For AI agents: a documentation index is available at /llms.txt. Markdown versions are available at matching .md URLs.

Skip to content

Redactor 6.x Release Notes

Redactor changelog archive

Historical Redactor 6.x updates, fixes, and API changes. Return to [latest release notes](release-notes.md).

6.6.0 – September 19, 2025

New Features

  • Data Drop via Flag File
    By dropping a file (can be empty) named DATA_DROP in the user directory will reset it completely on a restart except of license and the registered user data (name, credentials). This is intended to be used when Redactor acts as a worker for transient operations (no persistence required) in an API driven context. . Confirmation of this drop is in the main.log (...INFO [init] dropped all data except license and user data, reason: file).

Updates

  • Original Session Identifier Enforcement
    There was a unintended conflict possible when a session got originally introduced, stored in a bucket and then later on loaded from there again, but with a different session identifier given in the operation's prepareConfig. This will now cause an error, since session identifiers are pinned forever through their immutable presence in the manifest.json file (stored in a bucket along with all of the other data).

Bug Fixes

  • Session Cleanup Failed
    Due to a 3rd party library causing recursion the backend would crash. Did only happen in certain storage scenarios.

  • Missing Audio Data padded with Silence
    If a video only contains an audio portion at the beginning we pad the rest with silence. This facilitates intake, conversion, analytics and presentation.

6.5.2 – August 25, 2025

New Features

  • Open API Validation
    For API users: by setting the environment variable REDACTOR_API_SPEC_ENABLE=1 the OpenAPI 3.0 specification can be accessed under http://[host]:[port]/apispec/yaml (or http://[host]:[port]/apispec/ for JSON). For request and response validation set the environment variable REDACTOR_API_VALIDATE=1. Notice though that your existing API calls might need adjustment if e.g. extraneous data is present.

Updates

  • Automatic CPU Fallback if GPU Detection Fails
    If during the GPU detection phase an unexpected error occurs (happens e.g. on untested CUDA updates) Redactor will automatically switch into CPU mode, so you can still use it. The incident will be logged. This the equivalent of setting the REDACTOR_GPU_IGNORE=always environment variable.

  • Enhanced Reconnection Handling
    The transport layer (server version only) was overhauled to keep the client connected, even in unfavorable network conditions. Same as for more timely (30 seconds) detachment of clients, and the release of their sessions, if they cannot reach the server any longer. In the UI a Reconnecting... indicator will show when connectivity is lost, until reconnection happens, so you are aware of the current situation. Unless your networking won't allow websockets at all, you are encouraged not to use the fallback via the REDACTOR_FORCE_POLLING_TRANSPORT=1 environment variable any longer. Please remove it, or set it to ...=0 if you have been using this approach before.

6.5.1 – July 18, 2025

New Features

  • Bulk Object Selection via Shift-Click
    The ability to select a range of objects in the Objects List via Shift-Click has been added. This key was previously used to select individual objects, so that functionality is now provided via Control or Command keys.

    • Range selection: Click first object, hold Shift and click last object
    • Single selection: Hold Ctrl (Windows) or Command (Mac) and click to select or deselect individual objects.

  • View Keyboard Shortcuts in Settings
    A "View Keyboard Shortcuts" link has been added to the App Settings screen for quick and easy access to all available keyboard shortcuts.

Updates

  • GPU Acceleration Now Requires 4GB VRAM by Default
    GPU acceleration now requires a compatible NVIDIA GPU with at least 4GB of VRAM to ensure stability and performance for advanced computer vision models. Systems without a compatible GPU or with less than 4GB VRAM will use CPU mode by default (but this can be overridden).

    Based on our experience, GPUs with 3GB or more can typically run auto-detection with GPU acceleration enabled for most videos. However, we've set the default to 4GB to build in a cushion, as some videos may require more. This is especially important on Windows, where the operating system can use a portion of VRAM for other processes, reducing the amount available to Redactor and increasing the chance of running out of VRAM during auto-detection.

  • GPU Acceleration Control via Environment Variable
    Administrators can override the default GPU behavior by setting the REDACTOR_GPU_IGNORE environment variable with one of the following values:

    • auto (default): Enable GPU acceleration only if a compatible NVIDIA GPU with at least 4GB VRAM is detected.
    • never: Always enable GPU acceleration, even if the GPU has less than 4GB VRAM. May crash if VRAM runs out.
    • always: Always disable GPU acceleration and use CPU mode.

    This provides greater flexibility for deployments with different hardware or specific needs. If you encounter videos that require more than 4GB of VRAM and GPU acceleration is causing issues, you can set the environment variable to always to disable GPU mode and force processing on the CPU instead.

    If you experience such issues, it would also be helpful if you could send us your logs for review. This allows us to monitor details such as the type of video (resolution, fps, etc.) and the auto-detection settings (confidence threshold, object types selected, etc.) that were used, so we can better diagnose and optimize the VRAM defaults for future releases.

Bug Fixes

  • Crash During Speech Transcription in Desktop Version
    Fixed an issue that caused Redactor Desktop to crash when running speech transcription. The server version was not affected.
  • Export Metadata Not Fully Deleted on STATE_DROP
    Fixed a bug where optional metadata files generated during export were not always removed during a STATE_DROP operation.

6.4.4 – June 23, 2025

What's New and Improved

  • Download Logs from the Settings Screen Admin and Desktop users now have a “Download Logs” button in the Settings screen to easily collect logs without using the F2 keyboard shortcut or needing access to the server's filesystem. Users in the Supervisor or Users groups will not have access to this feature.

  • Display License Status and Expiration Date Each license in the Admin panel now clearly shows its expiration date and time. Expired licenses are highlighted in red for quick identification.

  • Expanded Profile Selection on Low-Spec Devices Systems with limited GPU memory (< 2.5GB) now have access to confidence threshold options when running the auto-detection process: Balanced (default), Higher, and Lower.

Bug Fixes

  • Expired Licenses Not Replaced with Valid License During Startup Resolved an issue where newly refreshed licenses were mistakenly deleted during startup if the previous license had expired.

  • Fixed Object Timeline Context Menu Timeline context menus now behave consistently and support localization. Previously, multiple context menus could open with each right-click on a timeline object.

  • Group Cleanup When Deleting Objects Deleting an object now correctly removes it from any groups it belonged to in the saved data.

  • Cleaner Application Logs Removed unnecessary debug logs from the browser to improve clarity and performance.

6.4.3 – June 12, 2025

Improvements

  • Improved S3 Integrity Checks: File uploads to AWS S3 now use built-in hashing from the AWS SDK to ensure upload integrity.
  • More Resilient Session Restoration: Session restoration from S3 and other storage buckets now works reliably even when optional overlay data is missing.

Bug Fixes

  • Fixed Unexpected Manifest Loading: Redactor no longer attempts to load the manifest.json file when the input path does not end in a trailing slash.
  • Preserved Session Data Without Overlay Files: Fixed a bug where imported session data would reset to defaults if overlay files were missing.
  • Synced Overlay Files to Storage: Overlay files are now correctly synchronized to either local or remote storage buckets to maintain additional editing context.
  • Resolved Crash from Build Tool Conflict: Replaced a problematic third-party hashing module to prevent a deployment-time crash.

6.4.1 – May 23, 2025

New Features

  • Improved Connectivity for Challenging Networks: Redactor now includes a fallback, long-polling connection option for networks that don't support WebSockets. Set the environment variable REDACTOR_FORCE_POLLING_TRANSPORT=1 to force the connection to use polling. If you experience frequent "Lost connection to server" disconnects, enable this new option to see if it helps.

Improvements

  • We’ve updated the Auto Detection settings screen to make it easier to understand each detection confidence option.

Bug Fixes

  • Fixed an issue where exported videos appeared darker or more washed out than expected. Videos now retain accurate colors and contrast.

6.4.0 – May 12, 2025

New Features

  • Object Timeline View: Visualize selected objects on a dedicated timeline, making it easier to review activity over time.
  • Object Grouping: Group multiple objects or tracks to simplify the management and review of related detections.
  • Object Selection Enhancements: You can now select items in the Objects List by hovering and clicking the new checkmark icon in the top-right corner. The previous method—holding the Shift key and clicking—still works.

Improvements

  • Improved Image Quality: Exported single images are now clearer and sharper. File sizes may be slightly larger as a result.
  • Updated Selection Styling: Selected items in the Objects List now feature a clearer border instead of the previous semi-transparent overlay and central checkmark.
  • TypeScript API Types: A typescript package is now included with API-enabled installations, offering documentation and examples for developers.

Bug Fixes

  • Fixed an issue where manually created objects didn’t display the “manual” tag in the main video list or object filter.
  • Resolved a memory issue that could lead to excessive usage or crashes when tracking long-duration objects, improving overall stability and performance.

6.3.2 - April 15, 2025

Bug Fixes
  • API: v4 data import was broken, now creating a manifest.json file on the fly

6.3.1 - April 11, 2025

New Features
  • Added Korean and Polish languages for speech detection
Updates
  • Simple version query with the /version?json URL path.
Bug Fixes
  • API: supported S3 ACL set for writing wasn't complete (missing aws-exec-read, bucket-owner-*)

6.3.0 - April 11, 2025

New Features
  • Selection improvements:
    • [Ctrl]+[D] to select the current object in the video editor
    • Invert Selection menu item in the objects list, useful for deleting many unwanted objects
    • Objects selection is getting preserved, will be restored when reopening a video later on
  • Environment variables are now getting validated, and documented under http://{server-url}/env/report?dev
  • Keyboard shortcut [F2] to quickly gather the logs as a ZIP file on the desktop without having to create a support ticket
  • API: if an operation's request descriptor is marked with { ... transient: true } the operation will be forgotten right after it finished, which avoids having to do explicit delete calls afterwards (with the outcome being reported via notifications)
  • API: operation notifications can be declared to be retried of the HTTP callback called, e.g. with { ... retries: 3 } to try 3x before giving up
  • API: added route /api/v11/ops:status to yield JSON overview of the operations count and how many are queued.
Updates
  • API: added environment variable REDACTOR_TLS_ACCEPT_UNAUTHORIZED to use TLS certificate leniency only in the Redactor process.
  • API: CORS improvements, added environment variable REDACTOR_CORS_ALLOW_ORIGIN=1
  • API: more efficient MD5 creation for S3 writes, added environment REDACTOR_API_SYNC_S3_SENDMD5=0 to skip it altogether
Bug Fixes
  • Slow zooming in der audio timeline fixed
  • Contact Support errors are now shown more prominently
  • Support dialog wasn't working correctly, now reporting errors if transmission failed.
  • Thorough cleanup if session couldn't be created.
  • Better crash recovery, not letting the UI become stuck.
  • API: v4 data import was broken, now creating a manifest.json file on the fly

6.2.0 - February 26, 2025

New Features
  • API: operation request queuing. Prevents overloading the system when too many requests are issued. Default request queue size is 1024 (environment variable REDACTOR_API_OP_QUEUE_LIMIT to change), anything beyond that will let request being denied with an HTTP busy response. Only 4 operations will run in parallel or actively respectively (REDACTOR_API_OP_MAX_RUNNING). Cancellation can still happen, even if in the queue.
  • Press [F1] for quick overview of keyboard shortcuts available.
Updates
  • API: webhook requests (callbacks for operation status) can now be repeated on failure, configurable via REDACTOR_EVENT_WEBHOOK_RETRIES=N environment variable. Recommended value is 3.
  • API: operations can be marked with { ... transient: true } in the request to make sure their state get removed automatically when they finished, suitable when combined with webhooks for receiving status (versus REST polling).
  • Expired licenses are kept and tried to be refreshed at startup time.
Bug Fixes
  • Duplicated error logging prevented, better error reason surfacing.
  • Discard sessions if an operation fails, not more leftover/build-up.
  • Contact support dialog (log attachment choice) was broken, transmitted data is now also smaller/uploads faster.
  • More thorough session (data) deletion, some parts got leftover/built up under certain circumstances.
  • The destroy command wasn't forwarded to iFrame embedded Redactor instances.
  • Fixed delayed user detachment issues (memory buildup on inactivity).

6.1.3 - January 22, 2025

New Features
  • Added new environment variable to help with debugging videos that fail to import. Setting REDACTOR_LOADVIDEO_ERROR_DUMPS=1 (increase 1 to how many debug folders/videos you want to keep on system) will copy the original and the converted videos (if any) into a local folder, mark this folder with the session ID, drop additional information (request etc) into the folder, and delete older folders if we run out of space/capacity.
Updates
  • Preventing the contact-support dialog from being submitted multiple times in a row, present a status message while transmitting.
  • Remove unknown tracks from media during import.
Bug Fixes
  • (Desktop) neither the Add Media button nor the File->Open menu would load media. Only dragging and dropping worked. Caused by Electron update in 6.1.2

6.1.2 - January 14, 2025

New Features
  • Upgrade Electron to latest version.
  • Pressing F1 on keyboard when in editor will show a list of keyboard shortcuts.
Updates
  • (API) Add support for submit MD5 of content if needed by S3
Bug Fixes
  • (Server) Fix issue with new installations getting stuck on initial setup screen.
  • (Desktop) Stepping through a video frame-by-frame could show the boxes in a different location than the rendered output. They were off by one frame.

6.1.1 - December 19, 20024

New Features
  • Replace native browser dialogs with application specific ones.
  • Allow uploading of media into current folder. Previously, all uploads would be placed in the root location.
  • Updated context menu on main page to work better with bulk operations.

6.1.0 - December 13, 2024

New Features
  • Folder Support - Media can now be moved and grouped into folders on the main page.
Updates
  • Improved handling of interlaced videos and those with odd resolutions during import conversion stage.
Bug Fixes
  • Improved support of filenames with uncommon characters (like #)
  • Fixed internal issue with persisted storage and write queueing causing issues (introduced in v6.0.5)

6.0.7 - December 10, 2024

New Features
  • (API) add ability to ignore SSL/TLS certificate issues for webhook URIs. Must be specified in request body as metadata etc_event_webhook_ignore_certificate_errors: true
Updates
  • (API) Add skipOnError request option für STATE_DROP for debugging purposes (error state is kept for querying)
Bug Fixes
  • (Desktop) The preview button was not working properly and would only show a black area instead of the exported media. This depends on the locally exported file still being available at viewing time.

6.0.6 - December 2, 2024

Updates
  • Improved error handling for speech transcription processing (inconclusive before).
Bug Fixes
  • (API) STATE_DROP was being skipped if sessionId wasn't provided in request body.

6.0.5 - November 11, 2025

New Features
  • (API) Ability to custom name the exported file when written back to storage/bucket using videoContext.outputContext.exportKey = "my exported file.mp4" in the request.
  • New keyboard shortcut [W] will cycle through the redaction objects on the screen
  • New keyboard shortcut [Delete] will delete the currently selected object
  • New environment variable REDACTOR_GPU_IGNORE. Defaults to 0. Set to 1 to disable the usage of any GPU in case of incompatibilities.
Bug Fixes
  • Fixed issue with audio redaction regions not muting during playback in editor.
  • Update the label used for Chinese language in speech transcription.

6.0.4 - October 28, 2024

New Features
  • New keyboard shortcuts added for manual editing. [A] and [D] keys to move backward and forward one frame, the same way as the left/right arrow keys do. After modifying the position or size of an existing redaction box, pressing the [S] key will apply the change to a single frame, like click the button of the same name.
Updates
  • If GPU is present, skip building the models that are not used in the UI.
  • Show the Export Range option in the Render and Export modal, even if no redactions have been specified. This is useful to just trim the video.
Bug Fixes
  • Displayed video editing features for audio files.

6.0.3 - October 1s7, 2024

Updates
  • AWS/S3 can now be used without explicitly providing credentials via Redactor environment variables. Counting on properly set up system for AWS (environment etc).

6.0.2 - October 15, 2024

Updates
  • Enhanced crash handling for processing videos.
Bug Fixes
  • Fixed issue with audio truncation.

6.0.1 - October 10, 2024

New Features
  • Audio only redaction is now possible.
  • (API) Audio regions can be passed in via API request
Updates
  • (API) Provided session identifier is now emitted in API progress messages

6.0.0 October 7, 2024

New Features
  • Head Detection added as option to auto-detect - replaces face detection, which is still available via the API.
  • Display custom metadata from API in the video list.
  • Search box added to projects page. It will search filenames and custom metadata if provided via API calls.
  • (API) support for specifying export video range in seconds.
  • (API) Emitted render event now carries video range as timestamps in seconds.
Updates
  • Rename Videos route to Media in Admin section
  • Removed Faces from auto-detection choices and replaced with heads.
  • Attempt to pull new license from remote licensing server if local one is expired, instead of becoming unlicensed.
  • Head/face box expansion can be specified via environment variable REDACTOR_PROCESS_(HEAD|FACE)BOX_EXPAND, API and also profiles.
  • Export dialog now displays only options which mater, e.g. audio redaction options will not be displayed if the media doesn’t contain audio.
Bug Fixes
  • (API) Filename/URI sanitation in the export URI needed (%, etc.)
  • Display the export range for audio-only redactions.
  • Fixed bug with exported video (missing last frame and playback issues).
  • Fix socket transport disconnection trigger on destroy (closing the wrong session).