This guide cuts through the complexity. You’ll get the current platform-by-platform requirements, the delivery formats that actually matter, the localization specs you can’t skip, and a clear framework for building a post-production workflow that won’t blow up at delivery.

Why OTT Delivery Specs Have Changed Dramatically Since 2022

What this means for producers is straightforward: your delivery checklist isn’t just a technical formality. It directly determines how quickly your content goes live, how many territories you can serve, and whether the platform’s localization pipeline can run at all. Get the specs wrong—even on metadata—and your content sits in a rejection queue while your launch date slips.

The real dynamic here is that platforms have quietly raised their baseline requirements while producers are still building finishing workflows designed for a 2019 delivery environment. The gap is costing independent producers real money and time.

Video Format Requirements by Platform: What Actually Differs

The honest starting point: there is no single universal delivery format. Platforms have converged on 4K UHD as the production standard for originals, but the HDR format requirements diverge sharply—and that divergence has significant cost implications for your finishing workflow.

Netflix

Netflix only accepts Dolby Vision-graded content—HDR10 is not accepted for new originals. This is a hard requirement, not a preference. Your colorist needs to be working in a Dolby Vision-calibrated environment, and your deliverable must include either embedded Dolby Vision metadata or a sidecar metadata asset. The delivery format for Netflix originals is IMF (Interoperable Master Format)—ProRes is accepted for certain legacy deliveries but IMF is the default for 4K HDR content. Frame rate support covers 23.976, 24, 25, and 30fps. Get your finishing timeline right: Netflix expects IMF assets 4–6 weeks before your launch date. Miss that window and you’re paying for expedited encoding.

Amazon Prime Video

Amazon supports both HDR10+ and Dolby Vision—a more flexible baseline than Netflix but with its own complexity. HDR10+ is the default on Samsung TVs; Dolby Vision applies on LG, Sony, and select other manufacturers. For audio, Amazon uses AAC for stereo and Dolby Digital Plus (E-AC-3) for surround, with Dolby Atmos available on select titles. Unlike Netflix, Amazon relies primarily on H.264 and VP9 codecs—Netflix has adopted AV1 for compression efficiency, which delivers roughly 30% better compression than VP9 but requires encode workflows capable of handling it.

Apple TV+

Apple mandates 4K HDR with Dolby Vision on virtually all originals—one of the stricter baseline requirements in the streaming landscape. Delivery uses Apple’s own asset pipeline via Transporter (the command-line delivery tool), with video assets packaged as .itmsp files including a metadata.xml descriptor. Apple also requires what it calls an Ambient Video Asset—a looping background video for the platform UI, delivered in multiple aspect ratios including vertical orientation for iPhone. It’s a spec most producers forget about until the delivery package gets kicked back.

The Pragmatic Default

For SDR HD content or productions that don’t have 4K HDR finishing budgets, Apple ProRes 422 HQ remains the most widely accepted mezzanine format across platforms. It’s supported by Netflix, Amazon, and most mid-tier OTT buyers as a delivery or mezzanine option, and reduces encoding errors in automated ingest pipelines. But for anything being positioned at premium tiers—originals, co-productions, flagship titles—IMF is where the industry is heading.

Audio Specifications: Dolby Atmos, Loudness Standards, and Stems

Audio rejections are the most common—and most expensive—delivery failure. Here’s what you need to nail.

Loudness. Netflix’s loudness spec is -27 LKFS (+/- 2 LU), dialog-gated, measured via ITU-R BS.1770-1. True-peak must not exceed -2dBFS. This is measured on a 5.1 rerender—not on the Atmos mix directly. Get your mix engineer to set true-peak limiters on all beds and objects at -2.3 or lower to maintain headroom. Miss this and the platform’s QC flags it automatically.

Dolby Atmos delivery. For Atmos mixes delivered in IMF, the supported frame rates are 23.976, 24, 25, and 30fps. Your mix must be done in a 7.1.4 room at minimum—Dolby certification for the room is not required, but the technical environment matters. Deliver as a Dolby Atmos BWAV ADM file. For Netflix originals, delivering a 5.1 rerender is required alongside Atmos; 2.0 stereo is optional. All mixes must be delivered at 48k/24-bit sample rate and bit depth.

Stems and M&E tracks. Platforms with international distribution ambitions—which is most of them—require full M&E (Music and Effects) stems delivered alongside the mix master. These stems enable foreign language dubbing without re-mixing from scratch. For Netflix, the 5.1 Dialog, Music, and Effects stems must sum to equal the 5.1 mix when combined. And if you’re working with the AI dubbing pipeline—companies like Deepdub or Papercup are now integrated into several platform localization workflows—clean, separated stems are what makes that pipeline viable at all. Deliver messy stems and you’re creating a localization bottleneck that costs you territories.

IMF—The Format That Changes Everything About Delivery Economics

If you’re still building your post-production workflow around individual platform deliverables—one encode for Netflix, another for Amazon, a separate one for Apple—you’re building the wrong way. IMF (Interoperable Master Format) is the industry’s answer to that inefficiency, and in 2026 it’s no longer an advanced technique. It’s the standard for premium multi-platform distribution.

Here’s the core idea. IMF acts as a single master from which multiple deliverables are derived, with the requirements for each platform governed by what are called CPLs—Composition Playlists. If the assets are the ingredients, the CPL is the recipe. One IMF package. Multiple platform outputs. One quality control pass. The “create once, publish many” model is the only rational approach to multi-platform delivery at the premium tier.

But IMF isn’t magic—it requires your post pipeline to be built correctly from the start. Your finishing facility needs to be working in an IMF-capable environment, your colorist needs to deliver the Dolby Vision metadata in the right format, and your audio deliverables need to be structured to slot cleanly into the IMF package. This is where post-production workflow planning becomes a distribution strategy conversation—not just a technical one.

Ramy Katrib, CEO of DigitalFilm Tree, has been explicit about this through the Global Post Network: the future of post is data-first and collaboration-first, with cloud infrastructure enabling the kind of real-time verification and hand-off that IMF packaging requires across distributed teams. The finishing facility you choose needs to have proven IMF capability—not just the software installed, but verified hero projects delivered to major platforms through that workflow.

And that’s where the Fragmentation Paradox becomes operationally dangerous. With over 10,000 VFX and post-production companies globally, most producers are sourcing their finishing partners from a network of 4–5 known relationships. The market has 100x more qualified options—including specialized IMF packaging facilities in the UK, Eastern Europe, and APAC that run at significantly lower cost than premium LA and London boutiques. But without verified capability intelligence, producers default to whoever they know and pay the markup.

Leon Silverman (Chair, MovieLabs; former Disney and Netflix exec) breaks down the 2030 Vision for cloud-native workflows and what it means for how producers should be building their delivery pipelines today:

Localization and Metadata: The Hidden Rejection Risk Most Producers Ignore

Here’s a stat worth internalizing: the video localization market is valued at $6.5 billion, driven almost entirely by media and entertainment. And it’s growing—because every platform with global distribution ambitions is localizing faster than ever, with AI dubbing pipelines that can generate multiple language versions simultaneously. But those pipelines depend entirely on what you deliver in your source package.

Subtitles. Deliver subtitles as VTT (WebVTT) files—this is the format most widely compatible across OTT ingest systems. Netflix additionally accepts PAC, DFXP, and STL format for approved subtitle deliveries. Do not burn in subtitles. If you’ve burned forced narrative text (foreign language inserts, title cards) into the picture, you need to also deliver a textless version of those shots so they can be localized. This is non-negotiable for any title destined for multi-territory distribution.

Fonts. List all fonts used in on-screen typography. If you’re using proprietary fonts, those licenses don’t transfer to the platform—the platform’s localization team must purchase their own copy. Not declaring this upfront is a common source of delivery delays that has nothing to do with picture or audio quality.

Metadata. This is where automated ingest pipelines either fly or die. Your metadata.xml must accurately describe every asset in the delivery package—content IDs, language codes, audio track types, subtitle associations, ratings classification by territory. For Apple TV+, content IDs are permanent—they cannot be changed once assigned. Get them wrong and you’re rebuilding the delivery package from scratch. As The Broadcast Bridge’s production delivery reference notes, meeting spec accurately on first delivery is “an increasingly critical business objective” as platform ingest pipelines become more automated and less tolerant of manual correction.

Per the AI localization guide we’ve published elsewhere, the AI dubbing breakthrough has collapsed what used to be a multi-month localization timeline into weeks. But that speed only works if your source M&E stems, textless elements, and subtitle files are delivered correctly from the start. Source asset quality is now the bottleneck in localization—not the technology.

Building a Cloud-Native OTT Delivery Workflow for 2026

The MovieLabs 2030 Vision isn’t a distant roadmap anymore. It’s the operating architecture being adopted by major studios and cascading into requirements for independent producers supplying those studios. The core principles—Zero Trust security, cloud-native workflows, real-time collaboration, and standardized digital distribution—translate into specific infrastructure decisions for producers delivering to premium OTT platforms.

What does this actually mean in practice? Four things:

• Work in the cloud from finishing — Don’t move assets physically between facilities. Cloud-based finishing (via services like Sohonet’s networking infrastructure or CREE8’s cloud workflow platform) eliminates the file transfer delays and version control problems that create delivery errors
• QC continuously, not at the end — Platform ingest QC is automated and unforgiving. Build QC checkpoints into your post workflow at picture lock, at color, and at audio, not just at final delivery packaging
• Build your IMF package first — Design your finishing workflow with IMF as the output target, then derive platform-specific deliverables. Don’t create individual deliverables and try to reverse-engineer an IMF package afterwards
• Verify your vendor’s platform credentials — Not all post-production facilities claiming IMF capability are equal. Ask for specific hero projects delivered to Netflix, Apple TV+, or Amazon through their IMF pipeline—with version history you can verify

This last point is where Vitrina’s vendor qualification intelligence provides real operational value. The difference between a facility that says it can deliver IMF and one that has delivered verified, accepted IMF packages to Netflix across 10+ titles is a difference that doesn’t show up in any sales deck. It shows up in production intelligence—and that’s exactly what Vitrina’s platform surfaces across the 10,000+ post and VFX companies it tracks globally.

Insiders recognize that the biggest cost driver in delivery failure isn’t the re-encode itself—it’s the 3–6 weeks of schedule compression that follows. Your distribution window slips. Your marketing spend loses its anchor. Your talent’s availability for press gets thrown off. The downstream cost of a rejected delivery package, for a title with significant P&A investment behind it, dwarfs whatever you saved by going with an unverified finishing facility.