diff --git a/config_help.md b/config_help.md
index 276cf8e..8bf06dd 100644
--- a/config_help.md
+++ b/config_help.md
@@ -42,6 +42,7 @@
 
 ```
 # Read the entire video and audio input (defaults).
+
 [a]
 [v]
 ```
@@ -49,6 +50,7 @@
 ```
 # Segment the video and audio, skipping irrelevant parts at the start and
 # end, and five minutes in the middle.
+
 [a] 1:35 25:00 30:00 54:27
 [v] 0:17 23:42 28:42 53:09
 ```
@@ -58,6 +60,7 @@
 # filler.wav. Split video into two corresponding segments, separated
 # by a filler frame generated by repeating the last frame of the first
 # video segment for the same duration as filler.wav.
+
 [a] 1:53 23:15
 [a:filler.wav]
 [a] 49:42 50:25
@@ -73,6 +76,7 @@
 # to the punch in time of the first frame, the punch out time of the
 # second-last frame, the punch in time of the last frame, and the punch
 # out time of the last frame, respectively.
+
 [a:audio.wav] 7:59 25:50 53:00 53:27
 [f:slide-000.jpg] 7:59 8:46
 [f:slide-001.jpg] 8:46 10:28
@@ -92,6 +96,7 @@
 
 ```
 # Extract slide images from a PDF and merge with the recorded audio.
+
 [a] 0:04 26:33
 [f:slides.pdf:0] 0:04 3:39
 [f:slides.pdf:1] 3:39 9:14
@@ -106,3 +111,22 @@
 [f:slides.pdf:10] 24:07 26:17
 [f:slides.pdf:11] 26:17 26:33
 ```
+
+```
+# The poor man's text overlay :). If you make a mistake and want to
+# include a correction, you can just pull the relevant slide from the
+# original PDF, annotate it as necessary, then insert that at the
+# appropriate point in the final render. You could also use a similar
+# technique to insert new audio, although that'll be trickier to get
+# the timing right.
+
+[a] 06:39 33:42
+[a:joiner.wav]
+[a] 54:01 54:12
+
+[v] 03:36 28:52
+[f:slide_11.pdf:0] 28:52 29:02
+[v] 29:02 30:39
+[f:^:last] @joiner.wav
+[v] 50:58 51:09
+```