revelo

KO | EN

gitlite — search

#javascript #python #hacktoberfest #react #ai #typescript #llm #go #golang #android #machine-learning #rust #deep-learning #linux

(★ 28)

A safe, fast Rust port of MediaInfoLib — reads technical and tag metadata from 177 audio, video, image, and container formats. Drop-in libmediainfo replacement (C ABI + CLI), differential-tested byte-for-byte against the mediainfo oracle.

revelo 최신버젼 다운로드

최종 버전 다운로드 (.zip)

// repository documentation

Was this content helpful?

★ 0 (0 ratings)

# Revelo

![revelo — read technical metadata from any media file, in pure Rust](https://raw.githubusercontent.com/vbasky/revelo/1dcc7cd/docs/banner.png)

A library and CLI for containers, audio & video codecs, image formats, subtitle
streams, archives, and embedded tags. A clean-room port of **MediaInfoLib**,
validated byte-for-byte against the C++ `mediainfo` oracle — plus an optional
**ExifTool-grade** maker-note layer, validated tag-for-tag against the `exiftool`
binary. Two reference tools' worth of metadata, in one pure-Rust crate.

**Name:** *Revelo* is Latin *to reveal/unveil* (and a backronym: **R**eveals **E**very **V**ideo & **E**ncoding **L**ayer **O**utput) — it reveals the technical metadata hidden inside media files.

[![crates.io](https://img.shields.io/crates/v/revelo?logo=rust&color=orange)](https://crates.io/crates/revelo)
[![Downloads](https://img.shields.io/crates/d/revelo?color=orange&label=downloads)](https://crates.io/crates/revelo)
[![docs.rs](https://img.shields.io/docsrs/revelo-core?logo=docsdotrs)](https://docs.rs/revelo-core)
[![CI](https://img.shields.io/github/actions/workflow/status/vbasky/revelo/ci.yml?branch=main&logo=github&label=CI)](https://github.com/vbasky/revelo/actions)
[![License](https://img.shields.io/crates/l/revelo-core)](https://raw.githubusercontent.com/vbasky/revelo/main/LICENSE)
[![MSRV](https://img.shields.io/badge/MSRV-1.88-blue)](https://www.rust-lang.org)
[![Parsers](https://img.shields.io/badge/parsers-180-blueviolet)](https://raw.githubusercontent.com/vbasky/revelo/main/docs/formats.md)
[![Read the write-up](https://img.shields.io/badge/Medium-Read%20the%20write--up-black?logo=medium&logoColor=white)](https://medium.com/@vbasky/porting-200-000-lines-of-c-to-rust-building-a-byte-identical-mediainfo-replacement-8e9b587d469a)
[![Stars](https://img.shields.io/github/stars/vbasky/revelo?style=social)](https://github.com/vbasky/revelo/stargazers)
[![Follow @vbasky](https://img.shields.io/github/followers/vbasky?style=social&label=Follow%20%40vbasky)](https://github.com/vbasky)

## How it reads a file

First revelo detects the format, then it walks the container's *native* on-disk
structure, reading fields box-by-box. Small inputs use a parallel parser race;
large inputs use a bounded sequential probe so speculative parser candidates do
not read the whole file before the real container parser wins. An MP4/MOV file is
a tree of boxes ("atoms") — `moov` holds the metadata, `mdat` the coded samples:

```text
  ftyp                          file brand: isom · qt · M4A …
  moov                          movie-level metadata
    mvhd                        timescale, duration → General: Duration
    trak                        one per track
      tkhd                      track geometry → Video: Width, Height
      edts ▸ elst               edit list / start delay
      mdia
        mdhd                    media header → Language
        hdlr                    vide·soun·subt·text → selects stream kind
        minf ▸ stbl             sample table
          stsd                  codec sample entry → Format (avc1/hvc1/mp4a…)
            avcC ▸ hvcC ▸ esds  codec config → Profile, Level, BitDepth
            colr ▸ pasp         colour, pixel aspect → primaries, DAR
            dvcC ▸ dvvC         Dolby Vision → HDR_Format
          stts                  time-to-sample → FrameRate
          stsz                  sample sizes → StreamSize, BitRate
          stco ▸ co64           chunk offsets
    udta ▸ meta ▸ ilst          iTunes tags, cover art → General: Title
  mdat                          coded samples — sized; scanned for x264/x265 tags
```

(`▸` marks a box that just wraps the next.) revelo walks every container's
*native* structure the same way — here are the other big families.

**Matroska / WebM** — EBML elements (`id ▸ size ▸ data`) under one Segment:

```text
  EBML                          header: DocType (matroska / webm), version
  Segment                       the single top-level element
    SeekHead                    index of the elements below
    Info                        TimestampScale, Duration, Segment UUID
    Tracks ▸ TrackEntry         CodecID → Format; Video/Audio sub-elements
                                → Width, Height, Channels, SamplingRate
    Tags                        Title, Encoder, cover art
    Cluster ▸ SimpleBlock       coded frames (sized; timestamps only)
```

**MPEG-TS** — a flat stream of 188-byte packets, demuxed by PID:

```text
  packet · 188 B                sync 0x47 · PID · payload-unit-start
  PAT  (PID 0)                  lists each program's PMT PID
  PMT                           elementary streams: PID + stream_type
  PES                           → Video/Audio: Format + codec params
```

**RIFF** — chunked; AVI and WAV share the layout:

```text
  RIFF 'AVI '                   LIST hdrl ▸ avih (frame rate, dimensions)
    LIST strl ▸ strh/strf       per-stream header + format → codec params
    LIST movi                   interleaved samples (00dc, 01wb, …)
  RIFF 'WAVE'                   fmt (codec, rate, channels) · data
    bext ▸ iXML ▸ axml          BWF metadata → Title, timecode, loudness
```

Whatever the container, the output is the same: one **General** stream for the
file plus one stream per track — **Video · Audio · Text · Image · Menu · Other** —
and, for photo files, dedicated **Exif · Iptc · Xmp · Icc · C2pa · MakerNotes**
sections (mirroring ExifTool's family-0 grouping).

## MediaInfoLib comparison

Revelo is a from-scratch port of MediaInfoLib v26.05. Every line is new Rust —
no C++ translation, no FFI wrappers, no generated bindings.

| | MediaInfoLib | Revelo |
| --- | --- | --- |
| **Language** | C++ | Pure Rust |
| **Memory safety** | Manual | Compile-time guaranteed |
| **Build** | `./Configure` + `make` + 10 system deps | `cargo build` (no system libs) |
| **Install** | `apt install mediainfo` / `brew install mediainfo` | `cargo install revelo-cli` or `brew tap vbasky/revelo && brew install revelo` |
| **Parser model** | Virtual `File__Analyze` hierarchy | `fn(&mut FileAnalyze) -> bool` flat table; parallel race for small buffers, bounded sequential detection for large buffers |
| **Output fidelity** | Reference oracle | Byte-equal XML (differential harness) |
| **License** | BSD-2-Clause | BSD-2-Clause |
| **Format support** | ~200 formats | 190+ parsers, 185 fields |
| **WASM** | No | Compiles on `wasm32-unknown-unknown` |

## ExifTool maker-note metadata (optional)

The core crate is a pure MediaInfoLib port and is **BSD-2-Clause**. For
camera-maker-note depth on par with [ExifTool](https://exiftool.org/), enable the
opt-in `exiftool-tables` feature:

```sh
cargo install revelo-cli --features exiftool-tables   # or: cargo build --features exiftool-tables
```

It pulls in `revelo-exiftool-tables`, a separate crate of tag tables **generated
from** ExifTool's source. Because those tables are derived from ExifTool (the same
terms as Perl), that crate — and any binary that links it — is **GPL/Artistic**,
not BSD. The licensing is deliberately isolated: the default build and everything
else in the workspace stay BSD-2-Clause; only a build that *opts in* takes on the
copyleft. A SaaS that runs the feature server-side stays unaffected (GPL conveys on
distribution, not network use).

With the feature on, 14 camera vendors decode maker-note metadata using the same
tag names and PrintConv value formatting as ExifTool (derived from ExifTool's own
tables). Parity percentages below are from `revelo-exif-diff` against real camera
samples; `—` means tagged under `exiftool-tables` but not yet quantified with a
dedicated sample corpus.

| Vendor | MakerNote | Sub‑IFDs | Parity |
| --- | --- | --- | --- |
| Canon | Yes | CameraSettings, ShotInfo, AFInfo1/2/3, MyColors, ContrastInfo, TimeInfo, AspectInfo, FaceDetect3, FileInfo, FocalLength, Panorama | 99% |
| Nikon | Yes (Type 2/3) | AFInfo, FlashInfo | 95% |
| Fujifilm | Yes | — | 98% |
| Olympus | Yes (Type 1/2) | Equipment, CameraSettings, RawDevelopment, ImageProcessing, FocusInfo | ~80% ᵃ |
| Sony | Yes | — | — |
| Panasonic | Yes | — | 92% |
| Pentax | Yes | — | — |
| Konica‑Minolta | Yes | — | 97% |
| Samsung | Yes | — | — |
| Sigma | Yes | — | — |
| Apple | Yes | — | — |
| Casio | Yes | — | — |
| DJI | Yes | — | — |
| FLIR | Yes | — | — |

ᵃ Olympus parity is measured from a Sony DSC‑S750 — this budget body writes
Olympus‑format maker notes, which exiftool attributes to the `Olympus` group. A
Sony‑proper maker‑note sample is still needed to quantify the `Sony` row.

Vendors not listed (Leica, Kodak, Ricoh, Epson, Kyocera, etc.) have no maker-note
parser and fall back to standard EXIF IFD tags only. ExifTool supports ~50 camera
makes — revelo targets the 14 most common.

## Project scale

| Metric | Value |
| --- | --- |
| Parsers | 190+ across 8 domains |
| Output fields | 185 (all known MediaInfoLib gaps closed) |
| Output formatters | 10 — XML / Text / JSON + 7 domain |
| Workspace crates | 17 |
| Tests | 800+ |
| License | BSD-2-Clause (core) · GPL/Artistic (optional `exiftool-tables`) |
| MSRV | Rust 1.88+ (edition 2024) |

## Format coverage

| Category | Parsers | Coverage | Formats |
| --- | --- | --- | --- |
| Containers | 42 | 98% | MP4/MOV, MKV/WebM, AVI, MPEG-TS, MPEG-PS, WAV, AIFF, Ogg, FLV, MXF, +32 more |
| Audio | 56 | 100% | AAC/ADTS, MP3, AC-3/4, DTS/DTS-UHD, FLAC, Opus, Vorbis, TrueHD/Atmos, Dolby E, PCM, ADM, MGA, CELT, MPEG-H 3D, **IAMF/Eclipsa Audio**, **AC-4 IMS/JOC**, SMPTE ST 302/331/337, +37 more |
| Video | 28 | 100% | AVC, HEVC, VVC, AV1, VP8/VP9, MPEG-2, VC-1, VC-3/DNxHD, ProRes, FFV1, H.263, MPEG-4V, Theora, Y4M, Canopus HQ, CineForm, Dirac, HDR Vivid, **Dolby Vision RPU**, **HDR10+**, **SL-HDR1**, **HLG/PQ**, **CTA-861**, +8 more |
| Image | 19 | 100% | JPEG, PNG, GIF, BMP, TIFF, WebP, ICO, PSD, DPX, EXR, DDS, BPG, PCX, TGA, ArriRaw, Amiga Icon, RLE, AVIF Gain Map, HEIF |
| Text/Subtitles | 21 | 100% | SubRip, TTML, Timed Text, PGS, DVB Subtitle, Teletext, EIA-608/708, CDP, SCC, N19, PDF, SDP, PAC, DTvCC, SCTE-20, Kate, CMML, ARIB STD-B24/B37, WebVTT |
| Archives | 11 | 100% | ZIP, 7z, RAR, TAR, gzip, bzip2, ACE, ISO 9660, ELF, Mach-O, MZ/PE |
| Tags | 12 | 86% | ID3v1/v2, APE Tag, Vorbis Comment, Lyrics3, EXIF, XMP, ICC, IIM/IPTC, C2PA, PropertyList, SphericalVideo |
| Reader | 4 | 100% | File, Directory, HTTP, MMS |

Complete format catalog with spec references: **[docs/formats.md](docs/formats.md)**

Complete field coverage with sources: **[docs/field_coverage.md](docs/field_coverage.md)**

## Architecture

The crate stack, from consumers at the top down to the foundation:

```text
┌────────────────────────────────────────────────────────────┐
│  CLI (revelo)        C ABI (revelo-cdylib)                 │  frontends
├────────────────────────────────────────────────────────────┤
│  revelo-export       revelo-reader                         │  output / input
├────────────────────────────────────────────────────────────┤
│  revelo-dispatcher — parser table + bounded detect()        │  dispatch
├────────────────────────────────────────────────────────────┤
│  container · audio · video · image · text · tag · archive    │  parsers (193)
├────────────────────────────────────────────────────────────┤
│  revelo-core (FileAnalyze engine)     revelo-util                │  foundation
└────────────────────────────────────────────────────────────┘
   revelo-diff — differential harness, diffs output vs the mediainfo oracle
```

## Crates

Fifteen workspace crates: a foundation layer, seven format-parser domains, the
output/API surface, and the differential test harness.

### Foundation

| Crate | Role | Status |
| --- | --- | --- |
| `revelo-util` | ZenLib port — `Ztring`, bit reader, integer/float types | Stable |
| `revelo-core` | Analysis engine — `FileAnalyze` byte reader, stream model, demux/trace, config dispatch, computed fields | Stable |
| `revelo-dispatcher` | Parser table (single source of truth) + bounded `detect()` | Stable |

### Format parsers

| Crate | Role | Status |
| --- | --- | --- |
| `revelo-parsers-container` | 42 containers — MP4, MKV, AVI, MPEG-TS/PS, WAV, Ogg, MXF, … | Stable |
| `revelo-parsers-audio` | 56 audio codecs — AAC, MP3, AC-3/4, DTS, FLAC, Opus, TrueHD, … | Stable |
| `revelo-parsers-video` | 28 video codecs — AVC, HEVC, VVC, AV1, VP9, ProRes, … | Stable |
| `revelo-parsers-image` | 19 image formats — JPEG, PNG, TIFF, WebP, DPX, EXR, … | Stable |
| `revelo-parsers-text` | 21 subtitle/caption formats — SubRip, PGS, Teletext, EIA-608/708, … | Stable |
| `revelo-parsers-tag` | 12 tag formats — ID3v1/v2, EXIF, XMP, Vorbis Comment, … | Stable |
| `revelo-parsers-archive` | 11 archive/binary formats — ZIP, 7z, RAR, TAR, ELF, Mach-O, … | Stable |

### Output & API

| Crate | Role | Status |
| --- | --- | --- |
| `revelo-export` | 10 output formatters — XML, Text, JSON + EBUCore/MPEG-7/PBCore/NISO/FIMS/Graph/reVTMD | Stable |
| `revelo-reader` | Input layer — File, Directory, HTTP, MMS | Stable |
| `revelo-cli` | The `revelo` command-line binary | Stable |
| `revelo-cdylib` | C ABI (`MediaInfo_*`) — drop-in replacement for libmediainfo | Stable |

### Tooling

| Crate | Role | Status |
| --- | --- | --- |
| `revelo-diff` | Differential harness — diffs revelo's output against the `mediainfo` oracle | Stable |

## Installation

```sh
cargo install revelo-cli                         # CLI (binary from crates.io)
brew tap vbasky/revelo && brew install revelo    # Homebrew (macOS / Linux)
cargo add revelo-core                            # library (core + dispatcher)
cargo add revelo-export                          # library (output formatters)
```

Or add individual parser crates as needed:

```sh
cargo add revelo-parsers-container
cargo add revelo-parsers-audio
cargo add revelo-parsers-video
cargo add revelo-parsers-image
cargo add revelo-parsers-text
cargo add revelo-parsers-tag
cargo add revelo-parsers-archive
```

## Library usage

```rust
// One-liner: detect, parse, extract tags — everything in one call
let meta = revelo::Metadata::from_file("photo.jpg").unwrap();

for (key, value) in meta.general() {
    println!("{key} = {value}");
}
for (key, value) in meta.exif() {
    println!("{key} = {value}");
}
```

For advanced control over the parse pipeline:

```rust
use revelo_core::MediaFile;
use revelo_dispatcher::detect;
use revelo_parsers_tag::parse_tags;

let bytes = std::fs::read("video.mp4")?;
let parser = detect(&bytes).expect("no parser matched");
let mut fa = MediaFile::new(&bytes);
parser(&mut fa);
parse_tags(&mut fa);

for stream in fa.streams() {
    for (key, value) in stream.iter() {
        println!("{} = {}", key, value);
    }
}
```

For XML output (byte-equal with mediainfo):

```rust
use revelo_export::to_xml;
let xml = to_xml(fa.streams(), "video.mp4");
```

## Design

| Principle | Description |
| --- | --- |
| **Pure Rust** | No C++ FFI, no system DLLs, no `pkg-config` — a single `cargo build` |
| **Harness-validated** | Every ported parser is diffed against the `mediainfo` oracle for byte-equal XML |
| **Detect + walk** | Parser-table detection prefers containers before elementary streams; small buffers race in parallel, while large buffers use a bounded sequential probe before the winner re-parses from a fresh state |
| **Container-native** | Each container is walked through its *own* on-disk structure (boxes, EBML, RIFF chunks, PES packets) — not a unified abstraction |
| **No unsafe** | `#![deny(unsafe_code)]` enforced workspace-wide |
| **BSD-2-Clause** | Permissive license, no GPL restrictions |

## Output formats

- **XML** — byte-equal with the MediaInfoLib oracle
- **Text** — 42-column layout, duration as `X s Y ms`
- **JSON** — MediaInfo-compatible `{media:{@ref, track:[...]}}` structure
- **EBUCore, MPEG-7, PBCore, NISO, FIMS, Graph, reVTMD** — 7 domain formatters

## Quick start

```sh
# build the workspace
cargo build --release

# inspect a file (text / json / xml)
cargo run -p revelo-cli -- --text path/to/media.mp4
cargo run -p revelo-cli -- --json path/to/media.mp4
cargo run -p revelo-cli -- --xml  path/to/media.mp4

# build the C ABI shared library (libmediainfo drop-in)
cargo build -p revelo-cdylib --release

# diff revelo's output against the mediainfo oracle
cargo run -p revelo-diff -- path/to/file
```

## Status

All layers are implemented and validated:

| Layer | Status | Detail |
| --- | --- | --- |
| Format parsers | ✓ | 194 parsers across 8 domains |
| Output fields | ✓ | 185 fields, all gaps closed |
| Output formatters | ✓ | XML/Text/JSON + 7 domain |
| Reader layer | ✓ | File, Directory, HTTP, MMS |
| Core infra | ✓ | demux/trace/config dispatch, channel math, IBI, MIME, computed |
| Element trees | ✓ | RIFF, Ogg, MP4 box tree for trace/debug |
| Multi-file | ✓ | BDMV M2TS concatenation, SRT/SST sidecars, duplicate resolution |

**Verified:** all tests pass against the differential harness (byte-equal XML for
the format subset with oracle samples).

**Blocked — unverifiable against the oracle:** `File_Apv` (APV), `File_Av2`
(AV2), `File_Ancillary` (SMPTE 436 VANC), and `File_Umf` (Ikegami UMF) are
intentionally **not** ported. Correctness here is defined as output-match
against `mediainfo` v26.05, and none of these can produce a sample *and* oracle
output to diff against:

- **APV** — ffmpeg 8.x ships only a raw APV muxer, no encoder; mediainfo v26.05
  doesn't surface APV, so there's no oracle target.
- **AV2** — no encoder exists anywhere yet; not in the v26.05 oracle.
- **Ancillary** — only exists embedded in MXF (no standalone sample), and is a
  ~1000-line VANC/CDP/AFD parser.
- **Umf** — proprietary Ikegami format with no obtainable sample.

Porting these would mean translating ~3.7k lines of C++ blind, with no way to run
the differential harness — contrary to the project's harness-validated workflow.
They become tractable only with real sample files (and, for APV/AV2, a newer
mediainfo). DCP PKL, the one validatable holdout, is implemented (BYTE-EQUAL).

## Future improvements

Planned features and high-impact areas for the next development phase:

### Output & Reporting

- ✅ **CSV export** — Machine-friendly format for pipeline integration
  (`--csv`).
- ✅ **YAML export** — YAML output format for pipeline integration (`--yaml`).
- ✅ **HTML report** — Self-contained visual report with collapsible sections and
  summary cards (`--html`).
- ✅ **Summary mode** — Aggregate statistics across a file collection: codec
  distribution, resolution ranges, bitrate profiles, container breakdown.

### Batch & Comparison

- ✅ **Glob / batch processing** — `revelo --json "**/*.mp4"` processes an entire
  directory tree, emitting NDJSON by default (`--json-array` for a single array).
  Other formats concatenate; single-file output is unchanged.
- **Diff mode** — `revelo --diff a.mkv b.mkv` to show which fields differ
  between two files (user-facing, distinct from the harness-oriented
  `revelo-diff`).

### Fidelity gaps

- ✅ **Elementary-stream extraction** — PES payload parsing for MPEG-TS
  (AVC SPS/PPS/SEI + AAC), VP9 keyframe headers in MKV/WebM, a full FLV tag
  demuxer (AVC/AAC), and AV1 OBU sequence headers in MP4. Validated against the
  mediainfo oracle; VP9 in MKV/WebM is byte-equal.
- **Blocked fields** — `Format_Settings_SBR` is now derived by the FLV AAC path;
  `FrameRate_Mode_Original` still needs a real-world VFR-normalized sample (the
  signal is absent from a normalized sample table, so it can't be synthesized
  without mislabelling genuine CFR files).

### Broader reach

- **Python bindings** via PyO3 — natural fit for the media analysis audience,
  with the existing C ABI (`revelo-cdylib`) as a proven bridge.
- **NPM package** — WASM builds already compile on `wasm32-unknown-unknown`; a
  documented JS API and NPM release would enable browser-side media inspection.

### Extraction

- **Cover art / attachment extraction** — `--extract-attachments` flag for MKV
  Attachments, MP4 Cover boxes, and ID3 APIC frames.
- **Subtitle extraction** — Dump subtitle streams to SRT/VTT from any container.
- **Thumbnail / keyframe offset** — Report byte offset of the first keyframe
  (no decoding — metadata-position only).

### Quality of life

- ✅ **Stream filtering** — `--video-only`, `--audio-only`, `--stream KIND:INDEX` to
  select specific tracks by index.
- ✅ **Container verification** — `--verify` flag with `IsComplete` field for
  structural integrity checks.
- ✅ **Format metadata** — `--inform-version`, `--inform-timestamp` for provenance.
- ✅ **Log file** — `--log-file <FILE>` to redirect output to a file.
- **Format conversion** — `revelo --to-json --from-xml` to transform between
  export formats without re-parsing.

## Status

`0.4.x` — 180+ parsers validated against the mediainfo oracle for byte-equal XML
output. All parsers are harness-validated; the remaining gaps are documented and
scoped. See the [status & roadmap](STATUS.md) for what's covered today and what's
planned on the way to `1.0`.

## License

BSD 2-Clause License — see [LICENSE](LICENSE) for details.

revelo

revelo 최신버젼 다운로드

오늘의 추천 프로젝트

Was this content helpful?

Recent Feedback

revelo

revelo 최신버젼 다운로드

오늘의 추천 프로젝트

Was this content helpful?

Recent Feedback

Download README