mediapipe

Bazel changes

• Bump MediaPipe version to 1.2.0
• Support Web Worker platform detection and Flutter Web telemetry overrides.
• Rollback due to breakage in gmscore.build.
• Fix GPU memory leaks when releasing AHWB-backed SSBOs and pooled textures.
• Fix nullptr exception in ImageToTensorCalculator
• Add support for F16 formats in ImageFrame
• Fix MediaPipe Tasks telemetry fields and audio_classifier.h license.
• Prevent possible lock-order-inversion deadlocks when locking input tensors in InferenceRunnerLiteRt
• Expose the remaining MlDriftDelegateOptions in InferenceCalculatorOptions
• Run face blendshapes on CPU when using the GPU delegate.
• Optimize Parakeet ASR streaming latency and decoder/runtime performance.
• Replace status_macros.h includes with absl/status/status_macros.h in mediapipe/framework
• Replace status_macros.h includes with absl/status/status_macros.h in mediapipe/calculators
• Replace status_macros.h includes with absl/status/status_macros.h in remaining mediapipe tasks, modules, and utilities
• Upgrade OpenCV to 4.10
• Enable F16 cv::Mat support in OpenCV conversions. OpenCV 4 was enabled in a previous CL.
• Update SPM files for 1.2.0

MediaPipe Tasks update

This section should highlight the changes that are done specifically for any platform and don't propagate to other platforms.

Android

• Add MediaPipe Tasks Android Kotlin TextToSpeech SDK and JNI bridge.
• Add LiteRT service support to InteractiveSegmenter on Android.
• Return topK distinct documents from SemanticRetriever when results are chunked.

iOS

• Open source the MediaPipe text proofreader and summarizer tasks.
• Add Apple Privacy Manifest (PrivacyInfo.xcprivacy) to MediaPipe Tasks iOS Swift Package Manager package.

MediaPipe Dependencies

• Update WASM files for 1.2.0 release
• Update genai wasm

Bazel changes

• Bump MediaPipe version to 1.1.1
• Add per-domain C libraries for Flutter SDK
• Fix GPU interactive segmentation on macOS.
• Add Flutter and Flutter Web host environments to MediaPipe Tasks C API.
• Add C API for intercepting Abseil logs in MediaPipe Tasks.
• mediapipe: Migrate to use new LiteRT GetOptions() API
• Migrate remaining NPU option callers to GetOptions()
• Update WASM and SPM files for 1.1..1 release

MediaPipe Tasks update

This section should highlight the changes that are done specifically for any platform and don't propagate to other platforms.

Android

• Enable LiteRT delegate for Interactive Segmenter on Android.

Javascript

• Update Retrieval SDK's code snippets

MediaPipe Dependencies

• Upgrade LiteRT dependency to v2.3.0 in MediaPipe WORKSPACE

Build and Bazel changes

• Remove the dependency on internal Bazel Apple rules resources.bzl in mediapipe/metal.bzl, using the metal rule provided by Bazel Apple rules.
• Add Python 3.13 and 3.14 support to the MediaPipe Bazel build.
• Upgrade LiteRT dependency to v2.2.0 in MediaPipe WORKSPACE.
• Move up declaration of kissfft in build configuration.
• Fix xla llvm download URL by using mediapipe-assets mirror.
• Fix Windows build.
• Remove obsolete unistd.h header imports.
• Configure MediaPipe for vendor/no-JNI Android builds.

Framework and core calculator improvements

• Bump MediaPipe version to 1.1.0.
• Add MediaPipe Universal Embedder, Semantic Retriever, and LiteRT-LM integration across C++, C, Python, and Android/Maven APIs.
• MediaPipe Decision: Add EmbeddingGemma 1 & 2 decision backend with multimodal state and option support.
• Implement core TaskPart and EmbeddingProvider interface.
• Refactor TextPart, AudioPart, and ImagePart into tasks.core and unify duplicates.
• Refactor TextEmbedder and ImageEmbedder to use executor abstractions (TextEmbedderExecutor, ImageEmbedderExecutor).
• Add HasFile, GetBundleManifest, and GetBundleId helpers to ModelAssetBundleResources.
• Add noexcept to move ctor/operator and add swap to Packet.
• Add method to create an empty packet in Java API.
• Use ABSL_ASSIGN_OR_RETURN and ABSL_RETURN_IF_ERROR in MediaPipe.
• Rename GeckoPreprocessorCalculator to SentencePiecePreprocessorCalculator.
• Add option to ignore missing encoded_media_bytes in UnpackMediaSequenceCalculator.
• Add LiteRT-LM based C++ Audio Embedder.
• C API for MediaPipe AudioEmbedder.
• Select the MediaPipe OpenGL device by Vulkan device UUID.
• Replace LiteRT delegate with use_litert option in MediaPipe BaseOptions.
• Expose LiteRT and NPU delegate options to MediaPipe Tasks.
• Only use public LiteRT API in LiteRT inference calculator.
• Support passing LiteRtSystemHandles to Environment in LiteRT Inference Calculator.
• Add support for specifying a signature key for LiteRT inference.
• Retain MTLCommandQueue directly instead of DrishtiMetalHelper in InferenceRunnerLiteRt.
• Use runtime layout for rank-0 LiteRT model outputs in InferenceRunnerLiteRt.
• Check GPU service availability in TensorsToDetectionsCalculator.
• Improve MediaPipe Metal GPU delegate error messages for unsupported configurations.
• Allow size mismatched tensors for GPU execution.
• Fix race condition in GpuResources.
• Fix deprecated CreateSourceTexture usage in desktop GPU example.

MediaPipe Tasks update

This section highlights the changes that are done specifically for each platform and don't propagate to other platforms.

Android

• Add plumbing needed for zero-copy AHardwareBuffer (AHWB) output from the GPU.
• Fix OpenGL ES and AHardwareBuffer synchronization when converting between write and read views.
• Wrap VC effects AHWB graph input as AHardwareBuffer (kRGBA32).
• Raise MediaPipeException instead of crashing in surface_output_jni on Android EGL surface creation failure.
• Fix crash in PathToResourceAsFile when TEST_SRCDIR is not set on Android.
• Made AppSearch vector store optional for SemanticRetriever Android SDK.
• Add Semantic Retriever Java API by prebuilding SQLite dependencies.
• Support in-memory audio and image representations in MediaPipe Tasks Android.
• Support multi-part multimodal embedding and vector store serialization in MediaPipe Tasks Android.
• Ensure that JNI methods are exported for Android Tasks.
• Remove legacy LITERT delegate flag from Android API.

iOS

• Update Swift Package Manager (SPM) manifest for MediaPipe Tasks iOS 1.1.0.
• Add iOS Retrieval SPM package (MediaPipeTasksRetrieval).
• Add missing MPPAudioRecord.h to MediaPipeTasksRetrieval XCFramework headers.
• Exclude MediaPipe C API (_Mp*) symbols from the iOS symbol rename

Build changes

• Remove MacOS prebuilts
• Replace ITUIOSTestUtil with ITUTestRunEnvironment in iOS tests.

Bazel changes

• Migrate Apple rules dependencies to MODULE.bazel.

Framework and core calculator improvements

• Add Gecko support to MediaPipe Text Embedder Task
• Store wgpu::Instance in WebGpuService.
• Supporting builds with DRISHTI_DISABLE_GL_COMPUTE define.
• Bump MediaPipe version to 0.10.36.
• • Abseil (com_google_absl): Updated from (2023-10-18) to (2025-01-14).
• Add Gecko support to TextEmbedder C API.
• Use standardized enum constants for Running Mode
• Report Mat type in debug logging.
• Add AddAudioHeaderCalculator
• Remove run_llm_inference.sh
• Add webgpu::Instance to WebGpuDeviceAttachmentManager.
• Fix MediaPipe nightly build failure and cleanup Bzlmod warnings.
• Fix null pointer access in Tensor::GetOpenGlTexture2dReadView() for AHWB backed Tensor.
• Adds SpectrogramFloatCalculator and InverseSpectrogramFloatCalculator to support float precision.
• Add support for Embedding Gemma models in TextEmbedder.
• Make wgpu::Instance a member of WebGpuAsyncFuture.
• Enforce namespacing conventions for MediaPipe Tasks C and Python APIs.
• Add methods to create and access Matrix packets using Direct ByteBuffers.
• Ensure AhwbGpuResources destructor cleans up OpenGL/EGL resources and forces AHWB usage completion.
• Update MediaPipe Python build Dockerfiles.
• Update base options and host environment.
• Introducing extra base options and corresponding conversion/get functions.
• Improve error message for option_value problems
• Support models with empty tags
• Add empty image checks to ScaleImageCalculator.
• Add gpu_postprocessing option for TensorsToSegmentationCalculator.
• Add text chunking utility
• Expand LiteRtService to support system runtime on Android.
• Add vImageExtractChannelAndConvertToFloat function.
• Compile with OpenCV 5's features
• Add documentation for LiteRT Inference Calculator accelerators
• Fix data race accessing GlContext::profiling_helper_
• Add a "prefer_ahwb" flag to MemoryManager
• Add Builder v2 and v3 APIs to change the node name
• Add verbose debug logging for buffer pooling
• Updates the manylinux Dockerfile and related build scripts to fix the
• Re-format GraphRunner
• Add EXIF orientation support to MpImageCreateFromFile and the Python Task API

MediaPipe Tasks update

This section should highlight the changes that are done specifically for any platform and don't propagate to other platforms.

Android

• Revert Java switch statement syntax update
• Introducing utils to convert BaseOptions <-> proto::BaseOptions.
• Release TextSummarizer and summarization_200m model
• Release TextProofreader and proofread_200m model
• Update TextSummarizer to load from file descriptor

iOS

• Add Gecko support to iOS Text Embedding API
• Remove iOS and Android LLM Inference Engine from release
• appId/appVersion/osVersion to common utility.
• Makes InteractiveSegmenter a legacy one, in preparation of releasing v2 / new interface.
• Add Documents path fallback for MPPFileInfo.
• Add TextProofreader and TextSummarizer to MediaPipe Tasks iOS.
• Add AudioToolbox framework to MediaPipeTasksCommon podspec.

Javascript

• Add Gecko support to MP Text Embedder for Web
• Add a Module.BAZEL to MediaPipe and migrate JS depdendencies to Aspect JS
• Add dummy logging framework
• Add MediaPipe Tasks Privacy Notice
• Add IIFE bundles for web task modules.
• Export InteractiveSegmenter methods for public visibility
• Allow MP Web LLM Inference to load .litertlm files containing web-formatted LLMs
• MP Web LLM Inference: add new experimental 'disableRewinding' flag for testing long-context support
• Cache running mode in VisionTaskRunner.

Python

• Add Gecko support to the Python Text Embedder API
• Remove MediaPipe Python GenAI components f
• TextSummarizer Python API OSS
• TextProofreader Python API OSS