Everything StegoPixel can do

Steganography hides the existence of a message, not just its contents. StegoPixel supports two carriers, and will also analyse and extract payloads that somebody else created.

What's included

• Optional encryption — AES-256-CBC with the key derived through PBKDF2 (SHA-256, 100,000 iterations).
• Optional error correction — a Reed–Solomon RS(255,223) code over GF(256) corrects up to 16 corrupted bytes per codeword, at roughly 14% payload overhead. It recovers from bit-level damage; it does not survive a lossy re-encode, and the app says so.
• Any file type — text, images, documents or archives all encode the same way.
• Detection and analysis — scan an image for a payload without extracting it, and inspect the carrier for signs of tampering.
• Capacity reporting — the available space for the chosen settings is shown up front, so a payload is never silently dropped.
• History — recent encode and decode operations are recorded so you can find what you did to which file.

Every image carries a second file inside it — the metadata. StegoPixel reads all of it, shows it grouped by source, and lets you change the parts that are actually writable.

Inspect

• EXIF, XMP and IPTC in a single grouped view, with a tag summary and per-source counts.
• Raw values — binary tags expose their hex representation for the tags where a decoded string is not the whole story.
• Anomaly detection — flags metadata that is internally inconsistent or out of place, which often indicates editing or a forged origin.

Edit

Editing is deliberately constrained, because most image formats can only have a small subset of tags written back reliably. StegoPixel exposes 19 typed writable EXIF tags — eight strings, three date/time, five numeric and two enumerated — and requires the value to match the tag's declared type rather than writing a raw string and hoping.

The eight string tags are the ones people reach for most often:

Custom key/value pairs are written as XMP in a dedicated urn:stegopixel:custom namespace, which keeps them namespaced and non-destructive. A 20-step undo history means a mistaken edit does not have to be repaired by hand.

Repeat

• Templates — save a set of metadata values once and reapply it across a shoot or a project.
• Validation — check values for consistency and get a report of what is wrong.
• Export — push the whole tag table out to JSON, CSV or HTML, including raw hex.
• Batch across folders — for many files at once, drive the metadata nodes from the workflow builder. The workbench itself is built around one image at a time.

Photos leak more than people expect: where you live, where you work, which device you own and when you were there. Privacy cleanup finds that data and removes it.

• Risk scan — a single scan rates what it found as high, medium or low risk and counts the offending fields.
• GPS removal — strips location tags, the single most identifying piece of metadata in a photo.
• Personal field removal — camera serials, author names, software history and the rest of the identifying set.
• Cleanup profiles — presets for different threat models, from “prepare for web sharing” to full anonymization, plus a minimal mode that keeps GPS on purpose.
• Custom sensitive fields — extend the detection set with your own field names.
• Save-copy default — cleaning writes a new file and leaves the original untouched unless you say otherwise.

A manipulation leaves traces in the statistics of an image long after the visual evidence is gone. StegoPixel gives you the standard instruments for finding them, each producing a result image you can zoom, compare and export.

Results are not just images. The file inspector keeps the digests — SHA-256 and SHA-512 — alongside the findings, and the whole session can be exported as a report.

Enhancement

Forensics is often about making something faint visible. Eight enhancement modes sit alongside the analysis panels: CLAHE, gamma correction, level adjustment, unsharp masking, local contrast, two auto-contrast variants and histogram equalisation — all applied non-destructively to the working copy.

Recipes

Rather than configuring eleven panels by hand, six built-in recipes chain the relevant analyses for a particular question — a quick tamper check, a deep scan, an AI-origin review — and collect the results into one summary.

Seven independent signals, combined into one verdict. All of them run on your machine — and they are hand-written statistical heuristics, not a trained neural network. There is no model file, no cloud inference and no labelled dataset behind them. That is precisely why the app reports confidence tiers rather than certainty.

Every verdict also lists which checks ran, were skipped or failed. A check that could not complete is reported as incomplete — never quietly treated as clean.

The verdict

Results resolve to one of three tiers, shown alongside the contributing signals and their individual confidences:

“Inconclusive” is a real outcome, not a failure. An offline detector that never admits uncertainty is not being honest with you.

Separately, the Home module produces an authenticity score for a loaded image — a weighted total that subtracts for high, medium and low-risk findings, hidden payloads, AI indicators and unverifiable provenance, and resolves to “Likely Authentic”, “Uncertain” or “Suspicious”. It is a summary of everything the app measured, with the breakdown shown so you can see which signal moved the number.

Finding a region of an image that has been copied from elsewhere in the same image — or a near-identical second file — is a copy-move problem, not a filename problem. StegoPixel compares image content, not file names or sizes.

• Copy-move detection — a block-level similarity engine with four matching algorithms, union-find clustering of matched regions, and displacement-vector filtering to discard coincidental texture repeats.
• Near-duplicate search — perceptual hashes (average, difference and DCT-based) find resized, recompressed and lightly edited copies that a byte comparison would miss.
• Image comparison — a side-by-side view with a sensitivity threshold, a red/green difference overlay, a changed-pixel count, a difference percentage and mean absolute error.
• Progressive folder scanning — a folder loads in batches with live counters rather than freezing the window, and a thumbnail strip with a worker-pooled cache keeps it responsive.
• Recent files and folders — reopen what you were last working on.
• Persisted workflow state — saved workflows and their full execution history live in a local SQLite database, with configurable retention, auto-vacuum and a size warning threshold.

Content hashes are SHA-256 and SHA-512; similarity uses average, difference and perceptual hashes.

Sometimes the only way to be sure is to look at the bytes.

• Raw binary inspection of any file, with a monospaced offset/hex/ASCII layout built for reading structure rather than just scrolling.
• Structure parsing — the file's segment and chunk layout, so you can see where metadata and payload regions actually live.
• Format detection — confirms what a file really is, independent of its extension.
• Byte interpretation — read integers, floats and timestamps out of the bytes under the cursor.

A finding you cannot hand to anybody else is not much of a finding. Everything StegoPixel discovers can leave the app in a form somebody else can read.

Reports are produced by the same engine the workflow system uses, so a report generated by hand and one generated by an automated pipeline contain the same information.

• Dark and light themes — designed as a matched pair, not one theme with the colours inverted.
• 12 interface languages — English, German, Spanish, French, Italian, Japanese, Polish, Portuguese, Russian, Simplified Chinese, Hindi and Arabic, switchable at runtime without a restart. Arabic switches the layout to right-to-left.
• Configurable hashing — choose which digests are computed for integrity reports.
• Sensible defaults — backup before destructive operations, warning modes for overwrites, and a thumbnail size preference, all adjustable.
• Notification centre — long operations report back without modal interruptions.
• Local logs — written to your user profile for troubleshooting, never transmitted.

Capabilities are format-dependent, and knowing which is which saves a lot of confusion. This matrix reflects what the current build actually does.

HEIC/HEIF, AVIF and camera RAW formats (CR2, CR3, NEF, ARW, RAF, DNG, ORF, RW2, PEF, SRW) appear in the format catalogue, but they cannot be opened — the image pipeline decodes only the six raster formats above. StegoPixel will not quietly pretend otherwise: an unsupported file is reported as unsupported rather than partially analysed.

Two more format facts worth knowing before you start:

• Pixel-embedded output is always written as PNG. PNG is lossless, so it is the only output container that preserves an LSB payload. Saving an encoded result over a .jpg path raises a format-mismatch warning rather than silently destroying the payload.
• Conversion targets exclude GIF. You can read GIF, but not write it — output formats are JPEG, PNG, TIFF, WebP and BMP.

Capacity, honestly

A metadata channel targets 44,984 bytes in JPEG, PNG, TIFF and WebP — and that figure is explicitly a typical maximum, not a guarantee. The app's own documentation notes that external EXIF readers may stop honouring metadata profiles beyond 64 KB, which is why the ceiling sits below it. It also depends on the image, the tags already present and the payload's own overhead.

Pixel capacity scales with resolution, the number of channels and the bit depth you select. Rather than estimate, the app shows the real available capacity for the specific file and settings before you encode, and refuses rather than dropping a payload that does not fit.