# **Add real-time audio separation to any pipeline** The AudioShake SDK integrates streaming-capable, real-time sound separation right into your app or self-deployed Enterprise service. AudioShake's stem separation SDK enables you to separate vocals, isolate instruments, remove music, and clean speech, in real-time, with industry-leading quality. Available on iOS/MacOS, Android, Windows, and Linux platforms, with local inference times optimized for each. You’ll always get the best speed wherever you deploy. [MORE ON AUDIOSHAKE'S SDK](https://developer.audioshake.ai/sdk) ## What is a stem separation SDK? Real-time separation processes live audio streams as they happen — isolating dialogue, vocals, or instruments instantly, without sending files to the cloud or waiting for post-production. AudioShake's stem separation SDK runs models locally on-device, making clean, separated audio available before it reaches the next stage of your pipeline. For dubbing and captioning, dialogue is isolated from crowd noise or background music the moment it's captured. In broadcast, music can be removed from streams to ensure rights compliance. For speech workflows, developers can turn messy, real-world audio into clean, structured inputs for ASR and LLM systems. And for music apps, stem-level control lets users interact and mix tracks in real time. - On-device, no cloud processing - 11ms dialogue isolation latency - Up to 200x real-time inference ## Separation models across voice, film, TV, and music AudioShake's SDK gives developers access to real-time music removal, dialogue separation, and instrument stem isolation – all with low latency performance, on-device, across iOS, Android, Windows, and Linux. ### DIALOGUE **Dialogue Isolation** [View product page →](/content/products/dialogue-music-effects-separation/index.html) Isolates spoken dialogue from background sound in real-time streams. Cleans voice inputs before they reach ASR, transcription, translation, or A1 audio engineer workflows — with ~25% improvement in ASR accuracy in noisy audio environment. ### DIALOGUE RT **Low-Latency Dialogue Isolation** New – 11ms latency Dialogue RT delivers 11ms latency for live broadcast workflows — built for live sports, news, commentary, transcription, and real-time speech applications. [Read more about Dialogue RT →](/content/post/audioshake-debuts-dialogue-rt-the-first-ai-dialogue-isolation-model-built-for-live-broadcast/index.html) ### COPYRIGHT **Commercial Music Removal** [View product page →](/content/products/music-removal/index.html) Removes copyrighted background music from live or streamed audio while preserving dialogue and effects. Built for broadcasters, sports commentary, and content platforms managing copyright compliance on live feeds. ### MUSIC **Instrument Stem Separation** [View product page →](/content/products/instrument-stem-separation/index.html) Isolates vocals, instruments, or up to 14 different instrument stems from any song in real time. Used by music apps, learning platforms, and DJ tools to give users stem-level control inside a consumer experience. ## Built for production workflows The SDK runs locally on-device, integrates in a few lines of code, and fits into mobile apps, desktop DAWs, live streaming platforms, embedded devices, and high-volume on-premise media processing workflows. - 11ms dialogue isolation latency–first to meet live broadcast threshold - 200x real-time inference speeds - 25% reported ASR accuracy improvement with SDK preprocessing ## Authentic and scalable speech recovery ### VOICE AI Improve transcription accuracy with speech isolation Models as small as 9mb 200x realtime 11ms latency NPU/GPU/CPU runtimes available for real-time performance Native support for low-res and high-res audio ### MUSIC Power music production, mixing, songwriting, and education apps Up to 14 instrument targets or joint 4-stem, 6-stem, drum kits, vocals available Up to 250x real-time processing (vocals) with per-platform optimizations Cross-platform SDKs available: iOS/MacOS, Windows, Linux, Android ### BROADCAST Remove copyright material from your audio Isolate clean dialogue from crowd noise Streaming-capable dialogue and music removal models Models as small as 9mb 200x realtime 11ms latency Support for hi-res audio ### ON PREM/SELF-DEPLOYED Run any of AudioShake’s edge or API models in your own cloud or offline All API models are available for safe and secure local inference Manage compute and process large amounts of data Streaming or batch API available ## Get started with our SDK ### SDK Bring sound separation to your edge device On-device inference, no cloud round-trip, under 50ms latency or better. Includes sample apps, integration guides, and demo code. Contact to access. [REQUEST ACCESS](/content/contact-us/index.html) ### API **Evaluate before committing to on-device** Full model access via our API. No hardware requirements. Same separation quality, cloud-based. Ideal for prototyping, batch processing, or teams not yet building for edge. [ACCESS NOW](https://developer.audioshake.ai/) ## FAQ No items found. AutoTune Mobile App Adds Stems [Mixing](/content/case-studies/autotune-mobile-app-adds-stems/index.html) Studio-Quality Stems on LANDR [Mixing](/content/case-studies/studio-quality-stems-on-landr/index.html) Clear Commentary for Live Sports Translation [Localization](/content/case-studies/clear-commentary-for-live-sports-translation/index.html) Tuned Global Unlocks Stems for Users [Mixing](/content/case-studies/tuned-global-unlocks-stems-for-users/index.html) Real-Time Fan Remixing [Mixing](/content/case-studies/real-time-fan-remixing/index.html) Mobile Stem Separation for Musicians [Mixing](/content/case-studies/mobile-stem-separation-for-musicians/index.html)