DCP Doctor checks a cinema or streaming package for the mistakes that get a delivery rejected: broken structure, bad hashes, expired keys, subtitle timing, JPEG 2000 forensics, audio levels and HDR metadata. Run a lighter check in the browser, or install the desktop app, which also reads every JPEG 2000 frame and the MXF essence.
Free, open source (AGPL)
Windows, macOS, Linux
SMPTE, Interop, IMF
Nothing uploaded
The desktop app is the thorough check. It reads the JPEG 2000 essence frame by frame. Drop in a DCP or IMP folder and get a pass/fail report with every warning called out. Free and open source, all platforms.
Prefer the command line? The same releases page has a standalone command-line build for Windows, macOS and Linux, and a command-line .deb, for scripting and watch folders. See Automation.
dcpdoctor as WebAssembly, in your browser. Point it at a DCP or IMP folder. Nothing is uploaded. It checks the structure, the PKL hashes, the CPL, the ISDCF name and the MXF and JPEG 2000 headers. It does not decrypt, verify signatures, scan every frame or measure loudness. For those, use the desktop app.
Or open the full-page checker for more room.
Grouped by the job in front of you, from a quick structure check to a full QC pass. Plain reports you can hand to a client or an operator.
Checks that ASSETMAP, PKL and CPL line up and that every asset they point to is actually present. Catches duplicate asset IDs, a CPL whose title does not match its annotation, and a missing CompositionMetadataAsset. A supplemental (VF) package resolves its references against the OV you pass with --ov.
Verifies the SHA-1 hash and size of every file, including the hash each CPL reel asset carries, so a corrupt or half-copied asset gets caught before it reaches a theater or a facility. --manifest also compares the files against a JSON list of names and sizes.
Validates the XML signatures and the certificate chain against the ST 430-2 rules, flags expired certs and malformed package XML, and catches an encrypted package that was left unsigned. The SMPTE and Interop XSD pass runs when xmllint is installed and the repository's schemas folder is found, as in a source build or the Docker image, or through DCPDOCTOR_SCHEMA_DIR. Otherwise the report says it did not run.
Confirms the 2K or 4K container, Flat / Scope / Full and aspect ratio, including the coded scope sizes (2048×858, 4096×1716). Flags a 4K track at an illegal frame rate. App 2E tracks are checked for an IMF JPEG 2000 profile, matching picture coding label, matching descriptor and codestream pixel depths, color primaries and transfer characteristic. Sampled frames are decoded to catch X′Y′Z′ samples behind an IMF label.
Reads every picture frame and checks the peak against the DCI cap of 250 Mbps at every resolution, with a warning within 5 percent of it. IMF has no such cap, so an IMP's peak and average are reported as an info note. frame-qc prints min, max and average for a folder of J2K frames.
One pass over every frame reports the track's codestream parameters (profile, tiles, TLM, POC, code-block size), the fattest frame against the DCI cap, and any parameter that changes partway through. The same scan runs on IMF AS-02 picture, minus the cinema-only rules.
Detects XYZ / DCI-P3 / Rec.709 and confirms 12-bit depth, with out-of-gamut sampling.
--hdr reads the PQ or HLG transfer from the picture descriptor and MaxCLL and MaxFALL from the CPL, and errors on a MaxFALL above MaxCLL. --dolby-vision reads the RPU of an HEVC track file. hdr-validate checks a file's HDR10 light levels against the values you expect.
The ST 429-10 reel is checked for both eyes, for a FrameRate twice the EditRate, and for essence that really is a stereoscopic J2K track. Each eye is scanned against half an edit unit's bitrate cap.
Validation errors on a true peak over -1 dBTP and warns when a sound track is silent on every channel. Auto-QC reports a channel's peak in dBFS when it clips.
EBU R128 (ITU-R BS.1770) loudness measurement, ISO 21727 Leq(m) against the content kind's limit (82 dB advertisement, 85 dB trailer), and normalization to a target when you need it. The default target is -23 LUFS.
Checks the channel layout (2.0, 5.1, 7.1, immersive), MCA labels in the sound MXF, and the CPL's MainSoundConfiguration against the actual channel count.
Detects Dolby Atmos (ST 429-18) and IMF IAB essence. Atmos reports the object count from its descriptor. IAB is reported as present, since its descriptor carries no count.
av-sync compares each reel's picture and sound as the CPL declares them, and reports the entry point offset and duration difference in frames.
Bv2.1 timed-text rules: in/out order, overlaps, minimum duration and gap, a too-early first subtitle, StartTime and EntryPoint, a cue that runs past its reel, a language that changes mid-composition, and a subtitle or caption missing from a reel.
Line count and length limits, closed-caption character-set and layout, required Language and LoadFont, schema validation of the DCST document, the 640 kB / 10 MB font caps, and the ST 429-5 rule that ResourceID repeats the document Id.
--accessibility reports whether the CPL names audio description, hearing-impaired or closed-caption tracks, and warns when it names none. The timed-text rules also read MXF-wrapped SMPTE subtitles, and encrypted timed text when a KDM is supplied.
Spots encrypted assets. A composition that mixes encrypted and clear essence is an error. Encrypted picture, sound and timed text are checked when you pass a KDM and the recipient's private key, on a DCP and on an IMP alike. Without a key the report says so rather than passing quietly.
Checks a KDM's validity window, its signature, that it names the right CPL, thumbprint length, and that the assume-trust marker is not mixed with a real device list. It is validated against the ST 430-1 and 430-3 schemas when the XSD pass can run.
Finds black frames, freeze frames, audio silence and clipping, with thresholds you can set, and names the time each one starts and ends.
HTML or PDF reports with package, track and per-track loudness summaries, plus a JPEG 2000 forensics table per picture track, with a client name and title. PDF needs weasyprint or wkhtmltopdf installed.
Diff two DCPs to see what changed in the structure and hashes, down to a perceptual picture fingerprint. frame-compare scores two packages or files frame by frame with PSNR and SSIM, and VMAF when ffmpeg has libvmaf.
validate --timeline writes an SVG of the reels, each labelled with its duration as a timecode at the reel's own edit rate.
profiles checks a package's resolution, frame rate, channel count, Atmos and 3D against built-in profiles for Dolby, Barco, Christie, GDC, IMAX and NEC servers and projectors. Separate playback warnings flag a 25, 30, 48, 50 or 60 fps package, 4K 3D, or a sound track that is not 8 or 16 channels, which many rooms will not play even when the package is legal.
facility-check reports whether a package is ready for theater ingest: the ASSETMAP, VOLINDEX, PKL, CPL and MXF files, namespaces, hashes, signing certificates and ISDCF name. On a SMPTE CPL it also checks the Chain, Distributor, Facility and Luminance metadata some facilities' QC requires.
--netflix checks the Netflix IMF delivery rules Photon applies to the ASSETMAP, CPL namespace, application and edit rate. Deep IMF conformance runs through Netflix Photon, which needs Java and the Photon jars from scripts/fetch_photon.sh.
Validates the content title against the ISDCF naming convention: content type, language, audio and resolution, with the audio field read against the ISDCF registry.
Repairs PKL and CPL hashes, PKL sizes, wrong SMPTE or Interop namespaces and invalid ContentKind, with a dry run to preview first. A signed CPL or PKL is left alone unless you pass --break-signatures.
Point it at an ingest folder and it checks each new package once every file its ASSETMAP names has finished copying.
imp-info lists an IMP's tracks with the raster, frame count, channels and sample rate read from each track file. mxf-extract pulls the JPEG 2000 frames or the PCM sound out of a track file. checksum-verify checks every PKL hash and size on its own.
A delivery check you can run on every workstation, free and open source.
| Feature | DCP Doctor | EasyDCP Plus |
|---|---|---|
| SMPTE Bv2.1 Compliance | ✓ | ✓ |
| Interop DCP Support | ✓ | ✓ |
| Hash Verification | ✓ | ✓ |
| J2K Bitrate Analysis (250 Mbps DCI cap) | ✓ | ✓ |
| JPEG 2000 forensics and App 2E descriptor checks | ✓ | - |
| Audio Loudness (EBU R128 + Leq(m)) | ✓ | ✓ |
| Dolby Atmos / IAB | ✓ | ✓ |
| Stereoscopic 3D | ✓ | ✓ |
| Up to 4K | ✓ | ✓ |
| DCP 24 to 60 fps, IMF App 2E 23.98 to 120 fps | ✓ | ✓ |
| SMPTE 428-7 / Bv2.1 timed text | ✓ | ✓ |
| Subtitle Font Validation | ✓ | ✓ |
| KDM Validation | ✓ | ✓ |
| CLI / Scriptable | ✓ | ✓ |
| HDR metadata (PQ and HLG transfer, MaxCLL/MaxFALL) | ✓ | - |
| Dolby Vision RPU in HEVC track files (profile, frames, MaxCLL/MaxFALL) | ✓ | - |
| Netflix IMF delivery rules | ✓ | - |
| IMF Validation (Photon) | ✓ | ✓ |
| Accessibility track detection | ✓ | - |
| Content Fingerprinting | ✓ | - |
| Out-of-gamut sampling | ✓ | - |
| Certificate Chain Trust | ✓ | ✓ |
| ISDCF Naming | ✓ | - |
| DCP Diff / Comparison | ✓ | - |
| JSON/HTML Reports | ✓ | - |
| macOS / Windows | ✓ | ✓ |
| Linux | ✓ | - |
| Price | Free (AGPLv3) | Commercial |
The command-line build does everything the app does, so you can batch a whole day of deliveries, watch an ingest folder, or wire checks into your own tools.
Run a full check and get a pass/fail exit code you can act on.
dcpdoctor /path/to/DCP dcpdoctor validate --bv21 --strict /path/to/DCP
Watch an ingest folder and check each new package as it lands.
dcpdoctor watch /ingest/incoming
Run a small HTTP service and POST a package path to check it. It listens on 127.0.0.1 unless you pass --bind, and --api-key requires a key on every request but /health.
dcpdoctor serve --port 8080
Pass several packages and get a summary table with the pass count at the end.
dcpdoctor /dcp1 /dcp2 /dcp3
Write the report as JSON or HTML, to a file or into each package's own folder.
dcpdoctor --json /path/to/DCP dcpdoctor validate --html -o report.html /path/to/DCP
Build the image from the repository. It carries ffmpeg, xmllint with the schemas, and Java with Photon. No image is published.
docker build -t dcpdoctor . docker run --rm -v /path/to/dcp:/data/dcp:ro \ dcpdoctor validate /data/dcp
The project is open source under the AGPL. The core and command line are written in Rust. The desktop app wraps them with Tauri.
# command line (needs Rust 1.89+, CMake, a C++ compiler and OpenSSL development files)
git clone --recurse-submodules https://github.com/PostPerfection/dcpdoctor.git
cd dcpdoctor/rust && cargo build --release
# desktop app (needs Node.js + pnpm as well)
# copy rust/target/release/dcpdoctor to gui/src-tauri/dcpdoctor-<target-triple>
# on Windows, also stage the VC++ runtime and vcpkg DLLs (scripts/bundle-windows-cli-dlls.ps1)
cd ../gui && pnpm install
pnpm tauri build
Full build notes and platform prerequisites are in the README. Source, issues and pull requests live on GitHub.
DCP Doctor is free and open source (AGPL), for cinema and streaming post-production. Download it or check a package in your browser.
Part of the PostPerfection suite · Open to contributions from everyone