konverter/Sources/Konverter/Convert/AVConverter.swift

import AVFoundation
import CoreGraphics
import Foundation
import ImageIO

/// Audio and video, mostly through AVFoundation. Formats Apple cannot write
/// (MKV, WebM, AVI, WMV, MP3, FLAC, OGG, Opus, WMA) are delegated to ffmpeg
/// when it is installed — see `FFmpeg`.
enum AVConverter {

    // MARK: - Probing

    static func duration(_ url: URL) -> Double? {
        let asset = AVURLAsset(url: url, options: [AVURLAssetPreferPreciseDurationAndTimingKey: true])
        let seconds = CMTimeGetSeconds(asset.duration)
        return seconds.isFinite && seconds > 0 ? seconds : nil
    }

    /// Uses the synchronously loaded track list; called only while building the
    /// drag wheel, so it must not await.
    static func trackSummary(_ url: URL) -> (video: Bool, audio: Bool) {
        let asset = AVURLAsset(url: url)
        var video = false
        var audio = false
        for track in asset.tracks {
            switch track.mediaType {
            case .video where track.naturalSize.height > 0: video = true
            case .audio: audio = true
            default: break
            }
        }
        return (video, audio)
    }

    // MARK: - Conversion

    static func convert(
        source: URL,
        to format: Format,
        at destination: URL,
        quality: String = "Medium",
        bitrate: String = "128",
        cancellation: Cancellation? = nil,
        progress: ((Double) -> Void)? = nil
    ) throws {
        try cancellation?.check()

        switch format {
        case .gif:
            try makeGIF(source: source, at: destination, fps: 12, maxWidth: 480, cancellation: cancellation, progress: progress)
            return
        case .mp4, .mov, .m4v:
            try exportWithSession(
                source: source,
                destination: destination,
                fileType: format == .mov ? .mov : .mp4,
                preset: videoPreset(for: quality),
                cancellation: cancellation,
                progress: progress
            )
            return
        case .m4a, .mp3:
            if format == .m4a {
                try exportWithSession(
                    source: source,
                    destination: destination,
                    fileType: .m4a,
                    preset: AVAssetExportPresetAppleM4A,
                    cancellation: cancellation,
                    progress: progress
                )
                return
            }
            try FFmpeg.convert(source: source, to: format, at: destination, bitrate: bitrate, cancellation: cancellation, progress: progress)
            return
        default:
            // Everything else: try ffmpeg, then CoreAudio, then AVFoundation.
            if Toolchain.hasFFmpeg, Settings.shared.useExternalTools {
                try FFmpeg.convert(source: source, to: format, at: destination, bitrate: bitrate, cancellation: cancellation, progress: progress)
                return
            }
            if format == .wav || format == .aiff {
                try CoreAudio.convert(source: source, to: format, at: destination, cancellation: cancellation)
                return
            }
            throw ConversionError.missingTool(
                "\(format.displayName) needs ffmpeg. Install it with: brew install ffmpeg"
            )
        }
    }

    static func videoPreset(for quality: String) -> String {
        switch quality {
        case "Low": return AVAssetExportPreset960x540
        case "High": return AVAssetExportPreset1920x1080
        default: return AVAssetExportPreset1280x720
        }
    }

    // MARK: - Tools

    static func trim(
        source: URL,
        at destination: URL,
        start: Double,
        end: Double?,
        cancellation: Cancellation? = nil,
        progress: ((Double) -> Void)? = nil
    ) throws {
        let asset = AVURLAsset(url: source)
        guard let session = AVAssetExportSession(asset: asset, presetName: AVAssetExportPresetHighestQuality) else {
            throw ConversionError.failed("Could not prepare a trim for \(source.lastPathComponent).")
        }
        let duration = CMTimeGetSeconds(asset.duration)
        let startTime = max(0, min(start, duration))
        let endTime = min(end ?? duration, duration)
        guard endTime > startTime else {
            throw ConversionError.failed("The trim range is empty.")
        }
        session.timeRange = CMTimeRange(
            start: CMTime(seconds: startTime, preferredTimescale: 600),
            end: CMTime(seconds: endTime, preferredTimescale: 600)
        )
        try run(session: session, destination: destination, fileType: destination.pathExtension == "mov" ? .mov : .mp4, cancellation: cancellation, progress: progress)
    }

    /// Rebuilds the file with the audio track dropped.
    static func mute(
        source: URL,
        at destination: URL,
        cancellation: Cancellation? = nil,
        progress: ((Double) -> Void)? = nil
    ) throws {
        let asset = AVURLAsset(url: source)
        let composition = AVMutableComposition()
        var inserted = false
        for track in asset.tracks where track.mediaType == .video {
            guard let target = composition.addMutableTrack(withMediaType: .video, preferredTrackID: kCMPersistentTrackID_Invalid) else { continue }
            try target.insertTimeRange(CMTimeRange(start: .zero, duration: asset.duration), of: track, at: .zero)
            target.preferredTransform = track.preferredTransform
            inserted = true
        }
        guard inserted else { throw ConversionError.failed("No video track to keep.") }
        try exportComposition(composition, at: destination, cancellation: cancellation, progress: progress)
    }

    static func changeSpeed(
        source: URL,
        at destination: URL,
        rate: Double,
        cancellation: Cancellation? = nil,
        progress: ((Double) -> Void)? = nil
    ) throws {
        guard rate > 0 else { throw ConversionError.failed("Speed must be greater than zero.") }
        let asset = AVURLAsset(url: source)
        let composition = AVMutableComposition()
        let duration = asset.duration
        for track in asset.tracks {
            guard track.mediaType == .video || track.mediaType == .audio else { continue }
            guard let target = composition.addMutableTrack(withMediaType: track.mediaType, preferredTrackID: kCMPersistentTrackID_Invalid) else { continue }
            try target.insertTimeRange(CMTimeRange(start: .zero, duration: duration), of: track, at: .zero)
            if track.mediaType == .video {
                target.preferredTransform = track.preferredTransform
            } else {
                target.scaleTimeRange(CMTimeRange(start: .zero, duration: duration), toDuration: CMTimeMultiplyByFloat64(duration, multiplier: 1 / rate))
            }
        }
        for track in composition.tracks where track.mediaType == .video {
            track.scaleTimeRange(CMTimeRange(start: .zero, duration: duration), toDuration: CMTimeMultiplyByFloat64(duration, multiplier: 1 / rate))
        }
        try exportComposition(composition, at: destination, cancellation: cancellation, progress: progress, audioPitch: .varispeed)
    }

    static func resize(
        source: URL,
        at destination: URL,
        width: Int,
        cancellation: Cancellation? = nil,
        progress: ((Double) -> Void)? = nil
    ) throws {
        let asset = AVURLAsset(url: source)
        guard let track = asset.tracks(withMediaType: .video).first else {
            throw ConversionError.failed("No video track to resize.")
        }
        let natural = track.naturalSize.applying(track.preferredTransform)
        let sourceSize = CGSize(width: abs(natural.width), height: abs(natural.height))
        guard sourceSize.width > 0 else { throw ConversionError.failed("Unknown video size.") }
        let scale = CGFloat(width) / sourceSize.width
        let renderSize = CGSize(width: CGFloat(width).rounded(), height: (sourceSize.height * scale).rounded())

        let composition = AVMutableComposition()
        guard let videoTrack = composition.addMutableTrack(withMediaType: .video, preferredTrackID: kCMPersistentTrackID_Invalid) else {
            throw ConversionError.failed("Could not build the resize composition.")
        }
        let range = CMTimeRange(start: .zero, duration: asset.duration)
        try videoTrack.insertTimeRange(range, of: track, at: .zero)
        if let audio = asset.tracks(withMediaType: .audio).first,
           let audioTrack = composition.addMutableTrack(withMediaType: .audio, preferredTrackID: kCMPersistentTrackID_Invalid) {
            try audioTrack.insertTimeRange(range, of: audio, at: .zero)
        }

        let instruction = AVMutableVideoCompositionInstruction()
        instruction.timeRange = range
        let layer = AVMutableVideoCompositionLayerInstruction(assetTrack: videoTrack)
        let transform = track.preferredTransform.concatenating(CGAffineTransform(scaleX: scale, y: scale))
        layer.setTransform(transform, at: .zero)
        instruction.layerInstructions = [layer]

        let videoComposition = AVMutableVideoComposition()
        videoComposition.renderSize = renderSize
        videoComposition.frameDuration = CMTime(value: 1, timescale: 30)
        videoComposition.instructions = [instruction]

        try exportComposition(composition, at: destination, cancellation: cancellation, progress: progress, videoComposition: videoComposition)
    }

    static func snapshot(
        source: URL,
        at destination: URL,
        time: Double,
        cancellation: Cancellation? = nil
    ) throws {
        let asset = AVURLAsset(url: source)
        let generator = AVAssetImageGenerator(asset: asset)
        generator.appliesPreferredTrackTransform = true
        generator.requestedTimeToleranceBefore = .zero
        generator.requestedTimeToleranceAfter = .zero
        try cancellation?.check()
        let image = try generator.copyCGImage(at: CMTime(seconds: time, preferredTimescale: 600), actualTime: nil)
        try Raster.write(image, to: destination, format: .png, quality: 1, metadata: nil)
    }

    /// Frame-by-frame GIF. Long clips are capped so the result stays small.
    static func makeGIF(
        source: URL,
        at destination: URL,
        fps: Int,
        maxWidth: Int,
        maxDuration: Double = 20,
        cancellation: Cancellation? = nil,
        progress: ((Double) -> Void)? = nil
    ) throws {
        let asset = AVURLAsset(url: source)
        let duration = min(CMTimeGetSeconds(asset.duration), maxDuration)
        guard duration > 0 else { throw ConversionError.failed("Nothing to animate.") }

        let generator = AVAssetImageGenerator(asset: asset)
        generator.appliesPreferredTrackTransform = true
        generator.maximumSize = CGSize(width: maxWidth, height: maxWidth)
        generator.requestedTimeToleranceBefore = CMTime(seconds: 0.05, preferredTimescale: 600)
        generator.requestedTimeToleranceAfter = CMTime(seconds: 0.05, preferredTimescale: 600)

        guard let output = CGImageDestinationCreateWithURL(destination as CFURL, "com.compuserve.gif" as CFString, 0, nil) else {
            throw ConversionError.failed("Could not create \(destination.lastPathComponent).")
        }
        CGImageDestinationSetProperties(output, [
            kCGImagePropertyGIFDictionary: [kCGImagePropertyGIFLoopCount: 0]
        ] as CFDictionary)

        let frameCount = max(1, Int(duration * Double(fps)))
        let frameDuration = 1.0 / Double(fps)
        for index in 0..<frameCount {
            try cancellation?.check()
            let time = CMTime(seconds: Double(index) * frameDuration, preferredTimescale: 600)
            if let image = try? generator.copyCGImage(at: time, actualTime: nil) {
                CGImageDestinationAddImage(output, image, [
                    kCGImagePropertyGIFDictionary: [kCGImagePropertyGIFDelayTime: frameDuration]
                ] as CFDictionary)
            }
            progress?(Double(index + 1) / Double(frameCount))
        }
        guard CGImageDestinationFinalize(output) else {
            throw ConversionError.failed("Writing the GIF failed.")
        }
    }

    // MARK: - Internals

    private static func exportComposition(
        _ composition: AVComposition,
        at destination: URL,
        cancellation: Cancellation?,
        progress: ((Double) -> Void)?,
        audioPitch: AVAudioTimePitchAlgorithm? = nil,
        videoComposition: AVVideoComposition? = nil
    ) throws {
        let fileType: AVFileType = destination.pathExtension == "mov"
            ? .mov
            : (destination.pathExtension == "m4a" ? .m4a : .mp4)
        guard let session = AVAssetExportSession(asset: composition, presetName: AVAssetExportPresetHighestQuality) else {
            throw ConversionError.failed("Could not prepare the export.")
        }
        if let audioPitch { session.audioTimePitchAlgorithm = audioPitch }
        if let videoComposition { session.videoComposition = videoComposition }
        try run(session: session, destination: destination, fileType: fileType, cancellation: cancellation, progress: progress)
    }

    private static func exportWithSession(
        source: URL,
        destination: URL,
        fileType: AVFileType,
        preset: String,
        cancellation: Cancellation?,
        progress: ((Double) -> Void)?
    ) throws {
        let asset = AVURLAsset(url: source)
        guard let session = AVAssetExportSession(asset: asset, presetName: preset) else {
            throw ConversionError.failed("This file cannot be exported with the \(preset) preset.")
        }
        try run(session: session, destination: destination, fileType: fileType, cancellation: cancellation, progress: progress)
    }

    private final class ErrorBox: @unchecked Sendable {
        private let lock = NSLock()
        private var value: Error?
        func set(_ error: Error) { lock.lock(); value = error; lock.unlock() }
        func get() -> Error? { lock.lock(); defer { lock.unlock() }; return value }
    }

    private static func run(
        session: AVAssetExportSession,
        destination: URL,
        fileType: AVFileType,
        cancellation: Cancellation?,
        progress: ((Double) -> Void)?
    ) throws {
        if FileManager.default.fileExists(atPath: destination.path) {
            try? FileManager.default.removeItem(at: destination)
        }

        let box = ErrorBox()
        let ready = DispatchSemaphore(value: 0)

        if #available(macOS 15.0, *) {
            let session = session
            Task.detached(priority: .userInitiated) {
                do {
                    try await session.export(to: destination, as: fileType)
                } catch {
                    box.set(error)
                }
                ready.signal()
            }
        } else {
            session.outputURL = destination
            session.outputFileType = fileType
            session.shouldOptimizeForNetworkUse = false
            session.exportAsynchronously { ready.signal() }
        }

        // Poll for progress and cancellation while the export runs.
        while ready.wait(timeout: .now() + 0.2) == .timedOut {
            if cancellation?.isCancelled == true {
                session.cancelExport()
                throw CancellationError()
            }
            progress?(Double(session.progress))
        }

        try cancellation?.check()
        if let error = box.get() { throw ConversionError.failed(error.localizedDescription) }
        if session.status == .failed {
            throw ConversionError.failed(session.error?.localizedDescription ?? "Export failed.")
        }
        if session.status == .cancelled { throw CancellationError() }
        progress?(1)
    }
}