Check PNG text chunks in ManYue: distinguish metadata encoding from drawn lettering

A PNG text chunk does not make drawn Chinese lettering searchable or guarantee that a reader displays metadata as dialogue. Prepare an original cover and external report, inspect text encoding and scope, then compare visible images in ManYue. Separate attached text, pixel lettering and application search capabilities, recording each result independently.

Make an image distinct from its description

Create an original four-hundred-by-three-hundred cover whose pixels say “灯下小册” (“Booklet under the lamp”), while attached text is planned as “原创测试，勿作正文” (“Original test, not body text”). Different strings identify the layer. Record size, markers and copy name and preserve the source. Do not draw metadata into pixels and call that metadata display. If writing text chunks is unavailable, leave them absent and use a paper report. Do not alter private comic notes or attribution as test material.

Check tEXt keyword and text encoding

Inspect tEXt if present, separating keyword from text rather than treating them as filename and extension. Its text uses Latin-1; arbitrary Chinese UTF-8 bytes cannot simply be inserted and labelled valid Chinese tEXt. Use the original English control Original sample and record actual output. Keywords have length and character restrictions, so do not place the whole description there. A decoded string alone needs documented tool behaviour or chunk evidence to establish encoding; visible English does not uniquely identify the chunk type.

Use iTXt to retain a Chinese description

iTXt can carry the Chinese description as UTF-8 text. Keyword, language tag, translated keyword and text have separate positions, rather than one undifferentiated encoded line. Use Description, zh-Hans and the original Chinese description only if supported externally. iTXt permits compressed or uncompressed text; record its flag and report, not image sharpness. An English keyword does not make the description non-Chinese, and the language tag is not a translation instruction.

Compare copies with unchanged pixels

Keep A without text chunks and B with verified iTXt from the same source, matching size and pixels and naming copies separately. Attached chunks can change file bytes and hashes without changing pixels. If resaving also alters pixels or colour information, separate those conditions instead of attributing all appearance changes to text. Claim only the checks supported by reports; without pixel comparison do not claim exact pixel equality. Similar covers are a limited observation.

Inspect pixel lettering on the reading page

Open readable samples through an existing image-collection or supported-archive path in ManYue and check markers and “灯下小册”. No extra description appearing in B does not prove chunk loss; displaying every metadata field as body text is not promised. Visible drawn Chinese does not establish iTXt parsing. If reading a report PDF, label it separately: Description visible in that report is not the original image page exposing the field. Unperformed steps remain unperformed.

Check search scope independently

If using an existing visible search control, identify its scope—shelf entries, chapters or otherwise—before testing the drawn title and attached phrase separately. Record actual results without assuming PNG-internal text search. Present metadata and indexed metadata are different facts; iTXt does not OCR drawn text. A missing hit means this search did not find it, not universal unreadability. Do not change filename, entry title and metadata simultaneously merely to force a hit and obscure its source.

Check garbling and field attribution

For garbled metadata, check type, encoding and tool decoding rather than assume a font change validates bytes. Distinguish filtered fields from absent chunks and use normal parsers for compressed text. Blurry drawn characters concern pixels and zoom rather than UTF-8. Share original samples and minimal reports without private notes. Preserve copies during conversions so removing provenance or attribution from the only file is recoverable.

Reopen with a four-layer text record

Reopen A and B at the same markers, check the visible title and confirm description, chunk and language externally. Retain filename, pixel lettering, attached text and actual search scope separately rather than combine them into a full-text-search result. Explain tEXt versus iTXt encoding and report reading versus native metadata display. Reinspect after exporter changes: chunks may be preserved, rewritten or removed, and one visible image cannot guarantee every later save preserves them.

Related guides and sources

Check PNG bit depth in ManYue: an eight-bit index differs from eight-bit RGB samples; ManYue PDF missing-glyph checks: separate appearance, font metadata and searchable text.

Checked on November 15, 2026: Apple public information lists ManYue 2.0.5, requiring iOS / iPadOS 15.0 or later. See the official App Store for published capabilities; verify controls in the installed version.

Method reference: W3C: PNG textual-information chunks. Materials and numbers are original exercises. Record external reports, paper reasoning and actual reading observations separately, without inventing missing conditions. Sources do not establish new automatic analysis by the app. No step-by-step physical-device, private-server or all-output-environment testing is claimed.