You can run audio-to-MIDI inference in a browser, but the strongest documented examples here transcribe piano—not arbitrary mixtures of instruments. A practical design combines browser audio decoding, preprocessing, model inference, note-event decoding, and MIDI export. Treat “runs entirely in the browser” as an inference architecture, then verify and clearly describe every other network request the finished app makes.
What “polyphonic audio-to-MIDI” can reliably mean
Polyphonic means that multiple notes can sound at once; it does not, by itself, mean that a system can identify every instrument in a mixed recording. The best-supported examples for this browser-based design are piano transcription systems. Google’s Onsets and Frames is specifically about polyphonic piano music, and the Magenta.js music documentation describes an OnsetsAndFrames implementation for converting raw audio to MIDI in a browser. Neither establishes dependable transcription of arbitrary multi-instrument recordings.
As an Amazon Associate I earn from qualifying purchases.
Set the product’s scope accordingly. If the model is piano-focused, say so in the interface and documentation rather than implying that any song will become a clean, instrument-separated MIDI arrangement.
Quick wins for a faster PC:
Scan for outdated or missing drivers - takes under a minuteDriver Scan →Repair Windows errors before they cause bigger problemsFix Now →Fix the driver behind crashes, sound loss and screen glitchesFind Drivers →Build the complete browser-side pipeline
Inference is only one part of a usable converter. The application must take the user’s audio through decoding and preprocessing, prepare the model input, run the model, turn its output into note events, and serialize those events as a downloadable MIDI file. This is an architecture, not a tested implementation; supported formats, resampling behavior, memory requirements, and long-file limits must be validated in the product.
#1 Best Overall
- Next-Gen Music Production and Beat Maker Essential - USB-powered MIDI keyboard controller with 25 mini velocity-sensitive keys, optimized for studio or beat production, piano-style performance, synth leads, sample triggering
- Real-Time Control and Navigation - 8x assignable 360° knobs, a vibrant full-color screen and push/turn encoder for hands-on access to settings, presets, and DAW functions, without reaching for a computer
- Iconic MPC Pads with RGB Feedback - 8 velocity- and pressure-sensitive MPC pads deliver an iconic finger-drumming experience, plus dynamic visual feedback to match your performance in studio or on the go
- Studio Instrument Collection Included - A powerful VST/AU and standalone virtual suite packing 1000+ pro-grade drums, keys, synths, bass, FX from AIR, Akai Pro and Moog, plus MPK Mini IV integrated controls
- Pre-Mapped DAW Integration - Get producing in under 15 minutes with Ableton Live Lite 12, Logic Pro, FL Studio and more; comes with an expanded DAW-mapped transport section for uninterrupted workflow
- Accept a local audio file. Define the formats and file-size limits your implementation actually supports. The cited sources do not establish a standard input-format list for this converter.
- Decode and prepare the audio. Browser audio facilities provide processing building blocks through the Web Audio API. Your implementation still needs to decide and test channel mixing, sample-rate conversion, feature extraction, and handling of long recordings.
- Load the runtime and model. An ONNX Runtime Web deployment includes more than application code: its guidance identifies the JavaScript bundle, WebAssembly files where applicable, and model files as deployed assets. See the deployment documentation.
- Run inference in the browser. A browser model library such as Magenta.js or an ONNX model executed through ONNX Runtime Web can provide the inference stage. Choose a model whose task matches your advertised musical scope.
- Decode predictions into note events. Model scores must become note-on and note-off events, with any supported attributes such as velocity and timing adjustments. The model’s output is not automatically a finished, musically useful MIDI file.
- Serialize and offer the MIDI file. Convert decoded events into a valid MIDI file and provide a download. Validate the exported file in the music software your users are likely to use; the cited sources do not prescribe a serialization implementation.
Choose a model whose scope matches the promise
Onsets and Frames: browser piano transcription
Google’s Onsets and Frames work combines onset and frame predictions for polyphonic piano transcription and also predicts relative velocity. Magenta.js documents a browser implementation for raw audio-to-MIDI use. This is a concrete route to a piano-focused browser feature, not evidence that the same model separates arbitrary instruments.
Transkun: a detailed piano-transcription pipeline
Transkun describes a piano transcription system that scores possible time intervals, decodes event intervals with a semi-CRF using dynamic programming, and predicts attributes including velocity and refined onset and offset positions before producing MIDI. Its documented command-line pipeline is useful for understanding the inference and decoding stages, but does not establish browser performance.
Rank #2
- Music Production and Beat Maker Essential -USB powered MIDI controller with 25 mini MIDI keyboard velocity-sensitive keys for studio production, virtual synthesizer control and beat production
- Total Control of your Production - Innovative 4-way thumbstick for dynamic pitch and modulation control, plus a built-in arpeggiator with adjustable resolution, range and modes
- Native Kontrol Standard (NKS) Integration - Akai Professional and Native Instruments have partnered to bring NKS support to the MPK Controller series, get ready to Kontrol straight from your MPK
- Choose Your Exclusive Complimentary NKS Bundle - Browse and control Native Instruments presets and sound libraries; select one of three curated Komplete 15 Select bundles: Beats, Band, or Electronic
- The MPC Experience - 8 backlit velocity-sensitive MPC-style MIDI beat pads with Note Repeat and Full Level for programming drums, triggering samples and controlling virtual synthesizer / DAW controls
The Transkun project’s model card reports these checkpoint metrics on MAESTRO V3:
What’s actually slowing this PC down?
Pick the symptom - the matching free tool is one click away.
| Metric | Project-reported result | What it measures |
|---|---|---|
| Activation F1 | 0.9530 | Activation overlap under the project’s evaluation convention |
| Note-onset-plus-offset F1 | 0.9349 | Notes evaluated using onset and offset |
| Note-onset-plus-offset-plus-velocity F1 | 0.9296 | Notes evaluated using onset, offset, and velocity |
These are project-reported checkpoint results, not measurements of a browser build and not guarantees for arbitrary recordings. Transkun’s repository also notes that its evaluation module does not currently support multitrack MIDI. Sustain-pedal conventions matter when interpreting note durations: the shipped checkpoint was trained without extending notes through pedal duration, while some piano-transcription conventions do extend them. A comparison is meaningful only when its evaluation and pedal conventions are understood.
Rank #3
- Mini Keys, Major Performance - 25-key MIDI controller keyboard with responsive keys to play with feel and dynamics; Enjoy expressive songwriting, composing, jamming, playing melodies, bass lines, and more
- Full Range Experience - Play every note with ease; Use the dedicated octave up and down buttons to transpose the range of your software instruments plus a sustain button for expressive performances
- Feature Packed - Use the built-in arpeggiator and clock for creative control of any of your sound sources like software instruments, synthesizers, or USB-enabled modular gear, all by one USB cable
- Customized Production - With 8 programmable preset slots you can instantly recall your mappings for DAWs, virtual instruments, and more; Create, edit, and manage all your presets from the computer
- Versatile Performance - The USB Studio MIDI controller is perfect for your songwriting, beat-making, and DJ remixes whenever you're inspired; The sleek design fits alongside your studio equipment
A separate community ONNX port shows an ONNX pipeline in a Node.js command-line tool. It is evidence that an ONNX export path exists, not evidence that the port runs in a browser or meets browser speed and memory targets.
Select a runtime and plan a fallback
ONNX Runtime Web documents WebAssembly CPU execution as well as WebGL, WebGPU, and WebNN options. Provider availability varies across browser and operating-system combinations. In the documented matrix, WebGPU is unavailable in Safari and Firefox, and support is limited to selected Chromium combinations with minimum-version notes. Recheck the current matrix for the runtime version and browsers you intend to support.
Rank #4
- Compact Design with 25 Mini Velocity-Sensitive Keys: The portable N-25 midi keyboard controller, made up of 25 scaled-down velocity-sensitive keys, is a companion for you to create melodies and riffs full of emotion
- Simple Layout and User-Friendly Functions: Donner N-25 mini MIDI is beginner-friendly, featuring 4 user programmable banks and essential tools like arpeggiator, sustain, octave, and tap tempo for musicians with limited music theory background
- Mac, iOS, PC/iPhone/iPad Compatible: Just plug and play with Mac/PC via USB-MIDI, no extra drivers needed. Also, easily connect your MIDI controller to smartphone music apps with an OTG adapter
- Multifunctional Pitch-Bend Wheel & Cool Light-up Buttons: The pitch-bend function ensures a rich and expressive timbre. In addition, the light-up button will provide you a fun visual display of your playing
- 40 Entertaining Keyboard-Teaching Courses: Beginner to advanced keyboard courses designed for music arrangement novices. 40 engaging lessons cover all arrangement essentials to help unleash your creativity
Operator coverage also differs. ONNX Runtime’s Web tutorials state: “All ONNX operators are supported by WASM but only a subset are currently supported by WebGL, WebGPU and WebNN.” A sensible deployment path is to validate the chosen model with WASM first, then enable an accelerated provider only for browser, provider, and operator combinations that you have confirmed work. Preserve a fallback for unsupported or failing acceleration paths.
Free tools Windows power users keep installed
One-click scans. No signup required.
- Check that the selected model’s operators work with the provider you plan to use.
- Test the actual browser and operating-system combinations you claim to support.
- Measure inference time and memory on representative target devices; the cited sources provide no browser-specific benchmark for these models.
- Tell users when the model and runtime assets are downloading, and provide progress or cancellation if the implementation supports them.
Make delivery and network behavior explicit
Local inference means the audio can be processed on the user’s device rather than sent to a server for inference. It does not mean the web page makes no network requests: the application, runtime assets, and model may need to be downloaded, and analytics, crash reporting, or other services may make additional requests. ONNX Runtime’s documentation discusses on-device inference and possible offline use, while its deployment guidance identifies assets that need delivery.
Best Value
- 32-note velocity sensitive mid-size key keyboard. Pitch Wheel. Modulation Wheel. Octave Buttons(Up & Down). Transpose Buttons(Up & Down).
- Sustain Switch interface (sustain pedal does not include). USB B type: USB connect to Computer
- Attachment Type: USB cable
- Country of origin : Taiwan
Describe what actually happens in your application. Distinguish the initial download from later use, document any caching behavior you implement, and disclose telemetry or third-party requests. Say that audio never leaves the device only after verifying the complete application path, including upload handling, telemetry, crash reports, and external resources. “Inference runs locally” is a narrower and more defensible claim when only the model execution path has been established.
Validate the product claims before launch
Model metrics from a named dataset help characterize a checkpoint, but they do not answer whether a particular browser application will be fast, fit in memory, accept a given file, or produce a useful transcription for every recording. Test and document those product-specific behaviors rather than borrowing claims from model papers or command-line projects.
Quick Recap
- State whether the supported task is piano transcription or a broader instrument mix, and validate any broader claim independently.
- Test decoding, preprocessing, and export across the input formats and recording lengths you accept.
- Check MIDI timing, note durations, velocity, and pedal-related conventions in representative outputs.
- Measure first-use download cost, inference time, and memory on target devices.
- Verify which network requests occur during download and processing, and whether processing can continue offline after assets are available.
Product prices and availability are accurate as of the date/time indicated and are subject to change. Any price and availability information displayed on Amazon at the time of purchase will apply.
