Caption Length Counter

Where the caption gets cut, and why it matters

Instagram gives every caption the same 2,200-character ceiling. Go past it and the app refuses to post — there is no silent truncation, you simply cannot publish until you cut. That limit is generous enough that most people never meet it. The counter turns amber at 2,000 to give you warning, and red past 2,200.

The limit that actually shapes what you write is the other one: the preview. In the feed, a caption sits beneath the media in a collapsed state. Instagram shows a short opening fragment, appends “… more”, and hides the rest behind a tap. The widely used planning figure for that boundary is 125 characters, and it is what the counter above marks.

Be clear about what that number is. Instagram has never published the truncation rule, and the real cut is not a fixed character count — it is a line count measured after the text has been laid out. It moves with screen width, system text size, the language you are writing in, whether you used emoji, and whether you inserted a line break. On a large phone with small text you get more; with accessibility text sizes you get considerably less. 125 is a reliable planning convention, not a specification.

Why the first line does all the work

• The tap is a cost. Expanding a caption is a deliberate action taken while scrolling, and most people scrolling do not take it. Anything below the fold is written for the minority who already care.
• The preview travels. The opening fragment is what appears in notifications and shared previews. It is the only part of your caption with reach beyond the post.
• Instagram search reads caption text. Search matches keywords in captions, not only hashtags, so plain specific words near the start are doing work beyond persuasion. Instagram has not published how heavily it weighs caption position, so treat “front-load your keywords” as sensible practice rather than a known rule.

Every limit this tool tracks

Emoji are not one character in any consistent sense — a flag is two regional indicator codepoints, and a family emoji can be five or more joined by invisible connectors. This counter measures codepoints, which is how the limit behaves for ordinary text, but it means an emoji-dense caption can show a different number here than in another app. Near the ceiling, cut more than you think you need.

The hook analyser

The tool takes the first 125 characters and runs five structural checks over them. Each one that fires gives you what is wrong, why it costs you, and a concrete rewrite built from your own words — not a generic tip.

• It opens with a greeting. “Hey everyone!” spends the one visible line on something nobody scrolled for. The suggested fix quotes the sentence that comes after the greeting and tells you to start there instead.
• The point arrives late, or the line opens on a filler word. When the sentence has a comma sitting well into it, or begins with a setup clause or a word like so, honestly, basically, the tool offers the halves swapped round — the part after the comma first, the setup afterwards where it costs nothing.
• It is a question with nothing riding on it. A short open question with no number and no stakes in it usually gets ignored, because ignoring it costs the reader nothing. The fix suggests giving the question a consequence or answering it outright.
• The first line runs past the cut-off — or is close enough to it that a different screen width will do the cutting for you. Over 125 characters it is flagged red with the exact point to split at; over 100 it is amber, because your margin is thin.
• There is nothing concrete in it. No number, no name, no place, no price, no span of time. Generic openers read as filler even when the post underneath is good. This one only fires once the hook is at least six words long, so a deliberately short hook is not nagged at.

When nothing fires you get a clean verdict, phrased carefully: no structural problem was found, but whether the hook is interesting is still your call. That is the honest boundary of what an offline check can do. It can tell you the hook is being wasted on a greeting, buried after a comma or cut off mid-sentence. It cannot tell you whether anybody will care.

How it reads

The readability panel is built from exactly two measurements: the average number of words per sentence, and the share of your words that are seven letters or longer. Both are shown as raw numbers alongside the sentence and word counts.

There is deliberately no score and no named readability index. Those formulas were built for printed prose and validated on schoolbooks, not on captions with emoji and line breaks in them, and putting a single number on a caption would suggest a precision that is not there. What you get instead is a plain verdict — quick scan, normal or dense — and the two numbers it came from, so you can disagree with it.

Short sentences and short words scan quickly on a phone. That is the whole theory. If a caption comes out dense and you meant it to, ignore the verdict; if you did not, the two numbers say which thing to fix.

Comparing two drafts

Paste a second version into the Draft B box and a comparison table appears: characters, how much is hidden after the marker, first-line length, how many hook issues each version has, words, words per sentence, share of long words, the readability verdict, caption hashtags and emoji. The full report above stays focused on whichever draft you have selected, so you can switch between them without losing the comparison.

The table does not pick a winner, and the tool says so where you can see it. It knows which caption is shorter, which gets to the point sooner and which has fewer structural problems in the visible line. Whether your audience prefers the warmer one is a judgement it has never been in a position to make.

Saved drafts

Naming a draft saves both caption boxes and the hashtag block together, up to twenty of them, and saving under an existing name overwrites it. They live in local storage in this one browser: they do not sync to your phone, no account is involved, nothing is uploaded, and clearing your site data deletes them. The page makes no network requests at all, so there is nowhere for them to go. If your browser blocks local storage, the tool tells you the save failed rather than losing your work quietly.

Hashtags, mentions and the words you actually used

Everything in the caption gets pulled apart into chips: the words you repeated most, hashtags written inside the caption, hashtags in the separate block, and @ mentions. It is a quick way to see whether the caption says the thing you think it says.

The counting is where it earns its place. Caption tags and block tags are added together against the 30-tag limit, and duplicates are found across both. A tag that appears in the caption and again in the block does not count twice — it simply burns one of your thirty slots for nothing, and the tool names each one. Go over thirty and it tells you exactly how many to remove.

The move hashtags to first comment button copies the unique tags to your clipboard and shows the caption with them stripped out, without touching your caption box. Whether the first comment is better than the caption is not something Instagram has ever published a difference on; do it because you want a cleaner caption, not because of folklore. Building the set itself is the hashtag builder‘s job.

Protecting the paragraph breaks

Instagram keeps caption line breaks in most places, but blank lines still get trimmed when you go back and edit a live caption, or when some keyboards paste. The protect the paragraph breaks toggle replaces each blank line with a U+2800 braille blank — the same trick the bio line break generator uses — which survives trimming because it is a printing character rather than whitespace. It costs one character per gap, and the tool shows you both counts so you can see what the spacing bought.

How to write to the boundary

• Make the first sentence complete. If the cut lands mid-thought, a reader who does not tap gets nothing.
• Do not open with hashtags, credits or an emoji run. They spend the only characters with guaranteed reach on information nobody is scanning for.
• Say the specific thing. “Three mistakes that cost me a wedding booking” travels; “Some thoughts…” does not.
• Put the call to action twice in a long caption — once visible, once at the end. Six ready-made lines are in the picker, and they are starting points to rewrite in your own voice.
• Break the body into short paragraphs. Line breaks below the fold cost almost nothing.
• Avoid styled Unicode characters. Caption text is searchable, so substituting characters removes it from that index. Save the font generator for decoration.

The honest downside

Character counting is the most measurable part of writing a caption and the least important. A caption trimmed to exactly 125 characters with a keyword in the first three words is still a bad caption if the opening line is not worth reading, and the hook analyser is checking structure, not quality. It can see that you opened on “Hey guys”. It cannot see that your idea is boring.

There is also a lot of confident advice about optimal caption length — that short captions get more reach, or that long ones signal quality. Instagram has not published anything supporting either claim, and the studies circulated in support are usually observations of what popular accounts happen to do rather than evidence that length caused the outcome. Write the length the idea needs. If you want to know what is working on your own account, your Insights are better evidence than any generator; the analytics guide covers how to read them, and the captions, bios and hashtags guide collects the limits themselves.

Finally, the 125-character marker is a convention, and treating it as exact leads to over-engineering. Nudging a hook from 128 characters to 124 is not a meaningful optimisation when the real cut moves by twenty characters depending on the reader’s phone. Use the marker to notice that your hook is 300 characters long, not to shave four off a good one.

Questions people actually ask