19,7 @@ pub struct Reference {
 }
 
 impl Reference {
-    pub fn parse(address: &str, alt: Option<&str>, base: &Uri) -> Option<Self> {
+    fn parse(address: &str, alt: Option<&str>, base: &Uri) -> Option<Self> {
         // Convert address to the valid URI,
         // resolve to absolute URL format if the target is relative
         match Uri::resolve_relative(
@@ -55,16 +63,15 @@ impl Reference {
             Err(_) => None,
         }
     }
-    pub fn into_buffer(
+    fn into_buffer(
         self,
-        buffer: &gtk::TextBuffer,
-        position: &mut gtk::TextIter,
-        link_color: &gtk::gdk::RGBA,
+        buffer: &TextBuffer,
+        position: &mut TextIter,
+        link_color: &RGBA,
         tag: &super::Tag,
         is_annotation: bool,
-        links: &mut std::collections::HashMap<gtk::TextTag, Uri>,
+        links: &mut HashMap<TextTag, Uri>,
     ) {
-        use gtk::{TextTag, WrapMode, prelude::TextBufferExtManual};
         let a = if is_annotation {
             buffer.insert_with_tags(position, " ", &[]);
             TextTag::builder()
@@ -94,25 +101,151 @@ impl Reference {
     }
 }
 
+/// Image links `[![]()]()`
+pub fn image_link(
+    buffer: &TextBuffer,
+    tag: &Tag,
+    base: &Uri,
+    link_color: &RGBA,
+    links: &mut HashMap<TextTag, Uri>,
+) {
+    let (start, end) = buffer.bounds();
+    let full_content = buffer.text(&start, &end, true).to_string();
+
+    let matches: Vec<_> = Regex::new(REGEX_IMAGE_LINK)
+        .unwrap()
+        .captures_iter(&full_content)
+        .collect();
+
+    for cap in matches.into_iter().rev() {
+        let full_match = cap.get(0).unwrap();
+
+        let start_char_offset = full_content[..full_match.start()].chars().count() as i32;
+        let end_char_offset = full_content[..full_match.end()].chars().count() as i32;
+
+        let mut start_iter = buffer.iter_at_offset(start_char_offset);
+        let mut end_iter = buffer.iter_at_offset(end_char_offset);
+
+        buffer.delete(&mut start_iter, &mut end_iter);
+
+        if let Some(reference) = Reference::parse(
+            &cap["img_url"],
+            if cap["alt"].is_empty() {
+                None
+            } else {
+                Some(&cap["alt"])
+            },
+            base,
+        ) {
+            reference.into_buffer(buffer, &mut start_iter, link_color, tag, false, links)
+        }
+        if let Some(reference) = Reference::parse(&cap["link_url"], Some("1"), base) {
+            reference.into_buffer(buffer, &mut start_iter, link_color, tag, true, links)
+        }
+    }
+}
+/// Image tags `![]()`
+pub fn image(
+    buffer: &TextBuffer,
+    tag: &Tag,
+    base: &Uri,
+    link_color: &RGBA,
+    links: &mut HashMap<TextTag, Uri>,
+) {
+    let (start, end) = buffer.bounds();
+    let full_content = buffer.text(&start, &end, true).to_string();
+
+    let matches: Vec<_> = Regex::new(REGEX_IMAGE)
+        .unwrap()
+        .captures_iter(&full_content)
+        .collect();
+
+    for cap in matches.into_iter().rev() {
+        let full_match = cap.get(0).unwrap();
+
+        let start_char_offset = full_content[..full_match.start()].chars().count() as i32;
+        let end_char_offset = full_content[..full_match.end()].chars().count() as i32;
+
+        let mut start_iter = buffer.iter_at_offset(start_char_offset);
+        let mut end_iter = buffer.iter_at_offset(end_char_offset);
+
+        buffer.delete(&mut start_iter, &mut end_iter);
+
+        if let Some(reference) = Reference::parse(
+            &cap["url"],
+            if cap["alt"].is_empty() {
+                None
+            } else {
+                Some(&cap["alt"])
+            },
+            base,
+        ) {
+            reference.into_buffer(buffer, &mut start_iter, link_color, tag, false, links)
+        }
+    }
+}
+/// Links `[]()`
+pub fn link(
+    buffer: &TextBuffer,
+    tag: &Tag,
+    base: &Uri,
+    link_color: &RGBA,
+    links: &mut HashMap<TextTag, Uri>,
+) {
+    let (start, end) = buffer.bounds();
+    let full_content = buffer.text(&start, &end, true).to_string();
+
+    let matches: Vec<_> = Regex::new(REGEX_LINK)
+        .unwrap()
+        .captures_iter(&full_content)
+        .collect();
+
+    for cap in matches.into_iter().rev() {
+        let full_match = cap.get(0).unwrap();
+
+        let start_char_offset = full_content[..full_match.start()].chars().count() as i32;
+        let end_char_offset = full_content[..full_match.end()].chars().count() as i32;
+
+        let mut start_iter = buffer.iter_at_offset(start_char_offset);
+        let mut end_iter = buffer.iter_at_offset(end_char_offset);
+
+        buffer.delete(&mut start_iter, &mut end_iter);
+
+        if let Some(reference) = Reference::parse(
+            &cap["url"],
+            if cap["text"].is_empty() {
+                None
+            } else {
+                Some(&cap["text"])
+            },
+            base,
+        ) {
+            reference.into_buffer(buffer, &mut start_iter, link_color, tag, false, links)
+        }
+    }
+}
+
 #[test]
 fn test_regex_link() {
-    let cap: Vec<_> = regex::Regex::new(REGEX_LINK)
+    let cap: Vec<_> = Regex::new(REGEX_LINK)
         .unwrap()
         .captures_iter(r#"[link1](https://link1.com) [link2](https://link2.com)"#)
         .collect();
 
     let first = cap.get(0).unwrap();
+    assert_eq!(&first[0], "[link1](https://link1.com)");
     assert_eq!(&first["text"], "link1");
     assert_eq!(&first["url"], "https://link1.com");
 
     let second = cap.get(1).unwrap();
+    assert_eq!(&second[0], "[link2](https://link2.com)");
     assert_eq!(&second["text"], "link2");
     assert_eq!(&second["url"], "https://link2.com");
 }
 
 #[test]
 fn test_regex_image_link() {
-    let cap: Vec<_> = regex::Regex::new(
+    let cap: Vec<_> = Regex::new(
         REGEX_IMAGE_LINK,
     )
     .unwrap().captures_iter(
@@ -120,12 +253,38 @@ fn test_regex_image_link() {
     ).collect();
 
     let first = cap.get(0).unwrap();
+    assert_eq!(
+        &first[0],
+        "[![image1](https://image1.com)](https://image2.com)"
+    );
     assert_eq!(&first["alt"], "image1");
     assert_eq!(&first["img_url"], "https://image1.com");
     assert_eq!(&first["link_url"], "https://image2.com");
 
     let second = cap.get(1).unwrap();
+    assert_eq!(
+        &second[0],
+        "[![image3](https://image3.com)](https://image4.com)"
+    );
     assert_eq!(&second["alt"], "image3");
     assert_eq!(&second["img_url"], "https://image3.com");
     assert_eq!(&second["link_url"], "https://image4.com");
 }
+
+#[test]
+fn test_regex_image() {
+    let cap: Vec<_> = Regex::new(REGEX_IMAGE)
+        .unwrap()
+        .captures_iter(r#"![image1](https://image1.com) ![image2](https://image2.com)"#)
+        .collect();
+
+    let first = cap.get(0).unwrap();
+    assert_eq!(&first[0], "![image1](https://image1.com)");
+    assert_eq!(&first["alt"], "image1");
+    assert_eq!(&first["url"], "https://image1.com");
+
+    let second = cap.get(1).unwrap();
+    assert_eq!(&second[0], "![image2](https://image2.com)");
+    assert_eq!(&second["alt"], "image2");
+    assert_eq!(&second["url"], "https://image2.com");
+}