Dalet Flex 2026.8.0 Release Notes

Dalet is excited to present the release of Dalet Flex 2026.8.0. 

This release of Dalet Flex provides the new features, quality and security enhancements, bug fixes, and other changes listed below. 

Required Versions for Unified Installer for FlexXtend 

• Xtend-flex-unified-installer: 4.0.49 

Plugins Removed from 2026.8.0

These legacy plugins have been completely removed from Dalet Flex from 2026.8.0.

Dalet Flex

New Features

Starred Assets

From this version of Dalet Flex, you can “STAR” objects in FlexMAM, assets, collections, UDOs to mark and easily retrieve them. Starred assets are stored across sessions and save time on repetitive searches.

Click the star icon next to an asset to add it to the Starred list in the left panel.

The total starred item count is displayed in the left panel.

The first 10 starred items are displayed in the left panel, click on View All to see all starred items. You can reorder items in the starred listing via drag-n-drop or using the arrow icons in the table header

Starred items are displayed in the Search Results/Left Panel with a grey star. Click on the star to remove the asset from the starred list.

NOTE: Starring review sessions in FlexREVIEW and displaying them in FlexMAM and add/remove star from three dots menu are planned for the next release of Dalet Flex.

Access dalet-cut-service using Flex SDK

This Epic enables Java Flex client SDK methods to consume dalet-cut-service endpoints, providing a more convenient way to interact with dalet-cut-service.

dalet-cut-service is currently used mainly to: 

• create AFF (Cut edl) from list of media sources
• extract TTML from AFF
• return list of clips contained in the AFF
• replace media inside clips in the AFF

These actions can now be done from the flex sdk.

Improve flex-panel-app – launch workflow wizard 

As part of an initiative to replace and improve the flex-panel-app, the launch workflow part of the flex-panels app has been implemented in the flex-wizard-app. Flex-wizard-app uses a modern framework that supports quick and easy updates. In this version of Dalet Flex, flex-wizard-app is used to launch workflows using single page wizards. Metadata fields used in the new wizard app have a new look and support the same functionalities as fields in metadata forms in FlexMAM.

Inherit media usage rights in clips

From this version of Dalet Flex, there is support for media usage rights (MUR) inheritance and display for clips created in Pyramid from source material in Dalet Flex.

The MUR on the source material is inherited by clips created from the source, and if the MUR on the source is modified, this is updated dynamically on the clip.

Use dalet-ai-service to support transcription of live feeds

From this release Dalet Flex supports near realtime transcription for live media using local models together with dalet-ai-service.

Support for hls and mpd growing media has been added for transcription. The job works for these formats even after the media is fully formed and no longer growing.

A typical workflow is:

• Ingest / Live Asset Detection: The growing media stream (HLS or MPD manifest) is ingested into Flex using a live-enabled storage resource.
• Workflow Execution & Groovy Script Trigger: A workflow step triggers a JEF Script Action running Groovy script. The script resolves the asset’s manifest/VFS location and calls POST /jobs on dalet-ai-service.
• Real-time Processing: The dalet-ai-service processes the growing stream via an open-source model (e.g., openai/whisper-large-v3) and produces incremental output files (.json, .ttml, .srt, .txt).

The transcription is triggered using the existing /jobs endpoint as below,

 curl -X POST --location 'http://localhost:18171/jobs' \       
--header 'Content-Type: application/json' \
--data '{
  "externalId": "Transcribe-12345",
  "job": {
    "jobType": "AIJob",
    "jobProfile": "SpeechToText",
    "jobInput": {
      "jobInputType": "SpeechToTextInput",
      "engine": "Model/openai/whisper-large-v3", // optional - System default can be set at flex/dalet-ai-service/s2t/defaultEngine
      "inputFile": "<file in VFS OR S3 signed url specification pointing to the manifest file - existing>",
      "outputLocation": {
        "jobOutputType": "SpeechToTextOutput",
        "targetLanguage" : "french", //Newly added
        "json": "<file in VFS OR S3 signed url specification - existing>",
        "ttml": "<file in VFS OR S3 signed url specification - existing>",
..
        "srt": "<file in VFS OR S3 signed url specification - existing>",
        "text": "<file in VFS OR S3 signed url specification - existing>"
      },
      "language": "en" // optional
    }
  }
}

The results are written to the specified output files as soon as the transcription is completed in almost near-realtime (in an incremental manner). 

 When the consul KV flex/dalet-ai-service/platform is set to FLEX, the transcription events are published so that JEF can consume them and create time-line annotations.

NOTE: JEF plugin support has not yet been implemented

Volume Sharing 

As dalet-ai-service begins supporting larger open-source models, sharing the model storage volume across service instances is strongly recommended. This prevents the same models from being downloaded and stored separately in every container, reducing storage consumption and download time. 

On-premises environments

 Use NFS to share the model volume across service instances. If NFS is unavailable, share the volume between dalet-ai-service instances running on the same node.

 Cloud environments

 Use EFS or another shared storage solution that supports access across all nodes.

Analyze visual content of image files with dalet-ai service 

A new VLM capability using Qwen3-VL-2B-Instruct to analyze visual content of image files has been implemented in dalet-ai-service. The implementation runs entirely in Dalet-ai-services with no cloud dependencies and uses a self-contained async REST API that accepts images and returns AI-generated natural language descriptions. 

The service supports an input of single/multiple (1-100) image files and the analysis returns the text inside images and the name of the people in an image.

The functionality is intended to be used to enrich metadata by analyzing the images/keyframe images of the assets by getting a useful description about it using the native AI capability of Dalet Flex.

Settings

A KV flex/dalet-ai-service/imageAnalysis/batchSize sets how many images should be used for processing at a given time. The default value is 1, so if 100 images are passed they are processed one by one.

In spite of this, analyzing a high-res image might cause OOM. In this case, use flex/dalet-ai-service/imageAnalysis/maxImagePixels to set the ceiling above which the image will be resized to that size. The default value for the maximum pixels is 1048576.

Usage

Use a POST command to start a MediaAnalysis job of type AnalyzeImage together with the image file location and in the Goal field type clear instruction of the analysis to perform on the image.

POST <dalet-ai-service>/jobs
{
"job": {
    "jobType": "AIJob",
    "jobProfile": "MediaAnalysis",
    "jobInput": {
      "jobInputType": "AnalyzeImage",
      "inputs" : [
            {
            "file" : "<file in VFS specification - details below>",
            "goal": "<Instruction of analysis to be performed on the given image, eg. Who are the people in this image?>"}]
    }
}
}
```

Once the JobID is generated, you can retrieve the response using a GET call. 

```
GET <dalet-ai-service>/jobs/<job-id-from-first-call>
{
"id" : "",
"date_created": "",
"status": "",
"job_output" : {
"analysisResults" : [{
"status" : "success",
"output" : "Output of the given analysis"
},
{
"status" : "error",
"error" : "Output of the given analysis"
}]
}
}

SAMPLE RESPONSE
{ 
    "id": "663", 
    "dateCreated": "2026-07-02T07:10:51", 
    "dateModified": "2026-07-02T07:11:29", 
    "status": "COMPLETED", 
    "jobOutput": { 
        "jobOutputType": "MediaAnalysisOutput", 
        "analysisResults": [ 
            { 
                "status": "success", 
                "output": "The people visible in this image are Neymar, Lionel Messi, and Cristiano Ronaldo." 
            } 
        ] 
    }, 
    "processedCaptions": null, 
    "error": "" 
}

The currently supported VFS file specification is one of the below,

For images in local file system:

{
                "protocol": "FILE",
                "path": "/data/images",
                "fileName": "image_analysis_input.jpeg"
}

For images in S3:

(Please note that the respective Key and secret are expected to be uploaded in the secret-service prior to calling this API).

{
"protocol": "S3", 
"bucket": "vfs-bucket", 
"path": "input/a.jpg", 
"secretUuid": "0f9c...-uuid"
}

Improve Tile/List View selection

To allow you to toggle between Tile View//List View with one click, the tile/list view switcher has been moved to the top level in the Asset Results row in FlexMAM.

Taxonomy group visibility filters respected in FlexMAM

From this version of Dalet Flex, taxonomies are filtered by user group visibility in the left panel in FlexMAM to restrict user access based on group permissions as well as account permissions. 

Dalet Cut

New Features

Support adding fixed CC tracks in Cut Video

From this version of Dalet Flex, you can manually add fixed CC tracks in Cut Video for four fixed languages: English, French, Italian and German.

Implement integration with Cut Audio from Dalet Flex

From this version of Dalet Flex, the dedicated app for Audio editing, Cut4Audio, has been integrated with Dalet Flex. Cut4Audio is a web-based audio editor that allows you to easily edit, mix, and export audio files directly from your web browser. Cut4Audio comprises audio tracks on the Timeline, view meter, audio effects and transitions, audio transport buttons.

Use the App Switcher to open Cut Audio.

From this release you can open these types of assets in Cut Audio directly from FlexMAM: 

• Video 
• Audio  
• Edit asset (EDL) 

Use the Open with option in the 3-dot menu on these asset types to open them directly in Cut Audio, only the audio tracks are displayed in the Cut Audio timeline.

Ingest Portal

New Features

Display Dalet Flex asset summary panel in Ingest Portal

From this version of Dalet Flex, you can open the asset summary panel for recording jobs directly in Ingest Portal instead of being redirected to a separate tab in the Browser. In the Job Info panel, click on the Go to Asset arrow.

When the summary panel opens, the Timeline shifts to the left and the selected job remains visible.

Click the icon (1) to close the summary panel or click the Manage button (2) to open the Manage page for the asset in Dalet Flex.

Display a Target view in Ingest Portal

From this version of Dalet Flex, a dedicated Targets view has been added to the Ingest Portal timeline. You can use this view to monitor target availability, manage target-based recordings, and switch between Sources and Targets.

Tabs on the Ingest Portal allow you to toggle between the Source and Target views.

You can use the search box to filter the sources and targets.

Any recording jobs that are created are displayed in both the Source and Target views.

You can schedule, view and edit jobs directly from the Target view, the recording panel is prefilled with the selected target.

Flex Xtend

New Features

Clean up file download in Flex Xtend

When you work with Flex Xtend the local working directories (cache) on Premiere machines fill with media files that are downloaded from Dalet Flex: aegraphics, images, dlts and mpd files. This can cause slowness and fill the disk space on the local machine. 

From this version of Dalet Flex, automatic cleaning of these directories by Oreo according to their last access time, and the configuration in the Oreo config.json. This functionality is disabled by default and the cacheManager section in the config.json lis as follows:

"cacheManager": {
"cacheCleanupPeriodInSeconds": 0,
"fileAgeThresholdInSeconds": 0,
"fileExtensionsToSkipInLocalStorage": [
".prproj"
]
}

To enable the cleanup, set values > 0.

By default, the .prproj files in the local storage are not purged, to purge these files together with the dlts and other media files,  remove the “.prproj“ entry from the array 

In addition, an auto download and relink functionality for projects that have been saved locally (not in Dalet Flex) has been implemented so they are usable, even if the Oreo cleanup deleted files that were used in these projects.

NOTE: Auto download and relink occurs for all opened projects, provided Flex Xtend is running and logged in to Dalet Flex.

The process is triggered when:

• Flex Xtend logs in 
• FlexXtend is notified of a project opening from Premiere (not import from Flex)

Changelog

2026.8.0

Enhancements DaletFlex

• When basic auth is used in conjunction with actAsUserId, API response 'href' response URLs include the account of the authorised user, not the actAs user.
• Allow service name of Enterprise to be looked up differently when running under K8s vs Consul service discovery
• Create FAS upload client (with CLI for testing purposes).
• xTend Flex - render on remote AME unaffected By Premiere project change.
• Add support for postgres for dalet-ai-service
• [Search] Implement search-fusion in flex-searchelastic-master.
• For transcription translation, extend existing /metadata-analysis/translate-text to use local models when configured. If no model is provided use legacy providers.
• Apply view permission check to storage resource for authenticated user before allowing upload. 
• Login App: Entra ID OIDC supported with no issues on login/logout.
• Index AIMetadata fields for WorkflowDefinitions in flex-indexelastic-service.
• Accept http endpoint representation of the image. AnalyzeImage accepts HTTP image URLs along with VFS locations.
• Update FSP to use FFMPEG 8.1.2.
• Wrapped method removeObjectId added in JEF groovy SDK to remove object from context.
• FSP: Update value calculation logic in FMPE to store value in dB. Add verification tests in FMPE and FSP and update documentation covering the value calculation.
• FlexXtend: Add Toaster message indicating missing file refresh status when opening a project in Adobe Premiere.
• Add aria labels for the icons in the header.
• Workflow Definition Tags: Include AI metadata fields for GET workflowdefinitions API using a flag GET /api/workflowDefinitions;includeAIMetadata=true.
• Split flex-solution-packages-provision-resources-cmbundle and exclude resource bundle in Dalet Tenant Service provisioning.
• Expose wizard custom iconId on the Wizard REST API.
• The waveformNewLocation key-value toggle, flex/flex-jobasyncexecutor-service/waveformNewLocation, was introduced in Flex 2023.11.0 to allow both old and new waveform storage locations during migration. Since Flex 2025.6.0 LTS it defaults to true. The KV is now obsolete and has been removed completely.
• Add a new conform preset: conform-high-res-25i-16mono to output XDACAM_HD422_50 interlaced (Top Field First), with 16 mono tracks.
• VFS growing folder support for non-local providers to support copy/move/listing.

Fixed Issues DaletFlex

• Fixed an issue with inconsistent naming, editing app renamed Cut Video and Cut Audio in all places.
• Fixed an issue where an odd waveform shape is displayed at the end of file for growing media.
• XTEND: Fixed an issue where partial check-in is performed more than once for the same segment.  
• Remote AME: Fixed an issue where a lot of source media is seen as offline in a render.
• FlexXtend import: Fixed an issue where importing an asset triggers a download even though the media file already exists in configured Shared Storage.
• [Xtend Flex] Fixed an issue where full check in is performed a second time for a clip that has already been rendered locally.
• Fixed an issue where the thumbnail on parent UDO asset manage page is not displayed when view from breadcrumb.
• Fixed an issue where it was not possible to preview a growing file in FMP if the file has accent in its name.
• Fixed an issue so that a conform action attaches an associated CC asset as a child of the conformed video asset and not to the edit asset.
• Fixed an issue where an NPE is incorrectly thrown when trying to download assets using the report definition when file information is missing for selected files. 
• Fixed an issue where modification of an existing metadata definition fails if a Single/Multi-Option variable contains more options than the configured optionsMaxLimit KV value even though no change is made to the option with the limitation.
• Fixed an issue where Sorry, there was an error while loading is displayed when trying to click the online archived assets from asset summary page.
• Fixed an issue where a System Error is displayed when saving configuration details of resources without providing mandatory input fields. After the fix, the appropriate error is displayed.
• Fixed an issue where data is not loading in Grafana dashboard under Flex Enterprise services for Workflow Quota per account/per definition.
• Fixed an issue on the Manage Page where audio track label shows -- instead of stream-number | stream-language | channels when stream name is missing.
• Flex Xtend: Fixed an issue where some imports get stuck when importing multiple projects.
• Add functionality to create placeholder for backup jobs and display them in Ingest Portal and FlexMAM.
• Fixed an issue where the custom flow status icon and tooltip are not displayed in tile view when fully zoomed out.
• Fixed an issue where dalet_service_info readiness metric is renamed/broken after Micrometer 1.15 upgrade.
• Fixed an issue where using the hotkey or player mark In Time icon in the Summary page  to make a time selection is not working.
• [Xtend Flex] Fixed an issue where Media checkIn button is disabled when dragging local media into a project.
• Fixed an issue where GET /transcodes[?type=<type>] is not retrieving the correct results.
• [Xtend Flex] Fixed an issue where partial check in is triggered instead of Media CheckIn for video files.
• Fixed an issue where an undefined object referenced error is displayed when performing MIXED mode non-package output chunked transcode job.
• Fixed an issue where delete/purge action is failing for segmented asset with NPE. 
• Growing file: View live asset\refresh the manage page for live asset deselect audio stream icon
• Simultaneous listing and deletion of folders has undefined behaviour on Posix filesystems
• Fixed an issue where launch workflow wizard does not prepopulate metadata values when prepopulateMetadataValues is set to true. 
• Fixed an issue where Launch workflow wizard with a secondary check variable does not show metadata for the signal action on the wizard page.
• Remove backward-compatibility support for singular BG_THUMBNAIL_FILENAME / BG_selectionStrategy_time job-context keys in ExtractThumbnailExecutor.
• [Xtend Pyramid] Fixed an issue where rendering a specific project with aegraphic and images, the render job is not created on remote AME. Issue caused by special characters ( ) in the file names.
• Fixed an issue where FFP unpack progress is not reported.
• Fixed an issue where an error or no result when polling restore status results in job failing or continuing to run instead of completing.
• Fixed an issue where intermittently the URL is not updated after navigating back from Starred to Asset listings.
• [UI] Starred – Fixed an issue where assets do not remain highlighted in the Starred left panel after navigating to and from the Manage page.
• Fixed an issue so that transcription writes the output to the file name specified in "SpeechToTextOutput" section in the Groovy Script.
• Fixed an issue where custom markers are not visible in custom marker tab.

List of Services upgraded between 2026.7.0 and 2026.8.0 

For the new 2026.8.0 release, most services have changed.