| 1 | import Foundation |
| 2 | |
| 3 | /// One source the models can separate out of a mix. |
| 4 | public enum StemKind: String, Sendable, CaseIterable { |
| 5 | case drums |
| 6 | case bass |
| 7 | case other |
| 8 | case vocals |
| 9 | case guitar |
| 10 | case piano |
| 11 | } |
| 12 | |
| 13 | /// Which HTDemucs variant to download or load. |
| 14 | public enum DemucsModel: Sendable, Hashable { |
| 15 | /// htdemucs: drums, bass, other, vocals. ~84 MB. |
| 16 | case fourStem |
| 17 | |
| 18 | /// htdemucs_6s: adds guitar and piano. ~84 MB. |
| 19 | case sixStem |
| 20 | |
| 21 | /// htdemucs_ft: one fine-tuned sub-model per stem, best quality. ~333 MB. |
| 22 | /// |
| 23 | /// Only drums, bass, other and vocals are available. Each requested stem |
| 24 | /// runs its own sub-model, so asking for fewer is proportionally faster. |
| 25 | /// Every selection shares the same downloaded weights. |
| 26 | case fineTuned(stems: [StemKind]) |
| 27 | |
| 28 | /// The fine-tuned model with all four of its stems. |
| 29 | public static let fineTunedAll = DemucsModel.fineTuned(stems: [.drums, .bass, .other, .vocals]) |
| 30 | } |
| 31 | |
| 32 | public struct DemucsModelInfo: Sendable, Hashable, Identifiable { |
| 33 | public let model: DemucsModel |
| 34 | /// The model's upstream name, such as `htdemucs`. |
| 35 | public let id: String |
| 36 | public let label: String |
| 37 | public let description: String |
| 38 | public let sizeMB: UInt32 |
| 39 | public let stems: [StemKind] |
| 40 | } |
| 41 | |
| 42 | /// A separated source at the input sample rate, the same length as the input. |
| 43 | public struct Stem: Sendable, Equatable { |
| 44 | public let kind: StemKind |
| 45 | public let left: [Float] |
| 46 | public let right: [Float] |
| 47 | } |
| 48 | |
| 49 | public struct DownloadProgress: Sendable, Equatable { |
| 50 | public let downloadedBytes: UInt64 |
| 51 | /// From the server's `Content-Length`, or the model's nominal size. |
| 52 | public let totalBytes: UInt64 |
| 53 | public let fraction: Double |
| 54 | } |
| 55 | |
| 56 | public struct SeparationProgress: Sendable, Equatable { |
| 57 | public let fraction: Double |
| 58 | public let chunkIndex: UInt64 |
| 59 | public let totalChunks: UInt64 |
| 60 | } |
| 61 | |
| 62 | public enum DemucsError: Error, Sendable, Equatable, LocalizedError { |
| 63 | /// The task was cancelled, or a progress handler asked to stop. |
| 64 | case cancelled |
| 65 | /// `separate` or `warmup` was called before `load`. |
| 66 | case notLoaded |
| 67 | /// `load` was called before the model's weights were downloaded. |
| 68 | case notDownloaded(modelID: String) |
| 69 | case invalidInput(String) |
| 70 | case downloadFailed(String) |
| 71 | case io(String) |
| 72 | case inferenceFailed(String) |
| 73 | |
| 74 | public var errorDescription: String? { |
| 75 | switch self { |
| 76 | case .cancelled: "The operation was cancelled." |
| 77 | case .notLoaded: "No model is loaded." |
| 78 | case .notDownloaded(let id): "Model \(id) has not been downloaded." |
| 79 | case .invalidInput(let reason): reason |
| 80 | case .downloadFailed(let reason): "Download failed: \(reason)" |
| 81 | case .io(let reason): "File error: \(reason)" |
| 82 | case .inferenceFailed(let reason): reason |
| 83 | } |
| 84 | } |
| 85 | } |
| 86 | |
| 87 | /// On-device music source separation (HTDemucs v4), GPU-accelerated through |
| 88 | /// Metal. |
| 89 | /// |
| 90 | /// An engine holds at most one loaded model. Loading, warming up and |
| 91 | /// separating on one engine run one at a time; create another engine to run |
| 92 | /// a second model side by side, memory permitting. |
| 93 | /// |
| 94 | /// ```swift |
| 95 | /// try await DemucsEngine.download(.fourStem) |
| 96 | /// let engine = DemucsEngine() |
| 97 | /// try await engine.load(.fourStem) |
| 98 | /// let stems = try await engine.separate(left: left, right: right, sampleRate: 44_100) |
| 99 | /// ``` |
| 100 | public actor DemucsEngine { |
| 101 | private nonisolated let core: FfiDemucsEngine |
| 102 | |
| 103 | public init() { |
| 104 | core = FfiDemucsEngine() |
| 105 | } |
| 106 | |
| 107 | // MARK: Model files |
| 108 | |
| 109 | /// Sets where model weights are stored. Call once at launch, before any |
| 110 | /// other Demucs call; sandboxed apps should pass a directory inside their |
| 111 | /// container, such as one under Caches. Without it, weights go to the |
| 112 | /// platform cache directory under `demucs-rs/`. |
| 113 | public static func configureCacheDirectory(_ url: URL) { |
| 114 | configureCacheDir(path: url.path) |
| 115 | } |
| 116 | |
| 117 | public static var models: [DemucsModelInfo] { |
| 118 | allModels().map(\.swift) |
| 119 | } |
| 120 | |
| 121 | public static func info(for model: DemucsModel) -> DemucsModelInfo { |
| 122 | modelMetadata(model: model.ffi).swift |
| 123 | } |
| 124 | |
| 125 | public static func isDownloaded(_ model: DemucsModel) -> Bool { |
| 126 | isModelCached(model: model.ffi) |
| 127 | } |
| 128 | |
| 129 | /// The downloaded weights file, or `nil` if it is not on disk. |
| 130 | public static func downloadedFileURL(for model: DemucsModel) -> URL? { |
| 131 | cachedModelPath(model: model.ffi).map { URL(fileURLWithPath: $0) } |
| 132 | } |
| 133 | |
| 134 | /// Removes the downloaded weights. Returns `false` if there was nothing |
| 135 | /// to remove. |
| 136 | @discardableResult |
| 137 | public static func deleteDownload(_ model: DemucsModel) -> Bool { |
| 138 | deleteCachedModel(model: model.ffi) |
| 139 | } |
| 140 | |
| 141 | /// Downloads the model's weights from Hugging Face. Returns straight away |
| 142 | /// if they are already downloaded. Cancel the calling task to stop; a |
| 143 | /// partial download is discarded. |
| 144 | /// |
| 145 | /// `progress` is called on a background thread. |
| 146 | public static func download( |
| 147 | _ model: DemucsModel, |
| 148 | progress: (@Sendable (DownloadProgress) -> Void)? = nil |
| 149 | ) async throws { |
| 150 | let listener = DownloadListener(handler: progress) |
| 151 | try await withTaskCancellationHandler { |
| 152 | try await translatingErrors { |
| 153 | try await downloadModel(model: model.ffi, listener: listener) |
| 154 | } |
| 155 | } onCancel: { |
| 156 | listener.cancel() |
| 157 | } |
| 158 | } |
| 159 | |
| 160 | // MARK: Loading |
| 161 | |
| 162 | /// Loads a downloaded model, replacing the current one. Returns the load |
| 163 | /// time in seconds. |
| 164 | @discardableResult |
| 165 | public func load(_ model: DemucsModel) async throws -> Double { |
| 166 | try await translatingErrors { |
| 167 | try await core.loadModel(model: model.ffi) |
| 168 | } |
| 169 | } |
| 170 | |
| 171 | /// Runs a dummy separation so the first real one does not pay for GPU |
| 172 | /// shader compilation. Worth calling once after `load`, off the critical |
| 173 | /// path. |
| 174 | public func warmup() async throws { |
| 175 | try await translatingErrors { |
| 176 | try await core.warmup() |
| 177 | } |
| 178 | } |
| 179 | |
| 180 | /// Frees the loaded model. Returns `false` if nothing was loaded. |
| 181 | @discardableResult |
| 182 | public func unload() async -> Bool { |
| 183 | await core.unloadModel() |
| 184 | } |
| 185 | |
| 186 | public nonisolated var isLoaded: Bool { |
| 187 | core.isLoaded() |
| 188 | } |
| 189 | |
| 190 | public nonisolated var loadedModel: DemucsModel? { |
| 191 | core.loadedModel()?.swift |
| 192 | } |
| 193 | |
| 194 | // MARK: Separation |
| 195 | |
| 196 | /// Separates stereo PCM at any sample rate into the loaded model's stems. |
| 197 | /// |
| 198 | /// Both channels must have the same, non-zero length. Audio is processed |
| 199 | /// in chunks of about 7.8 seconds; cancelling the calling task stops at |
| 200 | /// the next chunk boundary. `progress` is called on a background thread. |
| 201 | public func separate( |
| 202 | left: [Float], |
| 203 | right: [Float], |
| 204 | sampleRate: UInt32, |
| 205 | progress: (@Sendable (SeparationProgress) -> Void)? = nil |
| 206 | ) async throws -> [Stem] { |
| 207 | let listener = SeparationListener(handler: progress) |
| 208 | let stems = try await withTaskCancellationHandler { |
| 209 | try await translatingErrors { |
| 210 | try await core.separate( |
| 211 | left: left, |
| 212 | right: right, |
| 213 | sampleRate: sampleRate, |
| 214 | listener: listener |
| 215 | ) |
| 216 | } |
| 217 | } onCancel: { |
| 218 | listener.cancel() |
| 219 | } |
| 220 | return stems.map(\.swift) |
| 221 | } |
| 222 | |
| 223 | /// Separates mono PCM. The stems come back stereo with identical channels. |
| 224 | public func separate( |
| 225 | mono samples: [Float], |
| 226 | sampleRate: UInt32, |
| 227 | progress: (@Sendable (SeparationProgress) -> Void)? = nil |
| 228 | ) async throws -> [Stem] { |
| 229 | try await separate(left: samples, right: samples, sampleRate: sampleRate, progress: progress) |
| 230 | } |
| 231 | |
| 232 | /// How many chunks `separate` splits this many samples into, for sizing |
| 233 | /// progress UI up front. |
| 234 | public static func chunkCount(sampleCount: Int, sampleRate: UInt32) -> Int { |
| 235 | Int(numChunks(nSamples: UInt64(max(sampleCount, 0)), sampleRate: sampleRate)) |
| 236 | } |
| 237 | } |
| 238 | |
| 239 | // MARK: - Callback adapters |
| 240 | |
| 241 | /// Shared by both listeners: the progress callback returns `false` once the |
| 242 | /// Swift task is cancelled, which is how Rust learns to stop. |
| 243 | private final class CancellationFlag: @unchecked Sendable { |
| 244 | private let lock = NSLock() |
| 245 | private var cancelled = false |
| 246 | |
| 247 | var isCancelled: Bool { lock.withLock { cancelled } } |
| 248 | |
| 249 | func cancel() { lock.withLock { cancelled = true } } |
| 250 | } |
| 251 | |
| 252 | private final class DownloadListener: FfiDownloadListener, @unchecked Sendable { |
| 253 | private let handler: (@Sendable (DownloadProgress) -> Void)? |
| 254 | private let flag = CancellationFlag() |
| 255 | |
| 256 | init(handler: (@Sendable (DownloadProgress) -> Void)?) { |
| 257 | self.handler = handler |
| 258 | } |
| 259 | |
| 260 | func cancel() { flag.cancel() } |
| 261 | |
| 262 | func onProgress(progress: FfiDownloadProgress) -> Bool { |
| 263 | guard !flag.isCancelled else { return false } |
| 264 | handler?(DownloadProgress( |
| 265 | downloadedBytes: progress.downloadedBytes, |
| 266 | totalBytes: progress.totalBytes, |
| 267 | fraction: progress.fraction |
| 268 | )) |
| 269 | return true |
| 270 | } |
| 271 | } |
| 272 | |
| 273 | private final class SeparationListener: FfiSeparationListener, @unchecked Sendable { |
| 274 | private let handler: (@Sendable (SeparationProgress) -> Void)? |
| 275 | private let flag = CancellationFlag() |
| 276 | |
| 277 | init(handler: (@Sendable (SeparationProgress) -> Void)?) { |
| 278 | self.handler = handler |
| 279 | } |
| 280 | |
| 281 | func cancel() { flag.cancel() } |
| 282 | |
| 283 | func onProgress(progress: FfiSeparationProgress) -> Bool { |
| 284 | guard !flag.isCancelled else { return false } |
| 285 | handler?(SeparationProgress( |
| 286 | fraction: progress.fraction, |
| 287 | chunkIndex: progress.chunkIndex, |
| 288 | totalChunks: progress.totalChunks |
| 289 | )) |
| 290 | return true |
| 291 | } |
| 292 | } |
| 293 | |
| 294 | // MARK: - Conversions |
| 295 | |
| 296 | private func translatingErrors<T>(_ body: () async throws -> T) async throws -> T { |
| 297 | do { |
| 298 | return try await body() |
| 299 | } catch let error as FfiDemucsError { |
| 300 | throw DemucsError(error) |
| 301 | } |
| 302 | } |
| 303 | |
| 304 | extension DemucsError { |
| 305 | init(_ error: FfiDemucsError) { |
| 306 | switch error { |
| 307 | case .Cancelled: self = .cancelled |
| 308 | case .NotLoaded: self = .notLoaded |
| 309 | case .NotCached(let modelId): self = .notDownloaded(modelID: modelId) |
| 310 | case .InvalidInput(let reason): self = .invalidInput(reason) |
| 311 | case .Download(let reason): self = .downloadFailed(reason) |
| 312 | case .Io(let reason): self = .io(reason) |
| 313 | case .Inference(let reason): self = .inferenceFailed(reason) |
| 314 | } |
| 315 | } |
| 316 | } |
| 317 | |
| 318 | extension StemKind { |
| 319 | var ffi: FfiStemKind { |
| 320 | switch self { |
| 321 | case .drums: .drums |
| 322 | case .bass: .bass |
| 323 | case .other: .other |
| 324 | case .vocals: .vocals |
| 325 | case .guitar: .guitar |
| 326 | case .piano: .piano |
| 327 | } |
| 328 | } |
| 329 | } |
| 330 | |
| 331 | extension FfiStemKind { |
| 332 | var swift: StemKind { |
| 333 | switch self { |
| 334 | case .drums: .drums |
| 335 | case .bass: .bass |
| 336 | case .other: .other |
| 337 | case .vocals: .vocals |
| 338 | case .guitar: .guitar |
| 339 | case .piano: .piano |
| 340 | } |
| 341 | } |
| 342 | } |
| 343 | |
| 344 | extension DemucsModel { |
| 345 | var ffi: FfiDemucsModel { |
| 346 | switch self { |
| 347 | case .fourStem: .fourStem |
| 348 | case .sixStem: .sixStem |
| 349 | case .fineTuned(let stems): .fineTuned(stems: stems.map(\.ffi)) |
| 350 | } |
| 351 | } |
| 352 | } |
| 353 | |
| 354 | extension FfiDemucsModel { |
| 355 | var swift: DemucsModel { |
| 356 | switch self { |
| 357 | case .fourStem: .fourStem |
| 358 | case .sixStem: .sixStem |
| 359 | case .fineTuned(let stems): .fineTuned(stems: stems.map(\.swift)) |
| 360 | } |
| 361 | } |
| 362 | } |
| 363 | |
| 364 | private extension FfiModelMetadata { |
| 365 | var swift: DemucsModelInfo { |
| 366 | DemucsModelInfo( |
| 367 | model: model.swift, |
| 368 | id: id, |
| 369 | label: label, |
| 370 | description: description, |
| 371 | sizeMB: sizeMb, |
| 372 | stems: stems.map(\.swift) |
| 373 | ) |
| 374 | } |
| 375 | } |
| 376 | |
| 377 | private extension FfiStem { |
| 378 | var swift: Stem { |
| 379 | Stem(kind: kind.swift, left: left, right: right) |
| 380 | } |
| 381 | } |