Platform Delivery Spec Card
Pick Netflix, Prime Video, Apple TV+, Disney+, YouTube or BBC/DPP and get its delivery spec as one card — codec, loudness, captions, start timecode. Free.
Client review & approval for video, audio, images & PDF — one link, no per-seat fees.
Try FreePick a Platform
- Wrapper / codec
- Resolution
- Frame rate
- Color / HDR
- Audio
- Loudness #1 QC rejection reason
- Player-side normalization, not a delivery rejection threshold. Measure your own master against this target →
- True peak
- Captions / subs
- Start timecode
Partner specs change; always confirm against the current official delivery specification — .
All Six Platforms Side by Side
Feature / episodic values. Netflix trailers and promos measure loudness without dialogue gating — see the card above.
| Row | |
|---|---|
| Wrapper / codec | |
| Resolution | |
| Frame rate | |
| Color / HDR | |
| Audio | |
| Loudness | |
| True peak | |
| Captions / subs | |
| Start timecode |
Partner specs change; always confirm against the current official delivery specification for the platform you are delivering to.
About Platform Delivery Spec Card
Why delivery specs exist
A platform's delivery specification is a QC rejection list written in advance. Netflix, Prime Video, Apple TV+, and Disney+ all run automated and manual quality control against the exact wrapper, codec, resolution, loudness, and caption values in their published specs, and a mismatch on any row sends the deliverable back to the vendor rather than into the catalog. That is expensive in a way that has nothing to do with the creative work: a bounced delivery means re-encoding or re-mixing, resubmitting through the platform's compliance checker, and waiting in a QC queue again — commonly three to ten business days, sometimes longer during a platform's peak delivery season. On a title with a locked premiere date or a marketing campaign already scheduled, a spec-fail isn't a technical footnote, it's a missed date with a contractual or financial consequence attached. Reading the spec correctly the first time is cheaper than any other option.
A guided tour of one card
Each card on this page answers the same nine questions a QC operator checks, roughly in the order they check them. Wrapper/codec and resolution rule out the most common instant-reject: an MP4 sent to a platform expecting an IMF package, or a UHD-labeled file that's actually upscaled 1080p. Frame rate and color/HDR catch native-rate mismatches and HDR metadata that doesn't match the declared format (PQ vs HLG, missing Dolby Vision XML). Audio confirms channel layout and stem structure — a 5.1 mix delivered as one interleaved file instead of discrete stems is a routine Netflix bounce. Loudness and true peak are highlighted on every card because they are the single most common rejection reason across every platform here — see below. Captions/subtitles confirms the sidecar format the platform's caption pipeline actually ingests (IMSC1/TTML is not interchangeable with a plain SRT). Start timecode exists so slates, credits and chapter markers line up with the platform's automated media validation, which expects the first frame of picture at a fixed offset rather than at 00:00:00:00.
Loudness, decoded: LKFS vs LUFS, and two kinds of gating
LKFS and LUFS measure exactly the same thing — ITU-R BS.1770 integrated loudness — under two names from two standards bodies (LKFS from ATSC/US broadcast practice, LUFS from EBU practice); a −24 LKFS target and a −24 LUFS target are numerically identical. The difference that actually matters for delivery is gating: how much of the program the loudness meter is allowed to ignore while averaging. Netflix's −27 LKFS ±2 target is dialogue-gated — the meter measures loudness only during dialogue, so a loud, dialogue-free action sequence doesn't drag the average up. Prime Video, Apple TV+, and Disney+ specify full-program (BS.1770 relative) gating at −24 LKFS ±2, which uses the standard BS.1770 silence-relative gate rather than isolating dialogue specifically. Broadcast (BBC/DPP, EBU R128) works the same way at −23 LUFS ±1. YouTube doesn't gate anything at delivery, because it isn't a delivery spec at all: there's no rejection threshold, and instead YouTube's player normalizes playback loudness to roughly −14 LUFS after upload, turning louder uploads down (never up) to sit closer to quieter ones in the same session or playlist.
| Context | Target | Measurement | Mechanism |
|---|---|---|---|
| Broadcast (BBC/DPP, EBU) | −23 LUFS ±1 | ITU-R BS.1770 / EBU R128 | Rejection spec, full-program gated |
| Streaming — Netflix originals | −27 LKFS ±2 | ITU-R BS.1770 | Rejection spec, dialogue-gated |
| Streaming — Prime Video / Apple TV+ / Disney+ | −24 LKFS ±2 | ITU-R BS.1770 | Rejection spec, full-program gated |
| Web — YouTube playback | ≈ −14 LUFS | ITU-R BS.1770-based | Player-side normalization, not a delivery gate |
Measure your own master against whichever target applies with the LUFS Meter before you deliver — it's the fastest way to catch a loudness spec-fail before QC does.
What a spec-fail actually costs
Beyond the calendar slip, most platforms charge a re-QC fee for a second delivery attempt, and a broadcast slot missed for technical reasons can carry a contractual penalty separate from any re-delivery cost. The specs on this page are current as of 2026, but every platform revises its requirements over time — Netflix has updated its non-branded delivery spec multiple times in recent years, HDR formats keep getting added, and codec allowances shift as encoders improve. Treat this card as the fast pre-check, not the final word: confirm the exact current version against the platform's own spec portal (linked on every card) before you lock a master for delivery.
How to use the Platform Delivery Spec Card
Pick your platform — Netflix, Prime Video, Apple TV+, Disney+, YouTube or BBC/DPP.
Set the content type where it moves the answer, such as feature and episodic versus trailer and promo.
Work down the card row by row, then build the head of the master with the Broadcast Leader Builder.
Frequently Asked Questions
Netflix's non-branded delivery spec targets −27 LKFS ±2, measured with dialogue gating rather than full-program gating — the meter averages loudness only across dialogue-containing segments. True peak must stay at or below −2 dBTP. This applies to feature and episodic long-form deliveries; always confirm the current figure against Netflix's own Partner Help Center before final QC, since Netflix revises its spec periodically.
None, mathematically — both measure ITU-R BS.1770 integrated loudness on the same scale, so −24 LKFS and −24 LUFS are the same loudness. LKFS (‘Loudness, K-weighted, relative to Full Scale’) is the naming convention from US/ATSC broadcast standards; LUFS (‘Loudness Units relative to Full Scale’) is the EBU naming convention used in Europe and by most streaming platforms. You'll see LKFS on Netflix's spec and LUFS on EBU R128 broadcast specs and YouTube's normalization target — the underlying measurement is identical.
No — YouTube has no loudness rejection spec at all. Instead, YouTube's player normalizes playback loudness to roughly −14 LUFS after your video goes live: uploads mixed louder than that get turned down in playback (never boosted), so a hot master just plays quieter than it was mixed — it doesn't bounce at delivery. That's a fundamentally different mechanism from Netflix, Prime Video, Apple TV+ or Disney+, which will reject a file outright for missing their loudness target.
IMF (Interoperable Master Format, SMPTE ST 2067-21) packages a title as one set of shared essence — video, multiple audio configurations, multiple subtitle languages — plus small XML ‘Composition Playlists’ that describe how to assemble a specific version, instead of exporting a separate flat file per language or region. Netflix, Apple TV+ and Disney+ all require IMF for long-form originals because it lets them generate every territorial or language version from one mezzanine without re-encoding picture for each one.
AS-11 is an AMWA-defined constrained MXF file format for delivering finished programs to a broadcaster. DPP — the UK Digital Production Partnership, backed by the BBC, ITV and Channel 4 — defines the specific AS-11 profile (AS-11 UK DPP HD) used for UK broadcast delivery, including a required metadata set (title, series, compliance, contact information) embedded directly in the file rather than sent as a separate document.
Starting the program at one hour of timecode rather than zero leaves a full hour of addressable pre-roll for bars, tone, slate and countdown leader without the timecode ever going negative or rolling over at midnight — both of which historically confused tape-era broadcast equipment. It's now a near-universal convention across Netflix, Prime Video, Apple TV+ and Disney+ IMF deliveries; DPP broadcast delivery in the UK uses the same idea at 10:00:00:00 to leave room for a longer slate and countdown sequence.
Related Tools
Check your next delivery in VoiceDeck review
VoiceDeck adds client review & approval for video, audio, images & PDF — one shareable link, no per-seat fees, so every file ships in spec.