Make playback
your own.

Optional BOSS directional upscaling is implemented and enabled in Odyssey for eligible progressive SDR video. Automatic and directional modes use the BOSS reconstruction path, with bounded GPU resources and display-aware output sizing.
Source, geometry and device limits apply. HDR, Dolby Vision and interlaced pictures retain their appropriate playback path. MetalFX is a separate explicit option, not the accepted default.

Bounded thumbnail previews follow the touch or TV seek cursor on supported native and owned playback routes.
Coverage and latency vary by source and backend; live previews are limited to available history.

Compatible MKV sources use owned decoding to Apple sample-buffer output or a container bridge to AVPlayer.
These are distinct routes. Not every MKV codec, track layout or Dolby variant uses AVPlayer directly.

Explicit height-channel labels are converted to CoreAudio channel labels on eligible decoded PCM paths.
Negotiated output routing and actual height-speaker playback remain unverified. Channel labels do not establish object-audio or Atmos output.

Finite HTTP range caching and finite HLS resource staging support bounded storage, identity checks and offline reopening of supported sources.
DRM graphs, arbitrary live persistence and all provider configurations are not covered. Host integration controls cache policy.

Native MP4/MOV/M4V/HLS and owned-source preparation support preloading the next video while the current source remains active. Odyssey integrates prepared next-episode handoff and starts automatic Play Next at 0:00.
Preparation eligibility and transition latency depend on source, backend, processing and host lifecycle. No universal zero-delay guarantee.

An app-owned rolling buffer for supported raw TS sources provides bounded pause, rewind and return toward the live edge.
Rewind is limited to retained data. Provider HLS windows, long pauses, discontinuities and recovery have route-specific limits.

Local speech models generate timed subtitles from supported audio, with audio decoding/resampling, subtitle output, track selection and cancellation integrated in the SDK.
The host must provision a compatible local model. Supported media, model languages, transcription accuracy and sustained performance vary; this is not an always-on live caption guarantee.

Dolby Vision playback paths are implemented and used in Odyssey: Profile 5, Profile 7 MEL, Profile 7 FEL enhancement-layer reconstruction and Profile 8. Source identity, RPU metadata and explicit adaptation policy are preserved as separate concepts. The current Odyssey bridge includes the FEL-to-Dolby policy.
Completed denotes the working Odyssey playback implementation, including FEL support. Profile variants, source structure, selected adaptation, OS, device and display route determine output. P7 MEL-to-P8.1 adaptation is explicit. This is not blanket Dolby/HDMI certification, and unlisted profiles are not claimed.

Independent finite audio and video sources can share the playback clock with coordinated startup, seek and shutdown.
Live sources, arbitrary provider combinations and all loop modes are not covered.

Local and authorized network Blu-ray title reading, title/chapter selection and title streaming are implemented in the SDK.
Unencrypted supported Blu-ray structures only. Full disc menus, BD-J, decryption and DVD playback are not part of this feature.

Supported native and owned sessions can retain their playback owner and surface while moving between expanded and compact presentation.
Host navigation integration is required; FEL and compatibility-route coverage is not universal.

MainActor-isolated player control, asynchronous loading, cancellation and source-generation guards underpin the SDK.
Application integration must respect ownership and cancellation contracts; this is not a claim of architectural equivalence to another SDK.

Supported interlaced sources can use a bounded Metal spatial bob deinterlacer.
Dedicated VideoToolbox deinterlacing is not claimed. Cadence, quality and codec coverage depend on valid field metadata and the playback route.

AV1 playback is implemented on supported decode paths, using hardware acceleration where the device and backend provide it.
Codec profile and device limits apply. AV1 playback does not imply hardware decoding on every Apple device.

Authored WebVTT inline timings and ASS karaoke can drive word-specific subtitle presentation.
Plain captions do not acquire invented word timings. Full WebVTT region and CSS behavior is not covered.

Authored per-frame HDR10+ metadata is bounded, syntax-validated and attached to eligible Apple sample buffers.
Attachment does not prove display consumption or dynamic tone-mapping conformance. Device, renderer and sink validation remain separate.

Translation providers can convert timed text while retaining cue timing; local inference and selected embedded-text workflows are available.
Source languages, subtitle types and playback routes have explicit limits. Translation quality varies by model and language.

Supported owned text renderers use system caption font, size, color, opacity, background and edge preferences.
Appearance behavior is not identical across every native, compatibility, HDR and PiP route.

Encoded clip recording is available from supported owned H.264/HEVC playback and the live TS bridge.
Other native HLS/MP4, compatibility, Dolby, independent-input and repeat routes are excluded. Recording availability is explicit.

Eligible PQ video can use subtitle composition with an explicit caption-white luminance policy.
Dynamic Dolby Vision and HDR10+ composition are excluded from this path. No universal HDR subtitle-effects or display-conformance claim.

Selected media downloading, cache staging and packet-copy container conversion are implemented.
General-purpose transcoding, arbitrary format conversion and a complete durable download/export product workflow are not included.

External SUP/PGS subtitles support bounded import, timed display, selection, seek and clearing on supported routes.
Malformed-source, HDR composition and device/backend coverage have limits; not every image-subtitle format is supported.

Primary and secondary subtitle slots have independent selection, timing, layout and cleanup.
Native live/HLS/DRM, secondary closed captions and some PiP combinations are not all supported.

Local inference can translate selected English text cues continuously on the supported owned-primary route without scanning an entire remote file.
Native, compatibility, FEL and bitmap routes are excluded. Sustained real-time performance and language quality depend on the device and model.

Supported SDR output has saturation, brightness and contrast controls with bounded processing resources.
Nonneutral HDR and Dolby Vision picture adjustments are unavailable. SDR picture controls and optional video upscaling are separate settings.

Eligible owned subtitle composition can carry captions into supported PiP output.
Owned sample-buffer PiP is not available on tvOS through the same route; HDR, native and compatibility combinations have limits.

Actual Annex-B hardware decode paths exist, including the dual-layer FEL pipeline.
General raw elementary-stream and low-latency live coverage across all codecs is not complete.

Bounded embedded font attachments feed the actual libass renderer.
Font collections, fallback behavior, untrusted inputs and backend-specific presentation retain compatibility limits.

A bounded, representation-pinned compressed-byte cache can reuse nearby source data during seeks.
Latency depends on indexing, decoder work and provider validation; no fixed seek-time guarantee.

Playback paths cover common MP4, MKV, AVI and TS sources with multiple audio and video codecs.
Backend selection matters. The SDK does not support every possible format, codec profile or damaged file.

Real libass rendering supports styled subtitle images, embedded fonts and authored effects.
Universal ASS effect fidelity, animation, HDR and all-route reference parity are not claimed.

The same encoded recording implementation described by clip recording is available on its supported routes.
This is a duplicate catalogue requirement, not a second recorder. The clip-recording exclusions apply.

4K and eligible HDR/HDR10 playback and output paths are implemented.
Codec profile, device and route limits apply; display mode, color conformance and sustained delivery require appropriate hardware validation.

Native PiP and supported owned iPhone routes have integrated lifecycle and restoration support.
FEL, compatibility and owned tvOS routes are not universally covered; host scene and audio ownership must be integrated correctly.

Supported VideoToolbox decode and Metal processing paths use Apple hardware acceleration.
Acceleration is codec-, device- and route-dependent; software paths remain necessary for some media.

Explicit source field metadata drives supported deinterlacing policy and unsupported-state reporting.
Unknown field order is not guessed. Wider codecs and physical field cadence need route-specific validation.

Up to eight PCM channels and CoreAudio channel labels are supported on eligible routes.
Output depends on the device and selected audio route. Multichannel PCM and channel mapping are distinct from receiver-confirmed Atmos or object-audio passthrough.

Bounded authenticated SMB and trusted-network NFS file access, identity checks and directory browsing are implemented.
NAS compatibility and credential workflows depend on host integration. NFS AUTH_SYS is limited to trusted networks.

Text, styled ASS, image subtitles and supported closed-caption paths are implemented.
Selection, style and composition coverage varies by format and backend; not all combinations are available.

The owned playback path has coordinated audio/video looping within a selected range.
Other backends and perceptually seamless long-running boundaries are not universally verified.

Adapters for OpenSubtitles, ASSRT and Shooter feed supported subtitle import and selection workflows.
Provider accounts, quotas, configuration and service availability are external requirements; archive extraction and some delay cases are unsupported.

Native HLS uses AVPlayer adaptive variant selection, with exposed buffering policy and observed downshift and recovery behavior.
Provider/CDN conditions, waiting and stalls remain workload-dependent. No universal uninterrupted or fixed adaptation-time guarantee.