Last updated

Barcode and QR Code Reader

A photo of a barcode becomes its value. Free, no account.

VisionParse reads barcodes and QR codes out of a photograph free in your browser, with no account and no credit — several codes in one image all come back.

This deliberately does not use the AI engine. A decoder either resolves the bar pattern or it does not, and nearly every symbology carries a checksum that decides, so there is nothing to infer and nothing to invent. Six symbologies were tested in six photo conditions each — angled, blurred, halved in size, low contrast, glare across the label — and all 36 decoded correctly, with zero wrong values returned.

Last updated: 2026-08-06 · Measured by VisionParse engineering

  • Free · no account
  • JPG · PNG · WEBP · TIFF · BMP
  • 3 MB · every code in the image
  • Deleted within 1h
No image yetFree · no account

Drop a photo of a barcode or QR code

PNG, JPG, WebP, TIFF or BMP. EAN, UPC, Code 128, Code 39, ITF and QR are read; several codes in one photo all come back. No account, and the image is deleted straight after.

What this tool does, in numbers.

  • ACCURACY

    All 36 reads correct across six symbologies and six photo conditions, with zero wrong values returned.

  • SPEED

    A photograph decodes in tens of milliseconds, because no model is involved at any point.

  • ACCESS

    Free with no account: this runs on a classical decoder, so there is no credit to spend.

  • LIMIT

    Data Matrix and PDF417 are not read by our decoder — it returns nothing rather than guessing.

  • PRIVACY

    A checksum in the symbology means a partial read fails outright instead of returning a plausible wrong number.

  • RETENTION

    The image is deleted the moment the value is returned — a decode is finished when it answers.

Where the 36 of 36 reads, zero wrong values comes from.

Every figure on this page traces to a dated run on hardware we name. Nothing here is a vendor estimate.

Measured 2026-08-06

Claim
Every supported symbology read correctly in every photo condition tested, and nothing wrong was ever returned.
Method
zbar 0.23.93 on the VPS — a classical decoder, no AI credit. Full methodology.
Sample
EAN-13, UPC-A, Code 128, Code 39, ITF-14 and QR generated with known payloads, each put through six conditions: clean, photographed at an angle with perspective and hand blur, blurred, reduced to half size, contrast cut by 55, and a glare gradient across the label. Data Matrix and PDF417 were generated and tested too — both returned nothing.
Measured
2026-08-06 on Contabo Cloud VPS 8 — 8 vCPU AMD EPYC, 24 GB RAM, Ubuntu 24.04.4
Limitations
Generated symbols on white, not a crumpled label on a curved bottle under a shop light. Data Matrix and PDF417 are unsupported, and a code cropped or damaged past its error correction returns nothing at all rather than a partial value.
Contact
VisionParse@senithu.lk

What this tool gets wrong

HELD UP

Zero wrong values, which matters more than the decode rate. Every symbology here carries a check digit or error correction, so a partial read fails instead of returning something plausible. Across every condition tested the reader either produced the exact payload or produced nothing — and nothing is an answer you can act on, where a wrong barcode number quietly enters your stock system.

KNOWN LIMIT

Data Matrix and PDF417 are not supported. Both were generated and tested, and both returned nothing at all — our decoder does not read them. They are common on pharmaceutical packs and shipping labels, so if that is your case this tool cannot help yet, and saying so is cheaper than letting you find out with a real label.

KNOWN LIMIT

Our symbols were generated, not photographed off a shelf. Clean black on white, then degraded deliberately. A real label is curved, creased, wrapped around a bottle and lit by a shop ceiling. The conditions here approximate that and do not replace it — the honest read on this figure is that the decoder is sound, not that every shelf will behave.

Three steps, no account.

  1. Photograph the code

    JPG, PNG, WebP, TIFF or BMP. Angled and slightly blurred photos both decoded fine in testing.

  2. It decodes on our server

    A classical decoder, no AI and no credit. Several codes in one image all come back, each with its symbology.

  3. Take the value

    Copy any value with one click. The image is deleted as soon as the answer is returned.

What this tool accepts.

PropertyFree tool
Enginezbar 0.23.93 — classical decoding, free, no credit
Measured result36 of 36 reads, zero wrong values (2026-08-06)
Symbologies readEAN-13, EAN-8, UPC-A, Code 128, Code 39, ITF, QR
Not readData Matrix, PDF417 — tested, returns nothing
Multiple codesEvery code in the image is returned, each with its type
Input formatsJPG, PNG, WEBP, TIFF, BMP
AccessFree, no account, no credit
File retentionDeleted as soon as the value is returned

When to use the API instead.

This page converts one file now. To convert files from your own software, the same engine runs over HTTP from Starter upward.

# same engine, from your code
curl -X POST https://api.visionparse.app/v1/ocr \
  -H "x-api-key: $VISIONPARSE_KEY" \
  -F "file=@image.jpg"

→ 202 { "jobId": "c0df8e2c…" }

Use the API when

You convert more than a few files, convert on a schedule, or need results inside another system. Batch, webhooks and bounding boxes are API-only.

Frequently asked

Is this really free?

Yes, and for a reason worth stating: it runs on a classical decoder rather than the AI engine. A barcode has one correct answer that the bar pattern and its checksum decide, so paying a model to guess at it would cost you a credit to make a certain answer less certain.

What happens if the photo is bad?

You get nothing back rather than something wrong. Every symbology here carries a check digit or error correction, so a partial read fails outright — across all our test conditions the reader returned either the exact value or no value at all, never a plausible wrong one.

Which codes can it read?

EAN-13 and EAN-8, UPC-A, Code 128, Code 39, ITF and QR — all measured. Data Matrix and PDF417 are not supported: we generated both, tested both, and both returned nothing, so the page says so rather than letting you discover it with a real label.

Can it read several codes in one photo?

Yes. Every code found in the image is returned with its symbology, which is what a photo of a shipping label or a shelf usually contains.

Do you keep the image?

No. A decode is finished the moment it answers, so the file is deleted immediately rather than sitting out its usual hour.

Measured by VisionParse engineering. Method and raw data are published.Updated Benchmark · Method · Data handling