| --- |
| license: mit |
| library_name: coreml |
| pipeline_tag: audio-to-audio |
| tags: |
| - coreml |
| - core-ml |
| - ios |
| - macos |
| - apple |
| - on-device |
| - voice-conversion |
| - voice-cloning |
| - arxiv:2312.01479 |
| --- |
| |
| # OpenVoice V2 — Core ML |
|
|
| *Voice Cloning* |
|
|
| Zero-shot voice conversion. Clone a speaker from ~10s reference audio. |
|
|
| <p><img src="https://huggingface.co/mlboydaisuke/OpenVoice-V2-CoreML/resolve/main/media/2dfefeb539.gif" alt="OpenVoice V2 demo"></p> |
|
|
| Core ML conversion of [myshell-ai/OpenVoice](https://github.com/myshell-ai/OpenVoice) for on-device inference on iPhone, iPad and Mac. Converted with `coremltools`; the packages are stateless, so all sequencing and buffering lives in your Swift code. |
|
|
| | | | |
| |---|---| |
| | Task | audio to audio | |
| | Upstream | [myshell-ai/OpenVoice](https://github.com/myshell-ai/OpenVoice) | |
| | Packages | 2 | |
| | Download size | 58 MB | |
| | Minimum iOS | 17.0 | |
| | Peak RAM | ~500 MB | |
|
|
| ## Files |
|
|
| | File | Size | Compute units | SHA-256 | |
| |---|---:|---|---| |
| | `OpenVoice_SpeakerEncoder.mlpackage.zip` | 1 MB | `cpuAndGPU` | `c3f2a96aaf5ecb5c…` | |
| | `OpenVoice_VoiceConverter.mlpackage.zip` | 57 MB | `cpuAndGPU` | `ef3ce8a2d1564aef…` | |
| | **Total** | **58 MB** | | | |
|
|
| `compute_units` is not a suggestion -- it is the configuration the conversion was verified against. Moving a package to a different compute unit can silently change the numerics (FP16 attention overflow) or crash on the GPU. |
|
|
| ## Download |
|
|
| ```bash |
| hf download mlboydaisuke/coreml-zoo --include "openvoice/*" --local-dir ./openvoice |
| unzip './openvoice/openvoice/*.zip' -d ./openvoice |
| ``` |
|
|
| ## Use in Swift |
|
|
| ```swift |
| import CoreML |
| |
| let config = MLModelConfiguration() |
| config.computeUnits = .cpuAndGPU // as converted — see the table above |
| |
| // Unzip the .mlpackage, drop it into your Xcode target and Xcode compiles it |
| // at build time: |
| let model = try OpenVoice_SpeakerEncoder(configuration: config) |
| |
| // ...or compile a downloaded .mlpackage at runtime: |
| let compiled = try await MLModel.compileModel(at: mlpackageURL) |
| let model = try MLModel(contentsOf: compiled, configuration: config) |
| ``` |
|
|
| > This model is split into 2 Core ML packages that are driven in sequence from Swift. Load them one at a time, copy the outputs out of the `MLMultiArray` buffers and release each model before loading the next — two large Core ML models resident at once will OOM on an iPhone. |
|
|
| ## Demo |
|
|
| - **Sample app** — [`sample_apps/OpenVoiceDemo`](https://github.com/john-rocky/CoreML-Models/tree/master/sample_apps/OpenVoiceDemo), a standalone SwiftUI project. |
| - **Models Zoo** — this model is downloadable and runnable inside the [Models Zoo app](https://apps.apple.com/app/id6762083207) on the App Store, no build required. |
|
|
| ## Conversion |
|
|
| - Script: [`convert_openvoice.py`](https://github.com/john-rocky/CoreML-Models/blob/master/conversion_scripts/convert_openvoice.py) |
| - Pitfalls hit during conversion (FP16 overflow, ANE buffer limits, stride handling): [`docs/coreml_conversion_notes.md`](https://github.com/john-rocky/CoreML-Models/blob/master/docs/coreml_conversion_notes.md) |
| - Model index: [CoreML-Models](https://github.com/john-rocky/CoreML-Models) |
|
|
| ## License |
|
|
| The conversion inherits the upstream license: **MIT**. |
|
|
| ## Credits |
|
|
| - Upstream authors: [myshell-ai/OpenVoice](https://github.com/myshell-ai/OpenVoice), 2023 |
| - Core ML conversion: john-rocky (Daisuke Majima) |
|
|
