Bradford Lab is designed around professional media formats and delivery requirements, with attention to the details that matter in finishing workflows.
The formats that come through finishing are not a vocabulary problem — they are a precision problem. What follows is a working reference for the codecs, containers, packages, and standards Bradford Lab ingests, validates, and delivers, with the parameters that actually matter on the floor.
Apple ProRes is a family of intra-frame, variable bit rate codecs introduced in 2007 and now the de facto mezzanine for post in North America. The family spans six variants at 1920×1080: ProRes Proxy (target ~45 Mbps), ProRes LT (~102 Mbps), ProRes 422 (~147 Mbps), ProRes 422 HQ (~220 Mbps), ProRes 4444 (~330 Mbps), and ProRes 4444 XQ (~500 Mbps). Bit rates scale roughly linearly with resolution; a UHD ProRes 422 HQ master sits around 880 Mbps. The 422 variants are 10-bit 4:2:2, the 4444 variants are 12-bit and carry an optional alpha channel, and 4444 XQ is engineered for high dynamic range origination where the extra headroom matters.
In practice the 422 variants cover editorial and broadcast finishing, 4444 covers grading and VFX plates where chroma fidelity matters, and 4444 XQ covers HDR masters and beauty work. ProRes is almost always delivered in a QuickTime (.mov) wrapper; the MXF flavor exists but is rare outside specific broadcast houses. Common pitfalls: misreading bit rate as quality without checking the variant code, treating ProRes Proxy as a deliverable, dropping alpha when transcoding 4444 to 422, and tagging color metadata (transfer, primaries, matrix) incorrectly on the wrapper.
Bradford Lab parses ProRes variant codes, bit depth, chroma subsampling, alpha presence, and the QuickTime color tags (nclc / nclx) at ingest, and flags mismatches between the file's actual chroma sampling and the variant it claims to be.
Avid DNxHD is the HD-era counterpart to ProRes 422, with named bit rate tiers (36, 115, 145, 175, 175X, 220X, 444) where X denotes 10-bit and 444 denotes 10-bit 4:4:4. DNxHR is the resolution- independent successor, covering UHD and beyond with quality tiers labeled LB, SQ, HQ, HQX, and 444. LB and SQ are 8-bit 4:2:2 for offline and online proxy, HQ is 8-bit 4:2:2 finishing, HQX is 12-bit 4:2:2, and 444 is 12-bit 4:4:4. DNxHR HQX is the common online finishing tier for UHD work in Avid Media Composer.
DNx codecs ship in two containers: MXF OP1a for Avid-native workflows (where the codec lives alongside AAF metadata and frame-accurate timecode), and QuickTime for hand-off to non-Avid tools. Broadcasters frequently specify MXF OP1a DNxHD 145 or 220 as a finished master; streamers and theatrical clients prefer ProRes. Bradford Lab validates DNx variant, bit depth, chroma, color tags, and the MXF wrapper conformance (essence descriptor, edit rate, channel mapping).
A Digital Cinema Package is the theatrical delivery format. Picture is JPEG 2000 in an MXF wrapper at 2K (2048×1080) or 4K (4096×2160), 12-bit XYZ color, 24 or 48 fps, capped at 250 Mbps for 2K and 500 Mbps for 4K. Audio is uncompressed 24-bit PCM at 48 kHz or 96 kHz in a separate MXF. A package is described by three XML files: the CPL (Composition Playlist) defines the composition and reel structure, the PKL (Packing List) lists every asset with hashes, and the ASSETMAP plus VOLINDEX describe the on-disk layout. Multi-reel compositions split the show into reels (typically ~20 minutes each), which lets exhibitors handle errors per-reel and lets distributors patch a single reel rather than reship a feature.
Two standards coexist. Interop DCP is the older, more permissive specification, with subtitles authored as DCSubtitle XML. SMPTE DCP (ST 428) is the modern standard, with subtitles authored as SMPTE 428-7 timed text and stricter constraints on frame rates, audio configurations, and metadata. Most current servers play both, but festival and theatrical clients increasingly require SMPTE. Encrypted packages add a KDM (Key Delivery Message) bound to a specific server certificate, date window, and CPL; without the matching KDM the show will not play. ISDCF naming encodes the package identity in the Content Title Text: film title, content type (FTR, TRL, TSR, ADV, SHR, RTG-X), aspect ratio (F for flat, S for scope, C for full container), language pair (audio-subtitle), region and rating, audio configuration (51, 71, ATMOS), resolution (2K, 4K), studio code, distributor code, date, facility, package type (OV for original version, VF for version file), and standard (IOP or SMPTE).
Common pitfalls: subtitle reel boundaries that don't align with picture reel cuts (causing dropped subtitles at reel transitions), mixing Interop subtitles into a SMPTE package, missing or expired KDMs, incorrect ISDCF fields that pass naming validation but mislabel the content, and 7.1 audio that gets silently truncated to 5.1 by downstream tools. Bradford Lab runs structural validation across CPL, PKL, and ASSETMAP, confirms SMPTE versus Interop conformance, detects encrypted packages, verifies multi-reel composition coherence, validates ISDCF naming, and plays DCPs inline in the browser using a clairmeta and easyDCP-backed validation pass. 7.1 audio rendition support is currently deferred — the active pan filter truncates 7.1 to 5.1 without warning, and will be replaced when a real 7.1 package is available to test against.
H.264 (AVC) is the universal distribution codec. The profiles that matter in post are Main (8-bit 4:2:0, baseline web delivery), High (8-bit 4:2:0, the streaming default), and High 10 / High 4:2:2 for 10-bit and chroma-subsampled mastering. Level codes (4.0, 4.1, 4.2, 5.1, 5.2) cap resolution, frame rate, and bit rate combinations; level 5.1 is the practical ceiling for UHD H.264. H.265 (HEVC) is the successor, with Main, Main 10, and Main 4:2:2 10 the profiles seen in finishing. Main 10 is the carrier for HDR distribution: HDR10 static metadata, HDR10+ dynamic metadata, and Dolby Vision Profile 5 / 8.1 / 8.4 all ride on H.265 Main 10 in MP4 or MOV.
H.264 is appropriate for review screeners, web preview, SDR streaming, and broadcast contribution. H.265 is appropriate for UHD distribution, HDR mastering for streamers, and any deliverable where the platform spec calls for 10-bit HEVC with embedded HDR metadata. Bradford Lab parses profile, level, bit depth, chroma, HDR metadata payloads (MaxCLL, MaxFALL, mastering display primaries), and confirms the metadata matches the picture.
AV1 is the royalty-free successor to H.265, developed by the Alliance for Open Media. It delivers roughly 30 percent better compression than H.265 at equivalent quality and is now in active deployment at YouTube, Netflix, and Twitch for streaming distribution. JPEG 2000 is the picture codec inside DCP and IMF packages: intra-frame, wavelet-based, mathematically lossless at the top tier, and never used outside cinema and broadcast archive contexts. MPEG-2 remains the carriage codec for legacy broadcast contribution and ATSC 1.0 over-the-air transmission, and still appears in DVD and Blu-ray authoring. AVC-Intra is Panasonic's intra-frame H.264 variant for broadcast acquisition, common in P2 workflows at 50, 100, and 200 Mbps. XDCAM is Sony's long-GOP MPEG-2-based acquisition format, ubiquitous in ENG, news, and live sports for its small file sizes and fast metadata access. Bradford Lab ingests all of these as source media and transcodes to ProRes or DNx for finishing where the downstream tool can't handle them natively.
PCM WAV is the lingua franca of post audio: uncompressed linear PCM in a RIFF wrapper, 16-bit or 24-bit, with 48 kHz the universal post sample rate (96 kHz for high-resolution music masters, 44.1 kHz only for legacy music delivery). Broadcast WAV (BWF) extends WAV with a<bext> chunk carrying origination metadata: originator, originator reference, origination date and time, and crucially a Time Reference field expressed in samples from midnight. That time reference is the file's start timecode, not a content-alignment marker — a BWF recorded starting at 10:00:00:00 carries a Time Reference of 1,728,000,000 samples at 48 kHz, and any tool that reads BWF correctly will place the file at 10:00:00:00 on the timeline regardless of when it was imported.
Multichannel material is delivered as polyphonic BWF (one file, interleaved channels with iXML channel mapping) or as a folder of monophonic BWFs (one file per channel, named for routing). Production sound is almost always polyphonic; stem deliveries are split monophonic so each channel can be ingested independently. Bradford Lab parses the bext chunk, validates time reference against filename and slate metadata, and confirms sample rate, bit depth, and channel count match the project's declared spec.
Surround channel orderings vary by industry. SMPTE order (used in broadcast and most file-based workflows) for 5.1 is L, R, C, LFE, Ls, Rs. Film order (used in DCPs and on dub stages) for 5.1 is L, C, R, Ls, Rs, LFE. 7.1 adds Lss/Rss (side surrounds) and Lrs/Rrs (rear surrounds) in SMPTE, or Lc/Rc (left center / right center) in the legacy theatrical 7.1 used for some Dolby releases. Dolby Atmos for cinema is delivered as a bed (7.1.2 in most cases) plus up to 118 dynamic objects; the IMF and DCP carriage is a separate Atmos sidecar (.atmos), not embedded in the picture MXF.
Common stem deliverables: M&E (music and effects, no dialog, for international dub), DM&E (dialog, music, and effects, full mix minus narration), dialog stem (just spoken word), music stem (score and source music), and effects stem (Foley, hard FX, ambiences). Each is delivered in the show's native channel configuration. Common pitfalls: SMPTE versus film order mismatches that send the LFE to a surround channel, missing Lc/Rc channels on theatrical 7.1 deliveries, and Atmos beds delivered without the corresponding object track.
Modern loudness measurement is based on ITU-R BS.1770, which defines K-weighted, gated integration over the program duration. EBU R128 is the European broadcast standard: integrated loudness target of -23 LUFS with a tolerance of ±1 LU, maximum True Peak of -1 dBTP, and a Loudness Range (LRA) typically held under 20 LU for broadcast. ATSC A/85 is the United States broadcast standard codified by the CALM Act: integrated loudness target of -24 LKFS with a ±2 LU tolerance and -2 dBTP True Peak. LUFS and LKFS are numerically equivalent — the unit name differs by standards body.
Streaming platforms publish their own targets. Netflix specifies -27 LUFS integrated and -2 dBTP for original content masters; the platform normalizes on playback so over-loud material is turned down rather than rejected. Apple TV+ aligns with -27 LUFS for original mixes. Apple Podcasts and most podcast platforms target -16 LUFS for stereo programs. YouTube normalizes to roughly -14 LUFS. Theatrical mixes are not measured in LUFS at all — they are calibrated to 85 dB SPL pink noise per channel on a properly aligned stage, with dialog typically averaging around -31 LKFS.
Bradford Lab measures integrated loudness, short-term loudness (3-second sliding window), momentary loudness (400-millisecond window), True Peak, and Loudness Range against the target standard, and reports gated and ungated values so deliverables can be conformed to platform specs before shipping.
SRT (SubRip) is the simplest sidecar format: plain text with HH:MM:SS,mmm timecodes and minimal formatting. It's universally supported and universally limited — no positioning, no styling, no speaker identification beyond convention. WebVTT (.vtt) is the web standard, extends SRT with CSS-style positioning, color, and cue settings, and is the required format for HLS segmented subtitle playlists. TTML (Timed Text Markup Language) is XML-based and the foundation for SMPTE-TT, IMSC 1.0.1 / 1.1, EBU-TT-D, and Netflix Timed Text. IMSC is the modern interoperable profile and the format most streamers require for finished captions. EBU-STL (.stl) is the European broadcast standard, a fixed-width binary format still required by many EU broadcasters.
CEA-608 (line 21 analog) and CEA-708 (digital, in the H.264/H.265 elementary stream as user data SEI messages) are the broadcast caption standards required for FCC-compliant US distribution. 608 is ASCII-limited and supports up to four caption channels; 708 supports Unicode, eight services, and richer positioning. Embedded captions ride inside the video file as SEI; sidecar captions are separate XML or text files referenced by the manifest or playlist. For DCP, Interop subtitles use the DCSubtitle XML format and SMPTE packages use SMPTE 428-7 timed text in an MXF wrapper, with subtitle reels that must align with picture reel boundaries.
Bradford Lab validates SRT and WebVTT timing (overlap, reading speed, minimum gap), parses TTML and IMSC, runs CEA-608/708 character set checks against allowed glyph tables, validates DCP subtitle reel boundaries against the composition, and converts between formats with frame-rate awareness so 23.976-to-25 retiming doesn't introduce timing drift.
CMX 3600 EDL is the original timeline interchange format, dating from linear tape suites in the 1970s. It is plain text, one event per cut, with source and record timecodes, reel names, and limited transition support (dissolves and wipes). EDL remains the lingua franca for color conform: every grading system reads it, and it survives round trips that break richer formats. Its limits are real — one video track, no effects, no nested sequences, no audio levels — but for handing a cut from picture editorial to a colorist it remains the safest bet.
FCPXML is Apple's timeline format, originally for Final Cut Pro X and now widely supported as an interchange between Final Cut, Resolve, Premiere, and Motion. It carries multiple tracks, effects parameters, audio levels, markers, and metadata. AAF (Advanced Authoring Format) is Avid's interchange, complex but feature-rich: it preserves track-level audio routing, effects automation, embedded essence, and the full Avid bin structure, and is the standard for picture-to-audio handoff via the AAF to Pro Tools workflow. OpenTimelineIO (OTIO) is the modern open-source format from Pixar, with first-class support for Resolve, Hiero, and a growing list of tools, and adapters for EDL, FCPXML, and AAF that let it serve as a hub format for cross-application pipelines.
Bradford Lab parses EDL events, imports FCPXML, reads AAF, and supports OTIO across editorial- to-finishing handoffs, with conform reporting that flags missing source media, frame-rate mismatches, and clips whose reel names don't resolve to a known source.
Request early access to Bradford Lab.