This library is under heavy development!
This Swift Package extends OpenTimelineIO's Swift Bindings to provide functionality for Apple platforms.
The goal is to enable easy interoperability OpenTimelineIO and AVFoundation objects in a correct, lossless and useful manner to enable real world workflows.
- Load OpenTimeline IO files into OTIO
Timeline
objects and introspect using the standard Swift bindings. - Create AVFoundation
AVComposition
AVVideoComposition
andAVAudioMix
objects from an OTIOTimeline
instance allowing for playback viaAVPlayer
and export viaAVExportSession
orAVAssetWriter
to Quicktime compatible file formats. - Convert your own edits as
AVMutableCompositions
into a OTIOTimeline
This library should be compatible with the following Apple platforms:
- macOS
- iOS
- visionOS
but to date has only been extensively tested on macOS
-
The app should register itself as a reader of OTIO files.
- In theory should be able to double click a
.otio
file which will launch the app, or control click with open as.
- In theory should be able to double click a
-
To ensure the widest media support possible, please make sure to install Apple's Pro Video Formats package.
- If you have Final Cut Pro X, Compressor or Motion, this step is unnecessary.
- Please note that not all video formats are supported, not are image sequences. See Video Format Compatibilty
-
Thats it!
Assuming you have a basic AVPlayer
setup, this will let you import a .otio
file with basic jump cut editing.
See roadmap for transitions / effects.
do {
if
let timeline = try Timeline.fromJSON(url: url) as? Timeline,
let (composition, videoComposition, audioMix) = try await timeline.toAVCompositionRenderables()
{
let playerItem = AVPlayerItem(asset: composition)
playerItem.videoComposition = videoComposition
playerItem.audioMix = audioMix
self.player.replaceCurrentItem(with: playerItem)
}
}
catch
{
print(error)
}
Assuming you have succssfuly created an AVCompostion
- this will export a basic .otio
file without effects or transition metadata.
See roadmap for transitions / effects.
do {
let timeline = try compositon.toOTIOTimeline(named: toURL.lastPathComponent)
try timeline.toJSON(url: toURL)
}
catch
{
print(error)
}
- Conversion of OpenTimelineIO
RationalTime
to CoreMediaCMTime
- Conversion of OpenTimelineIO
TimeRange
to CoreMediaCMTimeRange
- Conversion of OpenTimelineIO
ExternalReference
to AVFoundationAVAsset
- Conversion of OpenTimelineIO
Timeline
to playable/exportable AVFoundationAVCompostion
AVVideoCompostion
andAVAudioMix
- Conversion of CoreMedia
CMTime
to OpenTimelineIORationalTime
- Conversion of CoreMedia
CMTimeRange
to OpenTimelineIOTimeRange
- Conversion of
AVCompositionTrackSegment
to OpenTimelineIOClip
with an embedded OpenTimelineIOExternalReference
which has url metadata - Conversion of
AVCompositionTrack
to OpenTimelineIOTrack
with eachAVCompositionTrackSegment
converted to OpenTimelineIOClip
associations - Conversion of
AVComposition
to OpenTimelineIOTimeline
with associatedTracks
converted
OTIO Formats | Status | Notes |
---|---|---|
OTIO | ✅ | |
OTIOD | ✅ | |
OTIOZ | ❌ | Planned |
NLE | Import from NLE | Export to NLE |
---|---|---|
Davinci Resolve 18.6 | ✅ | ✅ |
Davinci Resolve 19 | ✅ | ✅ |
Sample Project | Import to AVFoundation | Export from AVFoundation |
---|---|---|
ALab Trailer | ✅ | ✅ |
AWS Picchu Edit | ✅ (Decompress .otioz zip file and open the bundled .otio file) | ✅ |
OTIO-OC-Examples | ✅ | ✅ |
Generally if Quicktime or Final Cut Pro X Can support it, it should just work.
Video Formats | Status | Notes |
---|---|---|
H.264 | ✅ | Native |
HEVC | ✅ | Native |
Apple ProRes | ✅ | Requires Apple Pro Video Formats for some variants |
Apple Intermediate Codec | ✅ | Requires Apple Pro Video Formats for some variants |
MXF wrapped Pro Res | ✅ | For developers: Requires MTRegisterProfessionalVideoWorkflowFormatReaders() and VTRegisterProfessionalVideoWorkflowVideoDecoders() to be enabled |
MXF Wrapped DNx | ✅ | Requires Apple Pro Video Formats for Avid DNxHD® / Avid DNxHR®. For developers: Requires MTRegisterProfessionalVideoWorkflowFormatReaders() and VTRegisterProfessionalVideoWorkflowVideoDecoders() to be enabled - only some DNx variants work: |
MXF Wrapped DNxHD_36 | ✅ | |
MXF Wrapped DNxHD_80 | (codec specifically gets subsampled down to 1440 x 1080) | |
MXF Wrapped DNxHD_115 | ✅ | |
MXF Wrapped DNxHD_175 | ✅ | |
MXF Wrapped DNxHR_SQ | ✅ | |
MXF Wrapped DNxHR_LB | ✅ | |
MXF Wrapped DNxHR_HQ | ✅ | |
Image Frames | ❌ | Requires custom compositor |
Image Sequences | ❌ | Requires custom compositor |
Raw Formats (BRaw, Red, etc) | ❌ | Requires you to have SDK - manage decode and roll your own custom compositor |
- OpenTimelineIO's Swift Bindings for Swift interoperabiloty
- TimecodeKit for reading and parsing Time Code from AVAsset's
-
Enable viable metadata from
AVAssets
toExternalRefernce
and other objects where appropriate to faciliate correctness and robustness in conversion.- Waiting on #51
-
Enable robust support for transitions / effects metadata to be passed between AVFoundation and OTIO
- This requires thinking deeply how to support transitions and effects, and the infrasturcture required (custom instructions, composition renderers, etc).
- Input welcome!
- Why does RationalTime seconds, rate and value differ from a converted CMTIme?
RationalTime uses double as numerator and denimonator for rational time calculations. CMTime uses Int64 and Int32 for its value and time base. In order to get most accurate conversions, OpenTimleineIO-AVFoundation uses looks at the number of decimal places in a RationalTime representation and scales it to a whole number to maximally represent the same ratio without rounding or truncating which would happen with naive Double
<-> Int
casting.
See our Unit tests for conversion notes. In general, for integer frame rates (ie 120, 60, 30, 25, 24) there is zero loss conversions. For non integer frame rates like 59.97, 29.97, 24.98 (23.976) there may be very minor differences in the conversions as of today (our Unit testing uses an accuracy factor of 0.00000000001
)
If you have ideas on lossless conversion, PR's are welcome! This was a first pass :)
OpenTimelineIO and TimecodeKit, and Ozu.ai for supporting this effort.