Code 128 Barcode Generator

Guide: Code 128 code sets and the Code 39 charset walks through this analysis character by character, with the full Code 39 alphabet beside it. All guides are listed at /guides/.

The three code sets, in plain terms

Code 128 is really three alphabets sharing one set of bars, and the encoder may switch between them mid-symbol:

• Code set A — uppercase letters, digits, punctuation, and the ASCII control characters (tab, carriage return…). Control characters are the only thing A has that B lacks, and they are what force a symbol into A.
• Code set B — uppercase, lowercase, digits, and punctuation. Any lowercase letter in your input forces set B, because A simply has no lowercase.
• Code set C — digits only, packed two per symbol character. This is the compression trick: a run of digits in set C takes half the width it would in A or B, which is why an encoder jumps to C for runs of four or more.

The analyzer behind the readout belongs to this site's verification layer, not to the renderer. It records four facts about your text — whether every character is printable ASCII and which are not, whether a lowercase letter forces code set B, whether the input is digits only with an even count (pure code set C), and the longest run of consecutive digits — and writes one sentence of guidance. The methodology page fixes the division of labour: the analysis of "which characters force set A or B, when set C compression applies" is offered "as validation and advice; the actual code-set optimization inside the symbol is the renderer's, following the same rules."

Which characters force what is probed, not typed: at build time this page asks the analyzer about every single-character input from code point 0 to 255. It accepts 95; of those, 26 raise the lowercase flag that forces set B, 10 count toward a digit run, and the other 59 — uppercase letters, space, punctuation — force nothing. Control characters, the only thing that would force set A, lie outside the accepted range.

The sentences the readout can say

For an accepted input the analyzer picks one of 5 sentences, tested in a fixed order — even digit count; odd digit count; lowercase present; uppercase text with a digit run long enough to pay for a switch to set C; anything else. One fictional input per branch:

When an input qualifies twice, the order decides: the fictional lot-20261005 has a lowercase letter and a run of 8 digits, and gets the lowercase sentence. Two more fictional inputs show that a run is an unbroken stretch: in 12-345-6789 the longest is 4, because a hyphen ends one; in 1a2b3c it is 1, because every letter resets the count. The odd-count sentence recommends a leading zero because set C carries digits in pairs, but the analyzer never inserts it: a padded value is a different value. The guidance predicts from the published rules; it does not describe where the renderer switched sets.

What the validator does with your text, in order

Every generator here runs the same four steps before the renderer is called — normalize, enforce, compute or verify, report — and for Code 128 each has an unusual shape.

Normalize — nothing. Code 128 is the one symbology here whose validator changes no character. The EAN family strips spaces and hyphens, and Code 39 uppercases and says so; Code 128 does neither, because every printable character, space included, is data. The fictional input " padded " leaves validation as a payload of 8 characters.

Enforce — printable ASCII and a length cap. The accepted range runs from code point 32, the space, through 126, the ~. The cap is probed the same way, with ever-longer input: 80 characters are accepted and one more is refused — this tool's rule, not the symbology's.

Compute or verify — nothing visible. Code 128 carries no check digit you can type, so the checkDigit field of the validated payload is absent and the cross-check that re-derives the check digit of any validated payload carrying one passes trivially, returning true. The symbol's own mod-103 check character is the renderer's, computed inside the bars and never shown; anything you type is data.

That has a consequence worth seeing. Take the cited published EAN-13 vector 5901234123457 and a copy with its last digit changed, 5901234123458. The EAN-13 validator verifies the first and refuses the second: “Check digit mismatch: for 590123412345 the final digit must be 7, but you entered 8. Fix the number, or enter only the first 12 digits to have the check digit computed.” The Code 128 validator accepts both with the odd-count sentence above, because to it the digits are text. A GTIN's check digit is verified in the EAN, UPC, or ITF-14 generator, and encoding a product number here does not allocate or license it — see Do I need a GS1 barcode?

Report — one note. For the generator's own default example input BG-2026-0001 (not a published vector) the validator returns the payload unchanged and exactly one note, the guidance: "Uppercase text with digit runs — code set B (or A) for the text, switching to code set C where four or more digits run consecutively."

What this tool deliberately refuses

Control characters and accented text are rejected with an explanation, not encoded. Full Code 128 can carry both (sets A and FNC4 respectively), but FNC4 in particular is honored inconsistently by real-world scanners — the same label can read as café on one device and café on another. A generator that hands you that label without warning is not doing you a favor. If you need non-ASCII text scanned reliably, use a QR code, which carries UTF-8 by design.

The validator's real messages, each on a fictional input:

How to read them: the first names each offending character once, in order of first appearance, and a control character by its code point rather than an invisible glyph. The character check runs before the length check, and nothing is repaired — what the symbology cannot carry is reported, never rewritten.

Verified before you download it

After validation, the actual SVG shown on the page is rasterized and decoded by ZXing, an engine independent from JsBarcode, and the preview and download controls stay withheld under one of 2 reasons the verification code names explicitly:

• decode-failed — when the independent reader cannot read the rendered bars.
• payload-mismatch — when it reads a different text from the validated payload.

For Code 128 the comparison is exact — UPC-A and EAN-13 share one leading-zero equivalence (the same printed symbol under two names), but Code 128 has no alias — so on the padded example a decoded reading of the same text returns true and a reading with both spaces stripped returns false, withholding the barcode. What none of it can verify is print quality: no browser check can certify ink, substrate, scaling, contrast, or quiet zones after printing, so a physical production sample must still be scanned. See How the verifier decodes what you see .

What the ZPL download carries

The ZPL download is a text file of native printer commands, not an image. For the validated payload BG-2026-0001 the export engine's first note reads: “Subset selection is left in automatic mode — the printer picks Code 128 subsets.” One transport limit: ^ and ~ are ZPL control characters and cannot travel inside the field data, so although the validator accepts the fictional AB^CD, the ZPL export refuses it: “Code 128 ZPL export cannot carry "^": ^ and ~ are ZPL control characters inside ^FD data. This is a ZPL transport limitation, not a barcode limitation — the symbology itself can encode them, but this export cannot transmit them safely.” See the ZPL and EPS export guide.

Quiet zones and sizing

Like every linear barcode, Code 128 needs light margins on both sides; the downloads include a margin of ten module widths on each side, and you should preserve it in your artwork. Width grows with content — around 11 modules per character plus fixed overhead — so long payloads make wide symbols. If a label keeps failing to scan, the first three suspects are trimmed quiet zones, too-small printing, and low contrast, in that order.

Code 128 and Code 39: the same text, two validators

The nearest alternative for alphanumeric data is Code 39, whose fixed alphabet has 43 characters and no lowercase. The fictional serial-x passes this page's analyzer with its payload untouched; through the Code 39 validator the payload becomes SERIAL-X, with the change disclosed in a note: "Code 39 has no lowercase letters — your input was converted to uppercase." Neither validator guesses — one records a fact, the other changes the text and says so. The guide Code 128 code sets and the Code 39 charset lays both alphabets side by side.

Code 128 FAQ

Verify before production: the generator independently decodes its on-screen SVG and PNG export, but it cannot inspect your printer. Scan the physical sample at final size, or recheck final artwork with the Barcode Verifier. Everything runs locally; see the methodology page. For database merges, industrial printers, or high-volume workflows, use the barcode software decision guide — and if you print on Zebra-family thermal printers, the ZPL printing guide covers the native ZPL download.