Two containers for words, and most diary strips put things in the wrong one about a third of the time. The rule underneath is simple enough to apply while you’re drawing, and that’s the only place a rule is any use.
A bubble is heard, a caption is told
A speech bubble happens inside the panel’s moment. The words are being said now, by somebody in the picture, and the reader is present while it happens.
A caption sits outside the moment. It’s the voice of somebody who already knows how the day ended, and in a diary strip that person is you, this evening, looking back at this morning.
Nearly every decision falls out of that.
The test that settles most panels
Ask whether you’d want the exact words in ten years.
“Mum, is the wifi broken or is it just me” is a bubble. Somebody said it, in those words, and the words are the thing worth keeping.
“Spent the afternoon on the phone to the bank” is a caption. Nobody said that. It’s a summary, and a summary in a bubble reads as a person delivering a report to an empty room.
Never say it twice
The commonest error by a distance: a caption reading “I told her I was fine” sitting above a bubble reading “I’m fine.”
Pick one. The caption is quieter, the bubble is sharper, and running both makes the reader read the same sentence twice and then wonder what they missed.
The same applies to a caption describing what the drawing already shows. “It was raining” over a panel where it’s raining is a caption doing nothing but taking up room the rain could have had.
Thought bubbles are mostly gone
The cloud with the trailing circles reads as 1965. Modern comics barely use it, and diary comics have least use of all for it, because a thought bubble puts the thinking inside the moment when the natural home for your thinking is the narrating voice.
Use a first-person caption instead. Give it a different look if you want, a tinted box or an italic, and then keep that style meaning the same thing for the rest of the archive. A visual convention that changes meaning between March and September is worse than not having one.
Placement decides the reading order
Bubbles get read left to right and top to bottom, before the reader takes in anything else in the panel. So the first person to speak goes upper left, every time. Get it the wrong way round and the reader takes the conversation backwards and can’t tell you why the panel felt off.
Tails point at whoever is speaking. Crossed tails are unreadable, so when two people need to swap sides, swap them between panels rather than inside one.
Leave the room before you draw. Letterers place bubbles first for a reason, and a bubble added at the end always ends up sitting on somebody’s head.
Keep bubbles off the panel border. Touching it reads as a mistake, breaking it reads as a mistake too, unless you do it consistently enough that it becomes a style.
The word budget
A four-panel strip has room for about twenty-five words. Go past that and the drawings start working for the text rather than the other way round.
That’s five or six words a panel on average, so most panels get one short bubble or one short caption, and not both. Fifteen words is a long bubble. Two lines is comfortable. If a moment needs three lines of dialogue, it usually needs two panels instead.
Leave a panel with nothing in it
At least one wordless panel per strip is a good discipline and it’s free.
A face with no text on it is the strongest thing available in the format, and the reader does the work you’d otherwise have spent six words on. It’s also the fastest fix for a strip that reads as cluttered: cover the words in one panel and check whether anything was lost.
You are the narrator and a character at once
The particular problem of a diary comic is that both voices belong to you, and they’ll blur if you let them.
Keep them doing separate jobs. The captioned you knows what happens next. The bubbled you doesn’t, and shouldn’t behave as though it does. Once your bubble self starts commenting on the day like a narrator, the panels have stopped earning their place and you’re writing a paragraph with pictures round it.
Read something you made last month with the words covered. If the pictures still tell you roughly what happened, the split is working. If they don’t, you’ve written a paragraph and illustrated it, which is a fine thing to make and doesn’t need panels at all.