-
Notifications
You must be signed in to change notification settings - Fork 293
OpenTimelineIO Application Integrator's Guide
NOTE: This documentation is work in progress and will be in a high state of flux. Once it has stabilized we'll remove this notice.
OpenTimelineIO (OTIO for short) is an interchange format and API for editorial cut information that strives to be a superset of data models like EDL, ALE, AAF, and FCP 7 XML while providing a data and mental model that is accessible for people familiar with NLE (Non-Linear Editing) software.
OpenTimelineIO's key benefits over other formats include:
- Straightforward modeling of common timeline/editorial constructs: Timelines, Tracks, Clips, Effects, Transitions, and bins (as SerializableCollections).
- The ability to store nested JSON-like metadata on nearly any object in the hierarchy
- Well supported APIs that allow users to read/create files using python, C++, C, and Java (with more coming!)
- An open source community constantly improving the format and supporting users
With OpenTimelineIO an application can give users can get the rich data they want in a way they can easily access it using one of the provided language bindings and users can generate rich data for applications to leverage.
Integrating OpenTimelineIO with your applications allows for clearer and more accessible communication about creative decisions and technical metadata through production, finishing, and delivery. Preserving information like lens metadata, rough color grades, and notes from the director are just a few ways OpenTimelineIO can have direct impact on the filmmaking process. What metadata does your application generate that could preserve creative vision down the pipeline?
TODO: give an overview of the sections in this document
This document aims to cover common elements used in the OTIO schema and some typical patterns for how they're composed - there is more depth to the full schema than is covered in this document. The OpenTimelineIO Data Model Specification covers the schema objects in much deeper detail.
Applications integrating OTIO should rely on one of the implementations provided by the project to read and write .otio files, but to illustrated the core concepts, we'll include some JSON serialization as examples.
As a developer, you know your application best and only you will be able to identify the exact ways OpenTimelineIO will make your application more valuable in user workflows. This guide can, however, can help with some ways to think about what OTIO is and some of the unique benefits it provides.
The benefits of OTIO come from the fidelity of editorial data and metadata it communicates and ease of access to that data. What key information or creative decisions are captured or made within the application (e.x. edit points, color, notes/markup, camera report or script continuity data, etc.)? Is there somewhere data is getting lost or degraded when moving through the workflow? Where do users wish they could write scripts or automations to interface with the application?
Existing formats imported or exported from an application can give some hints about how interchange could be improved with OTIO. In what ways are .edl files overly simplistic for user workflows? Is there some metadata that isn't making it in our out of the application because FCP .xml don't carry it? Are users having troubles deciphering data in .aaf files? Are .ale metadata fields feeling overloaded or under-expressive? Are there custom XML, CSV, or JSON files that the application is reading/writing? Using OTIO in place of these formats can help users access data quicker and with fewer errors by allowing them to avoid writing custom parsing code.
Adopting OTIO as a replacement for EDLs or ALEs is a clear win but integrators should also challenge themselves to think outside just the common use cases. Because OTIO has an extensible schema and hosts rich metadata, its timelines and other editorial objects can be used to carry data not commonly house in other editorial formats.
Some more interesting (often experimental) examples of data that's been hosted in OTIO include:
- Timed Text
- Preservation of lens metadata
- Metadata from a screenplay (scene, location, characters, etc.)
- Markup, annotation, and notes
OTIO is really great at expressing how objects exist temporally - so anywhere there is a timing component to the data, it might be a good fit for OTIO.
If you think OTIO might be a good fit or want to talk to bounce ideas off OTIO developers and users, the OTIO community provides a few good avenues:
- The #OpenTimelineIO channel on the Academy Software Foundation Slack (registration required) hosts a lot of great conversation
- The project GitHub Discussions is a good place to ask questions too
OTIO files are designed to be human readable JSON (with some more explicit constraints around numeric types) and work is in progress on the OpenTimelineIO Data Model Specification to ensure they are easy to understand and useful beyond years to come and in to archival status.
While the specification provides detail about specific schema objects, their usage, and semantic meaning; integrators should avoid creating their own implementation unless there is a really compelling reason to do so. The core implementation provides a lot of subtle behavior that can be time consuming to reproduce and could impact an application's ability to effectively interoperate with OpenTimelineIO.
Beyond the main C++ implementation with Python bindings, language bindings are available for:
OTIO's implementation is in two layers:
- The serialization scheme for storing structured data along with a schema name and version for interpreting the data
- A set of schema which define types of objects and their serialized data fields
When interacting with OTIO, integrators will be dealing almost entirely deserialized schema objects. The implementation will handle mechanics of translating between in-memory objects and their serialized form. The OTIO implementation also provides a rich set of functionality for querying and modifying timeline data.
In addition to the "first-class" fields offered by OTIO schema objects, every object in OTIO has a field called metadata that allows applications and users to attach any arbitrary metadata they like as a nested dictionary structure. This is a unique feature of OTIO and provides some powerful uses in a production pipeline. Studios attach identifiers for internal production tracking systems, adapters store format-specific contextual metadata about the file they derived the OTIO from, and some users even attach metadata like source script page number or notes from a review system. The metadata dictionary is also a powerful tool in driving the evolution of first-class schema in OTIO.
OTIO allows for any arbitrarily deep collection of other dictionaries, lists, scalars, and even SerializableObjects in the metadata dictionary - the implementation enforces no structure, but there are some strong conventions implementers should follow.
Using a unique namespace helps avoid collision with metadata populated by others. The "namespace" key in this case is just the top-level key in the metadata dictionary under which an implementor places the dictionary containing their data. These top-level "namespace" keys denote informal sorts of sub-schema.
For example, a clip may have a metadata dictionary like the following:
{
"myAnimStudio": {
"sequence": "ABC",
"shot": "ABC010",
"shot_id": "7b3aaa14-8305-4fdd-87c2-b0b9d3f9dac7"
},
"cdl": {
"asc_sat": 0.9,
"asc_sop": {
"offset": [
1,
-0.0122,
0.0305
],
"power": [
1,
0,
1
],
"slope": [
0.1,
0.2,
0.3
]
}
},
"cmx_3600": {
"comments": [
"SOURCE FILE: ABC010.LAYOUT3.01"
]
}
}Note that information for studio production tracking, CDL values, and a comment field from the original CMX 3600 EDL all coexist as separated by their namespace keys.
In reading and writing OTIO, make a best effort to preserve the data in metadata dictionaries. If possible, only mutate the data under the namespaces your application controls. This ensures your application will be a "good citizen" and metadata populated upstream of your application will flow downstream through the pipeline without loss. In the previous example, the identifiers under the myAnimStudio namespace provide critical value to the pipeline. Likewise, creative decisions are stored in the cdl namespace, preserving them helps us ensure we maintain the artistic vision.
Part of OTIO is that all schema objects are serializable. .otio files are a JSON serialization of an object tree from OTIO with one root.
While there are no current plans to create alternate serialization formats, the implementation is built to allow creation of a new serialization format if needed. If this were to happen, a new file extension would be used.
Any object available in the OTIO schema can be serialized and de-serialized as the root object in a .otio file. This means implementers should check what kind of schema object results from an OTIO file and handle it appropriately. This also means there is a great degree of freedom in what kind of data can be communicated with OTIO: whole timelines, bins, and individual clips are among the possible objects that can be root objects in OTIO files.
While any OTIO schema object can be serialized at the top-level, most OTIO files will have either a Timeline or SerializableCollection as the root object.
For instance, in cases where older workflows might use an .edl file, an .otio file with a Timeline would be used. In cases where an .ale file was used, an .otio file with a SerializableCollection of Clip objects would be used.
OTIO has no schema for a project, this would typically be represented as a SerializableCollection containing Timelines, other SerializableCollections for folder or bin structure, and Clips for various media in the project.
OTIO offers a lot of the essentials for the most common types of cut information used in pipelines, but the team is also actively expanding the schema to enable more and more use cases. When possible, schema development is driven by real-world use in production.
The file format and implementation provide a number of useful tools to help implementors easily progress with OTIO's evolution.
Each object type has an independent schema version that is stored as an integer that increments each time there is a breaking change. This means non-breaking changes, like addition of fields, won't interfere with loading in older implementations, they just may not be available.
When the library encounters a serialized object with an older schema, it automatically migrates that object to the newest available schema, even "up-converting" data if needed. This means application implementors only need to concern themselves with handling the latest schema version when working with OTIO objects.
metadata dictionaries are a good place to store information that is not yet part of the "first-class" schema
Wherever possible, implementors should try to use the first-class fields available in in OTIO. However, when the schema does not yet represent valuable data, making it available early in a metadata dictionary allows pipelines to start using it and aids in the design of an "first-class" schema to be added to OTIO. Most new schema in OTIO starts as some extra information in a metadata dictionary, once it has been proven useful and the form has been refined, an update to OTIO will then promote that to new "first-class" schema.
Upgrading the OTIO release used in your software ensures you have maximum read compatibility with other applications. There are currently no mechanisms provided for downgrading the schema versions used within an OTIO file, so moving forward with the format helps users from having to make special concessions for your application.
OpenTimelineIO's schema provides familiar objects meant to directly correlate to the units editors interact with in an NLE while providing advanced compositions. Below are some of the most common schema objects and some guidance on using them.
Timeline objects represent one edit, sequence, or presentation. Timelines are named - as you might see a sequence named in an editor's project.
Using an appropriate namespace key, the Timeline metadata attribute is a great place to store timeline/sequence settings such as edit rate, video resolution, and audio sample rate. In the future many of these concepts will be given first-class attributes.
The global_start_time field on a Timeline object communicates the offset value for the start of the timeline. This is how you might set a start timecode of 01:00:00:00.
SerializableCollection objects store an ordered collection of other OTIO schema objects. They are commonly used to represent bins or folders in a project and can be nested to create deeper project structure. The name property corresponds to the name you see for the folder or bin within a project.
When creating an editorial composition, there are two important relationships items have to one another, sequential presentation (clip B comes after clip A), and simultaneous presentation (clip B is composited over clip A). In OTIO, a Track is used to compose items in sequential order and a Stack is used to compose items that are simultaneous. A Timeline object's tracks attribute references a Stack containing the timeline's audio, video, and other Tracks.
This Stack of Track items is directly analogous to the timeline interface where editors compose their clips in NLE software.
Commonly the only Stack used in an OTIO document is the tracks stack on the timeline. The children of the Stack is an ordered list of all the items (typically Tracks) that are coincident in "compositing" order.
In general, it is assumed that visual media uses alpha compositing, overlaying items one over the other. Audio media is assumed to use additive compositing, mixing all the tracks together so you hear them all. If another compositing method is used, somewhere under your application's namespace key the metadata on the Stack is a great place to store that information.
In addition to providing items in sequential order, a track has a kind. The track kind defines what type of output media the track composes. Track kind is deliberately a string. The values Audio and Video are provided via constants for the track kinds often used. Values outside the provided constants can be used, but it is expected that most application implementers will ignore tracks of a kind they don't understand and there may be certain utility methods that don't handle tracks of unknown kind.
The items in a timeline are either Clip or Gap schema objects. Clip objects specify media to use in the composition, whereas Gap objects can be used to offset Clip objects temporally.
These Items are single-use, meaning any given Item instance can be used in a composition once. The APIs provide ways of copying Items to create any number of identical instances for use multiple times in a composition. One way to think of items is as if each one were a strip of film being taped end-to-end in a track. A Gap is like inserting clear leader of some length, and a Clip is like using film from the camera - if you need to use the same strip of film twice, you have to make a copy of it.
A Clip represents an instance of using some subsegment of media in the timeline. This is directly analogous to clips you see in a standard NLE.
The source in and out points on a Clip are set using source_range. The source_range is specified as the time range within the the source media's available_range to select.
The media_reference on a clip specifies how to locate media assets composed in the timeline.
Gap
Gap represents an absence of media - in other words, a Gap does not contribute to the composition media, it only serves to offset other items in time - it can be thought of as clear filler.
OTIO doesn't embed media assets, instead it provides a way for applications to locate the appropriate media. Subclasses of MediaReference provide various methods for identifying the media composed within the timeline.
Use the media's global start time when setting available_range. When the start_time of the available range is the start timecode or first frame number of the media, applications are able to locate the correct media within a clip through a range of situations - especially when switching between representations that are trimmed with different handles.
When a specific file path/url is available for an asset, use one of the following reference types:
-
ExternalReference - Identifies an asset by URL (local files use
file://urls) - ImageSequenceReference - Identifies an asset as a set of sequentially numbered images
When there isn't a specific file path, use MissingReference. Any metadata that might help downstream consumers relink the media should be included in the reference's metadata dictionary.
If a clip's media is generated programmatically (e.x. solid color, white noise, color bars, tone, etc), a GeneratorReference should be used.
Following is a compilation of the guidelines scattered throughout the guide as a reference:
- Use a unique "namespace" for data your application stores within object
metadatadictionaries. - Preserve metadata to the best of your abilities.
- Check the type of top-level object when deserializing.
-
metadatadictionaries should be used to store important information that is not yet a "first-class" part of the OTIO schema. - Keep up-to-date with the OTIO implementation.
- Find all schema object name references and link out to appropriate documentation
- Create illustrations showing examples of the schema objects, maybe from otioview?
- Continue reorganization pass through specific schema objects section