On Sunday 15th October, Yella Umbrella released versionV6.2, a major update for our Stellar AD, Dubbing and Subtitle Timed Text editing tool.
We have added a range of new features for subtitling authoring, AD and Voice Over recording and QC to Stellar. These changes make Stellar a web based subtitling tool with all the feature of existing PC and Mac applications – and more.
In Version 6.2 you can now directly edit a wider range of subtitle file formats including the new IMSC-Rosetta format, the Cavena .890 format, .scc closed caption format, the Ooona ttml format, the LammdaCAP file format with vertical Japanes and rubies support, and the .ass Aegisub format. Direct editing removes the risk of converting between file formats when making small changes. In Stellar you can just open the file, edit and save.
Now that Stellar supports over 2,500 synthetic voices you can organise these into favourites lists that can be shared within your company. You can also now assign a synthetic voice to each Role (Character) in a dubbing file making it easier to produce professional dubbing and voice over with no need for a recording studio.
You can now use the ChatGPT AI engine to rephrase subtitles in most languages to fit into the available reading time. This is invaluable when translating into languages that tend to have longer words, like from English into German.
We have improved the ability to mark, cut, copy and paste groups of subtitles within and between files. Useful when manipulating files with common sections such as the intro section in a series which is the same in every episode.
You can now see the hotkeys for most actions just by hovering over the control.
Note: Layouts will move from the Files menu to the Settings menu
All the new features are covered in the Help Manual.
On Sunday 24th September, Yella Umbrella will release versionV6.2, a major update for our Stellar AD, Dubbing and Subtitle Timed Text editing tool.
We have added a range of new features for subtitling authoring, AD and Voice Over recording and QC to Stellar. These changes make Stellar a web based subtitling tool with all the feature of existing PC and Mac applications – and more.
In Version 6.2 you can now directly edit a wider range of subtitle file formats including the new IMSC-Rosetta format, the Cavena .890 format, .scc closed caption format, the Ooona ttml format and the .ass Aegisub format. Direct editing removes the risk of converting between file formats when making small changes. In Stellar you can just open the file, edit and save.
Now that Stellar supports over 2,500 synthetic voices you can organise these into favourites lists that can be shared within your company. You can also now assign a synthetic voice to each Role (Character) in a dubbing file making it easier to produce professional dubbing and voice over with no need for a recording studio.
You can now use the ChatGPT AI engine to rephrase subtitles in most languages to fit into the available reading time. This is invaluable when translating into languages that tend to have longer words, like from English into German.
We have improved the ability to mark, cut, copy and paste groups of subtitles within and between files. Useful when manipulating files with common sections such as the intro section in a series which is the same in every episode.
You can now see the hotkeys for most actions just by hovering over the control.
Note: Layouts will move from the Files menu to the Settings menu
All the new features are covered in the Help Manual.
IMSC-Rosetta: A New Era for Subtitle Formats – Bridging Broadcasting and Streaming
In the realm of media, delivering subtitles consistently across various platforms has posed challenges. Warner Bros. Discovery (WBD) identified the pressing need for an innovative subtitle format. This format should seamlessly suit both conventional TV broadcasts and contemporary streaming services. Historical subtitles have been fragmented, existing in diverse proprietary and generalized formats. However, none of these formats proved universally fitting for all content types and languages.
In response, WBD partnered with Yella Umbrella, a company with proven extensive experience in subtitle formats and linguistic support going back over 30 years. Their joint endeavour aimed to create a fresh subtitle file format that resolves these complexities. After evaluating multiple options, they selected IMSC V1.2 as a foundational framework. While IMSC V1.2 provided the framework, its inherent variability posed challenges in achieving consistent and reliable outcomes. The heart of the matter was devising a method to harness TTML’s potential while enforcing a standardized representation of subtitle text, timing, and style information, thereby eliminating the complexity of adapting generic TTML files for diverse purposes.
The outcome of this collaboration culminated in the IMSC-Rosetta subtitle file format. This innovation draws inspiration from IMSC V1.2, while streamlining the structure for clarity and uniformity. This rectifies challenges tied to XML-reliant formats, well known for their intricacy and incompatibility. IMSC-Rosetta champions ease of use, rendering parsing, modification, and creation accessible without necessitating mastery of exhaustive technical minutiae of XML and TTML.
IMSC-Rosetta retains the full spectrum of features seen in alternative subtitle formats, encompassing colour, outlines, boxing, and text placement. Its distinctiveness lies in its definitive construction, facilitating seamless translation across disparate formats and languages via the IMSC-Rosetta standard. For entities adhering to proprietary formats, transitioning to IMSC-Rosetta guarantees minimal feature loss and significantly reduced development efforts, in contrast to the more intricate IMSC or TTML routes.
IMSC-Rosetta’s versatility extends across a range of applications, serving as a solution for authoring, delivery, interim storage, and archiving. It stands out for catering to translation requirements, preserving the nuance created by subtitlers – often compromised during conversion processes.
The necessity for IMSC-Rosetta emerged from observations concerning the slow adoption of TTML. Despite its merits, complexities in implementation spawned an array of TTML-based formats that operated effectively as proprietary standards due to the demanding nature of implementation and the variable interpretations of standards. Existing implementations often overlooked subtleties that elevate the viewer’s experience, as compatibility with distinct media and streaming platforms took precedence.
IMSC-Rosetta emerges as a remedy for these disparities, streamlining development while setting forth a pathway to quality within a comprehensible, standards-compliant subtitle file format. Unlike many TTML-derived derivatives with limited reusability, IMSC-Rosetta catalyses change by offering seamless conversion to ‘proprietary’ IMSC formats.
This is especially timely ahead of the roll-out of Max, WBD’s streaming service that is currently live in the U.S. and is launching in Europe early next year with launches also planned for LatAm and Asia-Pacific.
Commencing September 8, 2023, IMSC-Rosetta will be accessible to all. The complete specification, coupled with samples, example source code, and a public wiki, will be accessible at https://github.com/imsc-rosetta. Queries, observations, and contributions are warmly welcomed through the issue-raising channel on the platform.
In an industry perpetually evolving, IMSC-Rosetta provides a method to store and deliver consistent, high-quality subtitling across languages and distribution channels.
By providing a single universal standard format, IMSC-Rosetta helps the whole media industry supply quality localised content while reducing production costs.
IMSC-Rosetta is the result of over 18 months work primarily by Simon Hailes of Yella Umbrella and Robert Cranfield of WBD, with input and review from Ooona, Broadstream (Screen), Eztitles, Edgeware (Cavena), Zoo, Deluxe, Iyuno, BBC Studios, Sandflow, and many others.
See IMSC – Rosetta files in use on the Yella Umbrella stand – 7.P01 & 7.P02 in the walkway between Halls 7 & 8.
Find out more at the IABM Round Table – 10:30am on 16th September in the IBC Partners Village.
Stellar’s New Shared Task Manager achieves an impossible task….
a 120min multi-voice VO dub and AD project – within a single week.
Here’s how it was done…
Stellar for VO and Dubbing
In February 2023 Hear Say Audio Description were asked to provide the voice over and AD audio for a 120 minute foreign language film, with a turnaround time of just one week.
The original plan was to use a single voice for all characters and the audio description. But with 16 main characters and a narrator, it was soon decided to split the work over three voice artists: one for the narration and AD, one for the female roles, and one for the male ones.
The inputs were the high res original media file in .mp4 format, and an English language SRT subtitle file. There was no separate M&E audio track without the original language audio available.
The client required both a mono wav file of the combined AD and voice-over audio, and a full mix with the program audio; this included dipping of the program audio during the new dialogue to dip the original foreign language dialogue.
Hear Say AD has traditionally been an audio description house working with freelancers, and doesn’t have a dedicated recording studio or digital audio software (DAW). Although each voice artist had their own home studio or sound booth, they were located in different countries.
Cara Edney of Hear Say AD had considerable experience using Yella Umbrella’s Stellar AD software, but had always used it as a single user. However, by early 2023 Yella Umbrella were doing beta trials of their new Task Manager that allowed multi-user shared working environments both for scripting and recording. Stellar had supported multiple role lip-sync dubbing since early 2022.
Although full lip-sync dubbing wasn’t required by the client, it was important that the script length closely matched the duration of the original dialogue to allow the voice artist to pace the dialogue correctly.
The first step was to create a Task within Stellar, and to import and check the script. The video was imported and converted to browse quality (quarter SD) to reduce the size of the file delivered to the remote users. The English language subtitle file was imported and converted to a dubbing format (srtdub). This format supports the association of audio clips to each block of dialogue and the assignment of a Role to each text block in the script.
The script was then checked, each role identified, and the dialogue assigned to a role. In some cases an original subtitle would cover the dialogue of two characters, in which case the title would be split and each half assigned to the correct role.
Stellar supports comments against each block of dialogue so that instructions, ideas or questions could be included in the script and be visible to all users.
Once the script had been checked for dialogue, the roles assigned and timing checked, the Task was ready for recording. Although Stellar supports multi-user real time recording sessions, it was decided to split the work into three Tasks, one for each voice artist. Yella Umbrella added two new features to support this, so that a script could be duplicated with some but not all the roles, and then later multiple scripts could be merged into a single master file.
The voice artists could then record separately from the other users, which greatly speeded up the recording phase to meet the very tight turnaround time.
Once the three Tasks were ready, each with its subset of roles, the Tasks were assigned to each of the voice artists. They received a notification email and could download each Task with their Roles in Stellar on their Mac or PC.
They could then review the timed script, read any comments, and start recording.
The Task Manager system supports automatic uploading of all script changes and audio recordings. Consequently, as work progressed the manager and other users could see progress in real time.
The project manager could hear all audio clips immediately after they were recorded, and so could provide feedback on style and audio quality. This greatly reduced the error rate and the turnaround time for correcting mistakes. Multiple audio-takes are supported, which means alternative audio clips can be stored, reviewed and selected for each section of dialogue. Because each section of dialogue (lines) has a separate audio clip, the timing and trimming of each take can be adjusted without affecting other sections of audio.
Stellar supports fading of the program audio, and this was vital due to the presence of the original dialogue in the master sound track. Audio dip timing and depth could be set and previewed within Stellar during the recording process, and afterwards in QC.
Once the voice artists were happy with their recordings, an initial QC was performed on each Task, and some minor changes made. Here the auto QC system within Stellar identified dialogue with no audio recording and other errors.
Then the three script files were combined, along with their associated audios, to create a new master script with all the roles in a new Task. Another QC was performed with a full preview replay, including all the new dialogue, original audio and fades. Any minor changes or re-records were then performed. Each voice artist could open the master version, re-record one or two changes, and the results were instantly available to all users.
The final stage was to produce the output audio mixes. A mono wav dialogue track, and a separate full program audio mix, were created using the audio mixer tool within Stellar.
Stellar alone was used for all stages of the workflow from scripting, editing, collaboration, recording, QC, final review, and output mixing.
Cara Edney, CEO of Hear Say Audio Description:
“We wouldn’t have been able to complete this intricate project without the shared task capabilities in Stellar, as it allows for different talents to access the latest versions, for a QC to occur efficiently and leave comments/feedback etc without having to download and import separate file versions. This greatly simplified the version control problems associated with a project of this size.”
And
– “The Stellar Task system also allows for quicker talent onboarding as a task can be created by an experienced Stellar user and then the scripter/VO only needs training for box and script creation/basic editing. These skills are often very intuitive and closely mimic other software workflows which the talent is already familiar with”.
– “The task system also increases asset security as only the projects are shared and not high resolution media assets.”
Matt Deakin, CEO of Yella Umbrella:
“Yella Umbrella were delighted to work with Hear Say AD on this challenging project. The new Task System was in late beta testing, but not available for general use. The size of the project at 120 mins, 17 roles and over 1600 audio clips, made this an excellent test of the new system. The project also highlighted useful workflow additions, such as splitting scripts by roles and recombining into a master script, that we have added as a standard feature.
Once the Task System is released later in September 2023 we look forward to working with many new customers on similar and even larger projects.”
Contact Usto arrange an online demonstration or come and see us at IBC 2023.
Automatic dialog removal – from Yella Umbrella and Audioshake
Yella Umbrella and Audioshake have joined forces in a groundbreaking AI technology collaboration that is set to transform the dubbing industry. Their cutting-edge solution addresses one of the major challenges faced by dubbing studios – how to produce top-notch voice-over and lip sync dubbing audio when the original M+E (music and effects) track is unavailable.
Traditionally, dubbing a program into another language meant contending with the presence of the original language dialog in the audio track. Access to the original M+E track, especially with older content or television programs, is a rarity. Conventional approaches involved ‘dipping’ the original dialog and incorporating the new language, but this often led to a loss of vital background effects and music, compromising the overall audio quality.
Leveraging their expertise in music filtering and their remarkable ability to separate vocals and instruments from music tracks, Audioshake now introduces an AI-based dialog removal solution. This innovative technology empowers dubbing studios to automatically generate an M+E track, irrespective of the original material’s availability.
Through seamless integration with Yella Umbrella’s Stellar virtual studio system, this collaborative effort automatically adds a separate M+E track to dubbing projects alongside the original dialog. During script preparation and translation stages, the original dialog proves indispensable, and serves as a reference for actors’ speaking styles during recording. However, when it comes to reviewing and final mixing, the M+E track emerges as the key to delivering a high-quality output.
The impact of this AI dialog removal system will be demonstrated live at the Yella Umbrella stand – 7.P01 and 7.P02 (located in the walkway between Halls 7&8) or at www.yellaumbrella.tv and on the Yella Umbrella YouTube channel – https://www.youtube.com/@yellaumbrella1315/videos.
Experience firsthand how this pioneering collaboration between Yella Umbrella and Audioshake is reshaping the dubbing landscape, elevating audio excellence, and setting new industry standards. Be a part of the future of dubbing with their game-changing AI-powered solution.
Yella Umbrella are proud to be sponsers of Media for All 10
as well as presenting a paper in the conference sessions.
In conjunction with Signapse we will be talking about:
Title:
At Home Signing: Increasing Accessibility forAudiovisual Contentby Enabling the Recording of Signing Tracks in anon-studio environment.
Abstract:
Stellar provides a full range of features for Subtitling, Audio Description and Dubbing creation. Just like in the case of dubbing where the original audio is replaced with a translated audio version in the target language, or subtitling where the audio is being transposed into on screen text, signing provides an extra layer of accessibility by enhancing the original video with a sign language overlay.
Stellar is based on a pluggable architecture which easily allows for the addition of video recording capabilities. As a result, signers would be able to record signing tracks from the comfort of their own home without the need of for a highly specialized studio environments. Once the original video signing snippets have been recorded, they can then be adapted based on the broadcaster’s requirements.
Signapse provide a new technology based on Generative Adversarial Networks (GANs) which allows the use of the original home-made signing videos and transform them into videos which are based on professionally recorded signers. (sign language to sign language ‘translation’).
Once the signing videos have been converted, Stellar would allow the users to adjust, preview and mix into an output as a full studio quality image.
Together we believe that this approach would increase accessibility to audio-visual media for the deaf community both by allowing more quality signed content for current budgets, but also would allow almost any signer from almost anywhere in the world to produce a studio quality signed video.
We will also be exhibiting the full range of Nebula and Stellar tools for subtitling, captioning, audio descriptiom voice-over, dubing and all forms of timed text and metadata.
Come and see us at the Conference or Contact Us to arrange an online demonstration.
How will chat engines help the process of subtitling in the future?
Subtitling is an essential aspect of video production, and it is crucial to get it right to ensure that viewers can understand the content fully. However, subtitling can be a time-consuming and often challenging process, especially when dealing with foreign language content or intricate technical jargon. This is where chat engines come in handy.
Chat engines are advanced computer programs that can analyze text and understand the context and nuances of language. They can be used to automatically generate subtitles for videos, making the subtitling process more efficient and accurate.
One of the most significant benefits of using chat engines for subtitling is the speed at which they can produce subtitles. Traditional subtitling methods often require a human to manually transcribe the spoken dialogue, which can take hours or even days. However, chat engines can generate subtitles in real-time, significantly reducing the time and effort required to create subtitles.
Furthermore, chat engines can also help to improve the accuracy of subtitles. As they analyze text and understand the context, chat engines can accurately translate foreign language dialogue, even if it contains idioms or colloquialisms that might be challenging for a human translator to understand.
Another benefit of using chat engines for subtitling is the potential cost savings. Traditional subtitling methods can be expensive, especially if you need to hire professional translators or transcriptionists. Chat engines, on the other hand, are cost-effective and can be integrated into your existing subtitling workflow, saving you time and money.
Of course, chat engines are not without their limitations. They may struggle with complex technical terms or names that are not in their database, and there is always the risk of errors or inaccuracies. However, these limitations can be minimized by using chat engines in combination with human oversight, which can catch any mistakes and ensure that the subtitles are accurate.
In conclusion, chat engines are a powerful tool that can help to streamline the subtitling process, improve accuracy, and reduce costs. As video content becomes more prevalent, the demand for subtitling will continue to grow, and chat engines will undoubtedly play a significant role in meeting this demand.
This opinion was produced by ChatGPT in answer to the question; Write an opinion piece on the question: ‘How will chat engines help the process of subtitling in the future?’ Is it accurate, or is it just regurgitated marketing speak?
The team at Yella Umbrella have been experimenting with real world problems that Chat Engines could help solve for professional subtitlers. The first of these was the task of text summarisation.
You’ve just translated an English template into German and some of the reading speeds are too long due to those pesky longer German words. So now you need to rephrase the German to say the same thing in the time available, as set by the original dialog timing and the reading speed rules you need to meet. Traditionally this was a task for a skilled translator, maybe now chat engines could help out?
We interfaced ChatGPT to Stellar and provided a Rephrase option. When asked to rephrase a subtitle in a set number of characters the Chat Engine responds with a shorter text that should match the timing constraints. So how well did it work?
English
German (Google Translate)
Rephrased German (Chat-GPT)
The original German translation was over length by several characters, when asked for a shorter version ChatGPT came back with a reasonable alternative which roughly translates as ‘Come back next year, audience guaranteed!’
While this is still very much an experiment we could see a workflow with alternative, sorter, translations provided when the initial translation is over length.
Would you use these new AI tools to speed up your work, or do you see them as a distraction?
We will be showing these new ideas, along with all of the Nebula and Stellar tools, at NAB in Las Vegas from 15th to 19th April on stand N2562M.
Let us know your thought and ideas on other ways these new AI services could help you.
Come and hear them at NAB or Contact Us to arrange an online demonstration.
Microsoft Azure and IBM Watson Voices added to Stellar
Yella Umbrella announces the addition of Microsoft and IBM computer voices to the Stellar Timed text and audio tools. Microsoft and IBM bring an additional 440+ voices in 140 languages or dialects to the existing 400+ voices from Amazon, Google, Acapela and Cereproc.
The continued improvement in the quality and controllability of computer generated (synthetic) voices over the last few years has been astonishing. It is now possible to control the style, speed and pitch of synthetic voices, and, in some cases, to have detailed control over prosody and intonation.
Synthetic voices can now be used for audio description, spoken subtitles and in some cases voice-overs and dubbing applications.
Yella Umbrella are committed to providing the best tools for media localization and access services.
Come and hear them at IBC or Contact Us to arrange an online demonstration.
From Mr. Neb and
everyone at Yella Umbrella.
IBC 2022 – Pods P01 & P02 – in the walkway between Halls 7 & 8
Cloud based Task Management System added to Stellar
Yella Umbrella will be demonstrating the new cloud based Task Manager as part of their Stellar toolset. Used in conjunction with the Stellar timed text and audio tools, the Task Manager provides the delivery, assignment, security and tracking of subtitling, audio description, voice over and dubbing Tasks.
Create a Task, with video, audio, text files and instructions and assign it to any user of the Stellar software, anywhere in the world. Task Manager support high level digital encryption (DRM) second factor authentication (SMS or email) and customised media spoiling.
Task Manager can track the progress on any task by any user, and all work done is synchronised to cloud storage, in real time.
Centralising the storage, distribution and tracking of all Tasks leads to significant time saving and cost reductions. No more emailing files or use of file delivery software. Automatic media ingest and upload options available to further speed up the process of starting a new Task.
Task Manager can be provided either as a shared cloud system from Yella Umbrella or as a private cloud system for larger customers.
With so much media moving from a broadcast environment on to streaming service there is a huge challenge to repurpose subtitles originally created for broadcast to forms suitable for a streaming platform. Often the video has been edited to remove ad-breaks or to add or remove colour bars, slates, headers or bumpers. Any change in the video run length or timing requires a matching edit in the associated subtitle files, of which there may be many different language versions. This process is known as Conforming.
As well as subtitle file format conversion Stellar from Yella Umbrella now provides automated realigning (conforming) of the subtitle file to match the new, edited, video. Using Speech to Text web services from Amazon and others, Stellar realigns the subtitles to match the dialog detected in the new video edit. Subtitles can be moved or deleted, any edits that can’t be done automatically will flag as errors.
When moving a complete archive from broadcast to streaming there can be tens or hundreds of thousands of subtitle files to conform. Stellar greatly speeds up the process so reducing costs and improving productivity.
