Looking for the previous docs site? Visit the legacy docs homepage. (opens in a new tab)
Agora Docs

Release notes

Updated

Information about changes in each release of Voice Calling.

Known issues

The list of known issues page is continuously updated as the systems evolve. Agora suggests you regularly upgrade to the latest version of the SDK, which includes new features, bug fixes, and improvements.

  • Unreal SDK v4.4.0

    MacOS 15 Sequoia beta system issue

    When the app path (including the App name) contains non-English characters (such as Chinese characters), the compiled Metal shader library may not contain any methods, causing the app to crash. See this article for the workaround.

  • Unreal SDK v4.2.1

    Audio module incompatibility (iOS)

    The audio module of the Agora Unreal SDK is incompatible with the AudioMixer module in Unreal Engine 5.3 and 5.4. If you use Unreal Engine 5.3 or 5.4, you need to add the following code to the /Config/IOS/IOSEngine.ini file in your project to disable Unreal Engine’s AudioMixer module and avoid conflicts.

    [Audio]
AudioMixerModuleName=

v4.5.1

v4.5.1 was released on March 13, 2025.

New features

  1. AI conversation scenario

    This version adds the AUDIO_SCENARIO_AI_CLIENT audio scenario specifically designed for interacting with the conversational AI agent created by Conversational AI Engine. This scenario optimizes the audio transmission algorithm based on AI-generated voice characteristics, ensuring stable voice data transmission even in weak network conditions with up to 80% packet loss. It enhances conversation continuity and reliability while adapting to complex network conditions.

Improvements

  • Reduces the SDK initialization time on specific device models. (iOS)

Issues fixed

This release fixes the following issues:

  • Apps that integrated the Agora SDK and set the targetSdkVersion to 34 encountered crashes when attempting to enable screen sharing for the first time on an Android 14 system. (Android)
  • When joining two or more channels simultaneously, and calling the takeSnapshotEx method to take screenshots of the local video streams in each channel consecutively, the screenshot of the first channel failed.
  • When using the pause method to pause playback, then calling seek to move to a specified position, and finally calling play to continue playback, the Media Player resumed from the position where it was paused, not the new specified position.
  • When using the Media Player, the file path of the media resource returned by the getPlaySrc did not change after calling the switchSrc method to switch to a new media resource.
  • When using Bluetooth headphones on specific device models for audio and video interactions, adjusting the phone volume would occasionally change the media volume instead of the Bluetooth volume. (Android)
  • During audio and video interactions, the local user occasionally experienced a black screen when watching the video streams of remote users. (Android)
  • On specific device models, after calling setCameraExposureFactor to set the exposure coefficient of the current camera at a specific angle of the device, the video screen occasionally became dark when the device was moved to another angle. (Android)
  • When playing a CDN live stream, the video occasionally froze for an extended period after recovering from an interruption. (Android)
  • When pushing video frames in i420 format to the channel, using CVPixelBuffer to handle these frames caused a crash. (iOS)
  • Calling setupLocalVideo to set up two views, then calling enableFaceDetection to start face detection, no face information can be detected in the subsequently passed views. (iOS)
  • In a screen sharing scenario, the receiving-end user saw a green line on the shared image. (iOS)
  • In the interactive live streaming scenario, after joining a channel to watch live streams using string user id, the audience members occasionally saw that the audio was not synchronized with the video.
  • Plugins sometimes did not work when using AI noise suppression and AI echo cancellation plugins at the same time.

v4.5.0

v4.5.0 was released on December 10, 2024.

Compatibility changes

This version includes optimizations to some features, including changes to SDK behavior, API renaming and deletion. To ensure normal operation of the project, update the code in the app after upgrading to this release.

  1. Changes in strong video noise suppression implementation

    This version adjusts the implementation of strong video noise suppression. The VIDEO_DENOISER_LEVEL removes VIDEO_DENOISER_LEVEL_STRENGTH. Instead, after enabling video noise suppression by calling setVideoDenoiserOptions, you can call the setBeautyEffectOptions method to enable the beauty skin smoothing feature. Using both together will help achieve better video noise suppression effects. For strong noise suppression, it is recommended to set the skin smoothing parameters as detailed in setVideoDenoiserOptions.

    Additionally, due to this adjustment, to achieve the best low-light enhancement effect with a focus on image quality, you need to enable video noise suppression first and use specific settings as detailed in setLowlightEnhanceOptions.

  2. Changes in camera plug and unplug status (macOS, Windows)

    In previous versions, when the camera was unplugged and replugged, the onVideoDeviceStateChanged callback would report the device status as MEDIA_DEVICE_STATE_ACTIVE(1) (device in use). Starting from this version, after the camera is replugged, the device status will change to MEDIA_DEVICE_STATE_IDLE(0) (device ready).

  3. 16 KB memory page size (Android)

    Starting from Android 15, the system adds support for 16 KB memory page size, as detailed in Support 16 KB page sizes. To ensure the stability and performance of the app, starting from this version, the SDK supports 16 KB memory page size, ensuring seamless operation on devices with both 4 KB and 16 KB memory page sizes, enhancing compatibility and preventing crashes.

  4. Changes in video encoding preferences

    To enhance the user’s video interaction experience, this version optimizes the default preferences for video encoding:

    • In the COMPRESSION_PREFERENCE enumeration class, a new PREFER_COMPRESSION_AUTO (-1) enumeration is added, replacing the original PREFER_QUALITY (1) as the default value. In this mode, the SDK will automatically choose between PREFER_LOW_LATENCY and PREFER_QUALITY based on your video scene settings to achieve the best user experience.
    • In the DEGRADATION_PREFERENCE enumeration class, a new MAINTAIN_AUTO (-1) enumeration is added, replacing the original MAINTAIN_QUALITY (1) as the default value. In this mode, the SDK will automatically choose between MAINTAIN_FRAMERATE, MAINTAIN_BALANCED, and MAINTAIN_RESOLUTION based on your video scene settings to achieve the optimal overall quality of experience (QoE).

New features

  1. Live show scenario

    This version adds the APPLICATION_SCENARIO_LIVESHOW(3) (Live Show) enumeration to VIDEO_APPLICATION_SCENARIO_TYPE. You can call setVideoScenario to set the video business scenario to show room. In this scenario, fast video rendering and high image quality are crucial. The SDK implements several performance optimizations, such as enabling accelerated audio and video frame rendering to minimize first-frame latency (no need to call enableInstantMediaRendering) to achieve better image quality and bandwidth efficiency.

  2. Maximum frame rate for video rendering

    This version adds the setLocalRenderTargetFps and setRemoteRenderTargetFps methods, which support setting the maximum frame rate for video rendering locally and remotely. The actual frame rate for video rendering by the SDK will be as close to this value as possible.

    In use-cases where the frame rate requirement for video rendering is not high (e.g., screen sharing, online education) or when the remote end uses mid-to-low-end devices, you can use this set of methods to limit the video rendering frame rate, thereby reducing CPU consumption and improving system performance.

  3. Filter effects

    This version introduces the setFilterEffectOptions method. You can pass a cube map file (.cube) in the config parameter to achieve custom filter effects such as whitening, vivid, cool, black and white, etc. Additionally, the SDK provides a built-in built_in_whiten_filter.cube file for quickly achieving a whitening filter effect.

  4. Local audio mixing

    This version introduces the local audio mixing feature. You can call the startLocalAudioMixer method to mix the audio streams from the local microphone, media player, sound card, and remote audio streams into a single audio stream, which can then be published to the channel. When you no longer need audio mixing, you can call the stopLocalAudioMixer method to stop local audio mixing. During the mixing process, you can call the updateLocalAudioMixerConfiguration method to update the configuration of the audio streams being mixed.

    Example use cases for this feature include:

    • By utilizing the local video mixing feature, the associated audio streams of the mixed video streams can be simultaneously captured and published.
    • In live streaming use-cases, users can receive audio streams within the channel, mix multiple audio streams locally, and then forward the mixed audio stream to other channels.
    • In educational use-cases, teachers can mix the audio from interactions with students locally and then forward the mixed audio stream to other channels.

  5. External MediaProjection (Android)

    This version introduces the setExternalMediaProjection method, which allows you to set an external MediaProjection and replace the MediaProjection applied by the SDK.

    If you have the capability to apply for MediaProjection on your own, you can use this feature to achieve more flexible screen capture.

  6. EGL context (Android)

    This version introduces the setExternalRemoteEglContext method, which is used to set the EGL context for rendering remote video streams. When using Texture format video data for remote video self-rendering, you can use this method to replace the SDK's default remote EGL context, achieving unified EGL context management.

  7. Color space settings

    This version adds the colorSpace parameter to VideoFrame and ExternalVideoFrame. You can use this parameter to set the color space properties of the video frame. By default, the color space uses Full Range and BT.709 standard configuration. You can flexibly adjust according to your own capture or rendering needs, further enhancing the customization capabilities of video processing.

  8. Other features

    • The onLocalVideoStateChanged callback adds the LOCAL_VIDEO_STREAM_REASON_DEVICE_DISCONNECTED enumeration, indicating that the currently used video capture device has been disconnected (e.g., unplugged). (Windows)
    • The MEDIA_DEVICE_STATE_TYPE adds the MEDIA_DEVICE_STATE_PLUGGED_IN enumeration, indicating that the device has been plugged in. (Windows)

Improvements

  1. Virtual background algorithm optimization

    This version upgrades the virtual background algorithm, making the segmentation between the portrait and the background more accurate. There is no background exposure, the body contour of the portrait is complete, and the detail recognition of fingers is significantly improved. Additionally, the edges between the portrait and the background are more stable, reducing edge jumping and flickering in continuous video frames.

  2. Snapshot at specified video observation points

    This version introduces the takeSnapshot [2/2] and takeSnapshotEx [2/2] methods. You can use the config parameter when calling these methods to take snapshots at specified video observation points, such as before encoding, after encoding, or before rendering, to achieve more flexible snapshot effects.

  3. Custom audio capture improvements

    This version adds the enableAudioProcessing member parameter to AudioTrackConfig, which is used to control whether to enable 3A audio processing for custom audio capture tracks of the AUDIO_TRACK_DIRECT type. The default value of this parameter is false, meaning that audio processing is not enabled. Users can enable it as needed, enhancing the flexibility of custom audio processing.

  4. Other improvements

    • In use-cases where Alpha transparency effects are achieved by stitching video frames and Alpha data, the rendering performance on the receiving end has been improved, effectively reducing stuttering and latency. (Android, iOS)
    • Optimizes the logic for calling queryDeviceScore to obtain device score levels, improving the accuracy of the score results.
    • Supports using virtual cameras in YV12 format as video capture devices. (Windows)
    • When calling switchSrc to switch between live streams or on-demand streams of different resolutions, smooth and seamless switching can be achieved. An automatic retry mechanism has been added in case of switching failures. The SDK will automatically retry 3 times after a failure. If it still fails, the onPlayerEvent callback will report the PLAYER_EVENT_SWITCH_ERROR event, indicating an error occurred during media resource switching.
    • When calling setPlaybackSpeed to set the playback speed of an audio file, the minimum supported speed is 0.3x.

Issues fixed

This version fixes the following issues:

  • When calling startScreenCaptureByWindowId to share the screen, the window capture area specified by regionRect was inaccurate, resulting in incorrect width and height of the screen sharing window seen by the receiving end. (Windows)
  • When the video source type of the sender is in JPEG format, the frame rate on the receiving end occasionally falls below expectations. (Android, iOS)
  • During audio and video interaction, after being interrupted by a system call, the user volume reported by the onAudioVolumeIndication callback was incorrect. (Android)
  • When the receiving end subscribes to the low-quality video stream by default and does not automatically subscribe to any video stream when joining the channel, calling muteRemoteVideoStream(uid, false) after joining the channel to resume receiving the video stream results in receiving the high-quality video stream, which is not as expected. (Android)
  • Occasional errors of not finding system files during audio and video interaction on Windows 7 systems. (Windows)
  • When calling followSystemRecordingDevice or followSystemPlaybackDevice to set the audio capture or playback device used by the SDK to not follow the system default audio playback device, the local audio state callback onLocalAudioStateChanged is not triggered when the audio device is removed. (Windows)
  • Occasional instances where the receiving end cannot hear the sender during audio and video interaction. (iOS)
  • During audio and video interaction, if the sender's device system version is iOS 17, the receiving end occasionally cannot hear the sender. (iOS)
  • In live streaming use-cases, the time taken to reconnect to the live room after the audience end disconnects due to network switching is longer than expected. (iOS)
  • No sound when playing online media resources using the media player after the app starts. (iOS)
  • Occasional instances of no sound in audio capture after resuming from being interrupted by other system apps during audio and video interaction. (iOS)
  • Calling startAudioMixing [1/2] and then immediately calling pauseAudioMixing to pause the music file playback does not take effect.

v4.4.0

v4.4.0 was released on August 29, 2024.

Compatibility changes

  1. Changes to API usage

    Starting from this release, the [Get](https://api-ref.agora.io/en/voice-sdk/blueprint/4.x/API/class_irtcengine .html#api_createagorartcengine) and [Release](https://api-ref.agora.io/en/voice-sdk/blueprint/4.x/API/class_irtcengine .html#api_irtcengine_release) methods are used for creation and destruction of engine instances, replacing the previous createAgoraRtcEngine and release methods. This change simplifies instance management. Details are as follows:

New features

  1. Voice AI tuner

    This version introduces the voice AI tuner feature, which can enhance the sound quality and tone, similar to a physical sound card. You can enable the voice AI tuner feature by calling the enableVoiceAITuner method. Pass in the sound effect types supported in the VOICE_AI_TUNER_TYPE enum to achieve effects like deep voice, cute voice, husky singing voice, and so on.

  2. Privacy manifest file (iOS)

    To meet Apple's safety compliance requirements for app publication, the SDK now includes a privacy manifest file, PrivacyInfo.xcprivacy, detailing the SDK's API calls that access or use user data, along with a description of the types of data collected.

    If you need to publish an app with SDK versions prior to v4.4.0 to the Apple App Store, add the PrivacyInfo.xcprivacy file to your Xcode project.

  3. Multi-camera capture (Android)

    This release introduces additional functionalities for Android camera capture:

    1. Support for capturing and publishing video streams from the third and fourth cameras:
    • VIDEO_SOURCE_CAMERA_THIRD (11) and VIDEO_SOURCE_CAMERA_FOURTH (12) in VIDEO_SOURCE_TYPE add support for Android, specifically for the third and fourth camera sources. This change allows you to specify up to four camera streams when initiating camera capture by calling startCameraCapture.
    • publishThirdCameraTrack and publishFourthCameraTrack in ChannelMediaOptions add support for Android. Set these parameters to true when joining a channel with joinChannel[2/2] to publish video streams captured from the third and fourth cameras.
    1. Support for specifying cameras by camera ID:

    A new parameter cameraId is added to CameraCapturerConfiguration. For devices with multiple cameras, where cameraDirection cannot identify or access all available cameras, you can obtain the camera ID through Android's native system APIs and specify the desired camera by calling startCameraCapture with the specific cameraId.

  4. Select different audio tracks for local playback and streaming

    This release introduces the selectMultiAudioTrack method that allows you to select different audio tracks for local playback and streaming to remote users. For example, in use-cases like online karaoke, the host can choose to play the original sound locally and publish the accompaniment in the channel. Before using this function, open the media file through the openWithMediaSource method and enable this function by setting the enableMultiAudioTrack parameter in MediaSource.

  5. Update video screenshot and upload

    To facilitate the integration of third-party video moderation services from Agora Extensions Marketplace, this version has the following changes:

    • The CONTENT_INSPECT_IMAGE_MODERATION enumeration is added in CONTENT_INSPECT_TYPE, which means using video moderation extensions from Agora Extensions Marketplace to take video screenshots and upload them.
    • An optional parameter serverConfig is added in ContentInspectConfig, which is for server-side configuration related to video screenshot and upload via extensions from Agora Extensions Marketplace. By configuring this parameter, you can integrate multiple third-party moderation extensions and achieve flexible control over extension switches and other features. For more details, please contact technical support.

    Additionally, this version introduces the enableContentInspectEx method, which supports simultaneous screenshot and upload of multiple video streams.

  6. ID3D11Texture2D rendering (Windows)

    As of this release, the SDK supports video formats of type ID3D11Texture2D, improving the rendering effect of video frames in game use-cases. You can set format to VIDEO_TEXTURE_ID3D11TEXTURE2D when pushing external raw video frames to the SDK by calling pushVideoFrame. By setting the textureSliceIndex property, you can determine the ID3D11Texture2D texture object to use.

  7. Wildcard token

    This release introduces wildcard tokens. Agora supports setting the channel name used for generating a token as a wildcard character. The token generated can be used to join any channel if you use the same user ID. In use-cases involving multiple channels, such as switching between different channels, using a wildcard token can prevent repeated application of tokens every time users join a new channel, which reduces the pressure on your token server. See Generate wildcard tokens.

    All 4.x SDKs support using wildcard tokens.

  8. Preloading channels

    This release adds preloadChannel and preloadChannelWithUserAccount methods, which allows a user whose role is set as audience to preload channels before joining one. Calling the method can help shorten the time of joining a channel, thus reducing the time it takes for audience members to hear and see the host.

    When preloading more than one channels, Agora recommends that you use a wildcard token for preloading to prevent repeated application of tokens every time you join a new channel, thus saving the time for switching between channels. See Generate wildcard tokens.

  9. Publishing video streams from different sources (Windows, macOS)

    This release adds the following members in ChannelMediaOptions to allow you to publish video streams captured from the third and fourth camera or screen:

    • publishThirdCameraTrack: Publishing the video stream captured from the third camera.
    • publishFourthCameraTrack: Publishing the video stream captured from the fourth camera.
    • publishThirdScreenTrack: Publishing the video stream captured from the third screen.
    • publishFourthScreenTrack: Publishing the video stream captured from the fourth screen.

    For one RtcConnection, Agora supports publishing multiple audio streams and one video stream at the same time.

Improvements

  1. Header file management simplification

    To reference other header files, you can add them uniformly to AgoraHeaderBase.h, simplifying the management of header files in the project and facilitating future maintenance.

  2. Channel media relay

    The number of target channels for media relay has been increased to 6. When calling startOrUpdateChannelMediaRelay and startOrUpdateChannelMediaRelayEx, you can specify up to 6 target channels.

Issues fixed

In Unreal Engine 5.3 and 5.4, adding the following code to the /Source/xxx.Target.cs file in your project to solve iOS memory allocation issues may lead to missing iOS compilation symbols. To address this issue, this version adds three function definitions (StdMalloc, StdRealloc, and StdFree) related to the standard library in the UnrealMemory.cpp file, ensuring proper handling of memory management needs when using the standard memory allocator (iOS).

if (Target.Platform == UnrealTargetPlatform.IOS)
{
    bOverrideBuildEnvironment = true;
    GlobalDefinitions.Add("FORCE_ANSI_ALLOCATOR=1");
}

v4.2.1

v4.2.1 was released on November 16, 2023.

This the first GA release of the SDK. It provides a simple and user-friendly real-time interactive solution for Unreal Engine users.

C++

The SDK for Unreal Engine provides a powerful set of C++ APIs, allowing developers to easily build games or apps with real-time interactive features. It can also leverage the powerful graphics engine functionality of Unreal Engine to add various visual effects and interactive performance to the app, providing users with a more immersive experience.

Blueprint

The SDK for Unreal Blueprint offers a visual interface and a set of intuitive and easy-to-use nodes. This enables developers to effortlessly implement real-time interactive features in games or apps within the Unreal project. No need to know C++ - just drag, connect, and configure nodes.

Read more:

Agora provides an open source example project on GitHub for your reference.

If you have already integrated the 3.x version of the SDK and wish to upgrade to this version, please refer to Migration Guide for upgrade instructions.

Contact Us

+1 (408) 879-5885

2804 Mission College Blvd.Santa Clara, CA, USA 95054

AgoraCopyright © 2026 AgoraAll rights reserved