ARCHITECTURE

Full marks here as Limecraft appears to offer a market-leading architecture for their product offering, built using a modern DevOps technology stack, giving media operations flexibility, scalability and robustness. The Limecraft platform has a service orientated architecture (SOA) and is microservices-based and dockerised/containerised. With relational, non-relational and graph databases, and using elastic search for indexing and searching, it has the core underlying technology to provide fast responses on different types of searches that other solutions that only use one database are unlikely to achieve. Edge is always installed on a customer's premises, usually on a laptop for small volumes, or an on Apple Pro or a beefed-up workstation for heavy-duty processing. Flow runs in the cloud, which can either be hosted by Limecraft or deployed on the customer provided cloud.  The solution is machine resource, human resource and service location-aware, so is able to make decisions on which to use when for optimal efficiency and minimal cost. It also claims to be able to automate machine resource scaling, with the ability to spin up additional machine resources on demand. Limecraft software can be single or multiple tenanted. Any user can securely share material and or metadata using symbolic links. Each share comes with specific access permissions (playout only, download, re-upload, commenting, verification). The forte of Limecraft lies in its ability to support customised workflows (ingest any format from any source, archive, restore from the archive, export in any format to any destination), and to keep track of all the versions throughout the lifecycle of the media.

​

INTEGRATION

Limecraft informs us that their platform is 'designed for maximum integration' with a flexible integration framework with plugins/connectors to a good selection of modern on-premise and cloud resources. In addition to the expected storage, transcoding, AQC and AV processing resources, Limecraft has experience in integrating with CRM, Rights Management, e-Commerce, scheduling and traffic systems, facilities management systems, HSMs, other manufacturers MAM, VOD and NRCS. In particular, Limecraft seamlessly integrates 3rd party cloud storage services operated by their customers, including but not restricted to Microsoft Azure and AWS. Limecraft has a framework agreement with an infrastructure provider that can easily make available capacity increments for transcoding, transcription, shot segmentation and similar services. In each contract, the holistic/demonstrated throughput is specified and the minimum ensured capacity. For example, a customer has the right to ingest up to X hours of new material per month, and a fair use policy is applied. The platform applies an intelligent load balancing method to distribute the available capacity of said services. Until recently, Limecraft did not offer integration to live processing or linear systems, or legacy SDI infrastructure. This means it doesn't natively integrate with VTRs, Flexicarts or linear playout systems (hence the dropping of one star in this category review), but they recently announced support for processing live feeds, albeit with a significant delay due to the cloud-based nature of the solution and the need to perform complex operations on the images.  Live processing includes transcription and the ability to create shot lists for post-production.

​

AUTOMATION

Limecraft doesn't offer a graphical workflow designer, which we believe is a very important feature in a platform's automation capabilities. Having said that, Limecraft claims that their approach of designing and configuring workflows using a structured language (JSON style) offers more detailed control over graphical user interfaces. In the end, system engineers will want to access the details anyhow, so this probably comes down to preference. If you have a highly technical team that has the experience and ability to use structure language, then this may indeed be more flexible than a graphical workflow tool. If you want less technical / more operational orientated users to be able to design and edit workflows, then this may not be the solution you need. One technology enabler we look for evidence of is the ability to allow media organisations to write their own scripted nodes in a workflow. This enables custom business logic and resource integrations to be added by the media organisation. For this, Limecraft supports service integrations (scriptable sequences of actions directly issued by the Limecraft Flow instance) or remote service integrations. In the latter case, local services or ("Cloud Connector") are deployed that are responsible for remote execution of certain tasks. Limecraft's platform supports several aspects of ingest automation - audio sync, automatic colour correction (configurable LUTs), automatic grouping of clips (Multicam) and on the mastering side, they support automated reformatting, subtitling, localisation and they are currently implementing Audio Description (AD). With regards to automated resource scheduling, the Limecraft platform supports timed actions, scheduled actions (using an internal calendar) and real-time reallocation of tasks (based on exceptions), but does not natively provide facility and equipment scheduling (for this it integrates with third-party systems - see Integration above). It also monitors machine resource queues and can automate the spinning-up of additional machine resources on-demand, both in the cloud and on-premise.​​​​

​​

METADATA

The Limecraft platform has good metadata capabilities. All the usual tools asset metadata definition tools you'd expect from a good MAM are there, including asset, object and other file definitions, and temporal metadata with definitions (although no logging button support, fixed or custom).  A nice spatial metadata feature on videos is available where you pin remarks to a spatial point associated with a specific fragment (TCIN-TCOUT). User-defined hierarchies (such as production/season/episode/version) are supported. Within a single account you can use several "productions" or libraries; use the navigation bar on the left side of the UI for collections, episodes, or work orders; and the grouping in the central view can be pivoted according to different presets (per type of user within the same production). Custom taxonomies and thesaurus are supported. Existing taxonomies can be used to feed navigation elements, dropdowns of custom fields or hashtags, or existing online repositories of entities can be referenced (e.g. Wikipedia). Human and Resource event metadata is captured, enabling tracking of actions and tasks. A built-in chat capability is a nice touch, with chat history kept. Like with so many of the systems reviewed, is built-in assets rights management and asset expiration date warning. Limecraft informs us that most of their customers manage rights by adding confidential documents and extracting certain fields to they can be used for business process automation. Detailed tracking of the different versions of media objects (instances) and complementary resources and their history and lifecycle is supported.​​

​

USER EXPERIENCE

The Limecraft platform user interfaces shown to us are very crisp, light and modern, differentiating themselves from most MAMs that tend to opt for the dark 'AVID' look and feel as standard. Having said this, Limecraft Flow and Edge, have the ability to 'inherit the look and feel' that the customer requires with a white-labelled approach to the UI design that makes it possible for large media organisations to create their private cloud. This also assumes delegated access control and capacity management. While widgets are customisable and component-based, such as parts of the UI (player, transcriber, subtitler, display of search results) and embeddable by third-party application developers, this does require skilled engineers to execute. The search and preview functionality all seem very powerful in the Limecraft platform, with both basic and advance search features. As mentioned above in the Architecture section, multiple types of databases and elasticsearch are used to optimise search speed and, as mentioned in the Information category, Limecraft supports customisable taxonomies and thesaurus.  The only search functionality area that seems missing compared to some other systems is AI-driven search recommendations. The Limecraft platform supports remote editing (solo editing), whereby the editor works from home on a cut-cut non-linear editor in the form of a storyboard editor that can export to a full-featured craft editor. This enables editors to cut on low-res proxies, which are automatically relinked using the original or high-resolution footage afterwards. The latter is a lights-out operation in the datacenter. There is currently no voice-over recording capability, no audio track selection and swapping, no object aware cropping, and no automated background removal. ​​​Since last year’s review, Limecraft included various tools for basic image processing. These include rotation and cropping for stills, as well as cropping, aspect ratio conversion, clipping and watermarking of moving pictures. In general, these are applied logically, and upon export, the user can opt to render the modifications.

​

ANALYTICS

The Limecraft platform supports daily/weekly/monthly reports of usage, capacity consumption, and human resource usage. Proactive messaging is supported if the account reaches certain high watermarks. All the tasks executed within a production can be reported on and grouped by a specific department or account. For visualising data in charts, so that trends and bottlenecks can be more easily seen, Limecraft relies on the third-party Datadog tools rather than having built-in chart capabilities, but Limecraft informs us though that built-in fixed and or configurable charts are on the roadmap. In case of anomalies such as 'exotic' file formats, circled shots, and metadata exchange status info, Limecraft use labels on top of thumbnails. For work orders, Limecraft use colours and icons to indicate the state of the workflow (waiting, in process, waiting for approval,...). Monitoring for metadata and encoding errors is supported, with customisable actions on detection (event handling). The number of errors is not currently monitored and broken down per user though. Financial tracking and reporting are supported, including cost tracking, cost dashboards, configurable billing information, historical cost vs budget etc. None of the systems we've reviewed covers all areas of analytics. For example, missing from the Limecraft analytics capability is any monitoring and reporting of quality of service across the supply chain (supplier quality issue tracking, delivery time vs target, delivery quality issue tracking), some human resource reporting (task queue length, human resource usage by location, by capacity), per program time to air and asset usage versus asset rights. ​

​

SECURITY

Limecraft appears to have taken security seriously, hence the strong score in this category. They inform us that they have written security documentation available and although we do not claim to be security experts, most security best practices we were looking for evidence of seem to be covered. At a coding level, Limecraft has indicated that secure coding and verification best practice is in place, including format-string attack prevention and buffer & integer overflow protection. Databases are kept on separate network segments, security focussed open source libraries are used, and transport layer security (TLS) is used for secure transfers. Resource quotas per container, run at root prevention and secure container registries are also implemented. For user authentication the Limecraft platform has secure centralised authentication, applying the principles of least privilege and authenticating first/authorising next. Functionality-based access control is achieved through secure password entry with a recovery mechanism. Single Sign-On (SSO) mechanisms and multi-factor authentication are fully supported. While inactivity timeouts are supported in the native app (Edge), the browser applications (Flow) do not deactivate themselves by default. Re-authentication for sensitive material is not. From a logging and forensics perspective, Limecraft log timestamps with source IP, monitors for suspicious IP addresses, and have an extensible logging framework. Exceptions are logged, handled centrally and any security breaches manually alerted as described in their Data Processing Agreement. Automated vulnerability scans and container image signing are also implemented. With regards to prevention, the Limecraft platform includes injection and input prevention including query parametrisation/binding variables, object query limitations, whitelists for syntax checking, server-side input validation. Its RESTful API has appropriate security measures, including hiding all API security clues.​​​

​​​

ARTIFICIAL INTELLIGENCE

When reviewing AI/ML we are looking for evidence of the platform automating content identification, recognition and detection of attributes in the content, creation and placement of content, technical processing, and operations management. In the area of content identification and association, the Limecraft platform has the ability to track content similarities. When using Limecraft Edge for ingesting content an identifier is inserted into the header of each file which can be used for similarity detection and version tracking. The identifier combines different visual and audio features to ensure identity. For video analytics such as facial recognition, voice detection and visual text extraction (OCR), the Limecraft platform applies a best of breed strategy. It preferably acts as a broker, integrating Microsoft Video Indexer, Google Vision and Amazon Rekognition, and it also integrates AI aggregators (Valossa and GrayMeta). Apart from these, Limecraft ensures access to specialised services such as Mobius labs, Speechmatics and Vidrovr. Doing so, Limecraft effectively hides the underlying complexity of the different point solutions and helps to avoid vendor lock-in. Limecraft is currently working on audio-based language detection, considered one of the key deliverables of a Horizon 2020 project that ended in March 2021. Limecraft uses a proprietary and patented matching algorithm to automatically cut shots into scenes.  Clusters of meaningful concepts from the (tran)script are used to generate unique semantic identifiers to automatically cut shots into scenes and to identify 'different points of view'. Limecraft informed us that about a quarter of their clients, those that create fiction content, use this technology. The Limecraft platform support AI-powered technical processing, including smart encoding, smart up/down conversion, technical quality checking, transcription, multi-language translation and contextualised captioning.