| --- |
| license: apache-2.0 |
| library_name: coreml |
| pipeline_tag: text-to-speech |
| base_model: hexgrad/Kokoro-82M |
| base_model_relation: quantized |
| tags: |
| - coreml |
| - core-ml |
| - ios |
| - macos |
| - apple |
| - on-device |
| - tts |
| - styletts2 |
| - istftnet |
| - arxiv:2306.07691 |
| --- |
| |
| # Kokoro-82M — Core ML |
|
|
| *Multilingual TTS* |
|
|
| English + Japanese text-to-speech. 24 kHz. StyleTTS2 + iSTFTNet vocoder. Multiple voices. |
|
|
| <p><img src="https://huggingface.co/mlboydaisuke/Kokoro-82M-CoreML/resolve/main/media/2ff88dfdc5.gif" alt="Kokoro-82M demo"></p> |
|
|
| Core ML conversion of [hexgrad/Kokoro-82M](https://huggingface.co/hexgrad/Kokoro-82M) 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 | text to speech | |
| | Upstream | [hexgrad/Kokoro-82M](https://huggingface.co/hexgrad/Kokoro-82M) | |
| | Packages | 4 | |
| | Download size | 724 MB | |
| | Minimum iOS | 17.0 | |
| | Peak RAM | ~1000 MB | |
|
|
| ## Files |
|
|
| | File | Size | Compute units | SHA-256 | |
| |---|---:|---|---| |
| | `Kokoro_Predictor.mlpackage.zip` | 69 MB | `cpuAndGPU` | `af1d55dc842980c3…` | |
| | `Kokoro_Decoder_128.mlpackage.zip` | 219 MB | `cpuAndGPU` | `cece0d072f5ba6aa…` | |
| | `Kokoro_Decoder_256.mlpackage.zip` | 219 MB | `cpuAndGPU` | `36d5e16d5c5ccb50…` | |
| | `Kokoro_Decoder_512.mlpackage.zip` | 219 MB | `cpuAndGPU` | `0a44484c327e4fe8…` | |
| | `kokoro_vocab.json` | 1 KB | `-` | `70abefbe8a1c8865…` | |
| | **Total** | **724 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 "kokoro/*" --local-dir ./kokoro |
| unzip './kokoro/kokoro/*.zip' -d ./kokoro |
| ``` |
|
|
| ## 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 Kokoro_Predictor(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 4 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/KokoroDemo`](https://github.com/john-rocky/CoreML-Models/tree/master/sample_apps/KokoroDemo), 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_kokoro.py`](https://github.com/john-rocky/CoreML-Models/blob/master/conversion_scripts/convert_kokoro.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: **Apache-2.0**. |
|
|
| ## Credits |
|
|
| - Upstream authors: [hexgrad/Kokoro-82M](https://huggingface.co/hexgrad/Kokoro-82M), 2024 |
| - Core ML conversion: john-rocky (Daisuke Majima) |
|
|
