Фикс: конвертация голосовых сообщений desktop - iOS

Rosetta/Core/Network/TransportManager.swift

@@ -152,6 +152,16 @@ final class TransportManager: @unchecked Sendable {
     private static let maxDownloadRetries = 3
 
     func downloadFile(tag: String, server: String? = nil) async throws -> Data {
+        try await downloadFile(tag: tag, server: server, onProgress: nil)
+    }
+
+    /// Downloads file with optional progress reporting.
+    /// Progress callback fires on MainActor with values 0.0–1.0.
+    func downloadFile(
+        tag: String,
+        server: String? = nil,
+        onProgress: (@MainActor (Double) -> Void)?
+    ) async throws -> Data {
         let serverUrl: String
         if let explicit = server, !explicit.isEmpty {
             serverUrl = explicit
@@ -172,12 +182,41 @@ final class TransportManager: @unchecked Sendable {
         var lastError: Error = TransportError.invalidResponse
         for attempt in 0..<Self.maxDownloadRetries {
             do {
+                if let onProgress {
+                    // Stream bytes for progress reporting
+                    let (asyncBytes, response) = try await session.bytes(for: request)
+
+                    guard let httpResponse = response as? HTTPURLResponse else {
+                        throw TransportError.invalidResponse
+                    }
+                    guard httpResponse.statusCode == 200 else {
+                        throw TransportError.downloadFailed(statusCode: httpResponse.statusCode)
+                    }
+
+                    let expectedLength = httpResponse.expectedContentLength
+                    var data = Data()
+                    if expectedLength > 0 { data.reserveCapacity(Int(expectedLength)) }
+                    var lastReportedKB = 0
+
+                    for try await byte in asyncBytes {
+                        data.append(byte)
+                        let currentKB = data.count / 4096
+                        if expectedLength > 0, currentKB > lastReportedKB {
+                            lastReportedKB = currentKB
+                            let fraction = Double(data.count) / Double(expectedLength)
+                            await onProgress(min(fraction, 1.0))
+                        }
+                    }
+                    await onProgress(1.0)
+                    Self.logger.info("Download complete: tag=\(tag), \(data.count) bytes")
+                    return data
+                } else {
+                    // Fast path — no progress needed
                     let (data, response) = try await session.data(for: request)
 
                     guard let httpResponse = response as? HTTPURLResponse else {
                         throw TransportError.invalidResponse
                     }
-
                     guard httpResponse.statusCode == 200 else {
                         Self.logger.error("Download failed: HTTP \(httpResponse.statusCode)")
                         throw TransportError.downloadFailed(statusCode: httpResponse.statusCode)
@@ -185,6 +224,7 @@ final class TransportManager: @unchecked Sendable {
 
                     Self.logger.info("Download complete: tag=\(tag), \(data.count) bytes")
                     return data
+                }
             } catch {
                 lastError = error
                 if attempt < Self.maxDownloadRetries - 1 {

Rosetta/Core/Services/SessionManager.swift

@@ -839,16 +839,21 @@ final class SessionManager {
     }
 
     /// Builds a data URI from attachment data (desktop: `FileReader.readAsDataURL()`).
+    /// Voice attachments use hex encoding (Desktop parity: `Buffer.toString('hex')`).
     private func buildDataURI(_ attachment: PendingAttachment) -> String {
-        let base64 = attachment.data.base64EncodedString()
         switch attachment.type {
         case .image:
-            return "data:image/jpeg;base64,\(base64)"
+            return "data:image/jpeg;base64,\(attachment.data.base64EncodedString())"
         case .file:
             let mimeType = mimeTypeForFileName(attachment.fileName ?? "file")
-            return "data:\(mimeType);base64,\(base64)"
+            return "data:\(mimeType);base64,\(attachment.data.base64EncodedString())"
+        case .voice:
+            // Desktop parity: raw bytes as hex string.
+            // Desktop DialogInput.tsx:214 sends Buffer.from(audioBlob).toString('hex'),
+            // Desktop MessageVoice.tsx:125 reads Buffer.from(blob, "hex").
+            return attachment.data.hexString
         default:
-            return "data:application/octet-stream;base64,\(base64)"
+            return "data:application/octet-stream;base64,\(attachment.data.base64EncodedString())"
         }
     }
 

Rosetta/Core/Services/WebMOpusConverter.swift

@@ -0,0 +1,709 @@
+import AudioToolbox
+import AVFoundation
+import Foundation
+import os
+
+// MARK: - WebM/Opus → M4A Converter
+
+/// Converts WebM/Opus audio (from Desktop) to M4A/AAC playable by AVAudioPlayer.
+/// iOS doesn't support WebM container natively. This parser extracts Opus frames from
+/// the EBML/WebM container, wraps them in a CAF file (Apple's native container that
+/// supports Opus), then transcodes to M4A via AVAudioFile.
+enum WebMOpusConverter {
+
+    private static let logger = Logger(subsystem: "com.rosetta.messenger", category: "WebMOpusConverter")
+
+    /// WebM/EBML magic: `0x1A 0x45 0xDF 0xA3`
+    static func isWebM(_ data: Data) -> Bool {
+        data.count >= 4 && data[0] == 0x1A && data[1] == 0x45 && data[2] == 0xDF && data[3] == 0xA3
+    }
+
+    /// Main entry: WebM/Opus Data → M4A Data (or CAF fallback).
+    /// Returns nil only if WebM parsing fails entirely.
+    static func convertToPlayable(_ webmData: Data) -> Data? {
+        guard let parsed = parseWebM(webmData) else {
+            logger.error("WebM parse failed (\(webmData.count) bytes)")
+            return nil
+        }
+        guard !parsed.frames.isEmpty else {
+            logger.warning("WebM contains 0 Opus frames")
+            return nil
+        }
+        logger.info("WebM parsed: \(parsed.frames.count) frames, \(parsed.sampleRate)Hz, \(parsed.channels)ch")
+
+        // Decode Opus → PCM via AudioConverter, then write as WAV
+        if let wavData = decodeOpusToWAV(parsed) {
+            logger.info("WebM→WAV conversion succeeded (\(wavData.count) bytes)")
+            return wavData
+        }
+
+        // Fallback: try CAF container route
+        if let cafData = buildCAF(parsed), let m4aData = transcodeCAFtoM4A(cafData) {
+            return m4aData
+        }
+
+        logger.error("All conversion paths failed")
+        return nil
+    }
+
+    // MARK: - WebM/EBML Parser
+
+    private struct ParsedWebM {
+        let frames: [Data]
+        let sampleRate: Double
+        let channels: UInt32
+        let codecPrivate: Data?
+    }
+
+    private static func parseWebM(_ data: Data) -> ParsedWebM? {
+        let bytes = [UInt8](data)
+        var pos = 0
+
+        // EBML header (0x1A45DFA3)
+        guard bytes.count >= 8,
+              bytes[0] == 0x1A, bytes[1] == 0x45, bytes[2] == 0xDF, bytes[3] == 0xA3
+        else { return nil }
+        pos = 4
+        guard let (headerSize, hLen) = readVINTValue(bytes, at: pos) else { return nil }
+        pos += hLen + Int(headerSize)
+
+        // Segment (0x18538067)
+        guard pos + 4 <= bytes.count,
+              bytes[pos] == 0x18, bytes[pos + 1] == 0x53,
+              bytes[pos + 2] == 0x80, bytes[pos + 3] == 0x67
+        else { return nil }
+        pos += 4
+        guard let (_, segLen) = readVINTValue(bytes, at: pos) else { return nil }
+        pos += segLen
+
+        var sampleRate: Double = 48000
+        var channels: UInt32 = 1
+        var codecPrivate: Data?
+        var audioTrackNum: UInt64 = 1
+        var frames: [Data] = []
+        frames.reserveCapacity(512)
+
+        while pos < bytes.count {
+            guard let (elemID, idLen) = readElementID(bytes, at: pos) else { break }
+            pos += idLen
+            guard let (elemSize, sizeLen) = readVINTValue(bytes, at: pos) else { break }
+            pos += sizeLen
+
+            let elemDataEnd = min(pos + Int(elemSize), bytes.count)
+
+            switch elemID {
+            case 0x1654AE6B: // Tracks — parse children (TrackEntry)
+                parseTracks(bytes, from: pos, to: elemDataEnd,
+                            sampleRate: &sampleRate, channels: &channels,
+                            codecPrivate: &codecPrivate, audioTrackNum: &audioTrackNum)
+                pos = elemDataEnd
+
+            case 0x1F43B675: // Cluster — extract SimpleBlocks
+                parseCluster(bytes, from: pos, to: elemDataEnd,
+                             audioTrackNum: audioTrackNum, frames: &frames)
+                pos = elemDataEnd
+
+            default:
+                pos = elemDataEnd
+            }
+
+            if pos <= 0 || elemDataEnd <= 0 { break }
+        }
+
+        return ParsedWebM(frames: frames, sampleRate: sampleRate,
+                          channels: channels, codecPrivate: codecPrivate)
+    }
+
+    private static func parseTracks(
+        _ bytes: [UInt8], from start: Int, to end: Int,
+        sampleRate: inout Double, channels: inout UInt32,
+        codecPrivate: inout Data?, audioTrackNum: inout UInt64
+    ) {
+        var pos = start
+        while pos < end {
+            guard let (id, idLen) = readElementID(bytes, at: pos) else { break }
+            pos += idLen
+            guard let (size, sizeLen) = readVINTValue(bytes, at: pos) else { break }
+            pos += sizeLen
+            let childEnd = min(pos + Int(size), end)
+
+            if id == 0xAE { // TrackEntry
+                parseTrackEntry(bytes, from: pos, to: childEnd,
+                                sampleRate: &sampleRate, channels: &channels,
+                                codecPrivate: &codecPrivate, audioTrackNum: &audioTrackNum)
+            }
+            pos = childEnd
+        }
+    }
+
+    private static func parseTrackEntry(
+        _ bytes: [UInt8], from start: Int, to end: Int,
+        sampleRate: inout Double, channels: inout UInt32,
+        codecPrivate: inout Data?, audioTrackNum: inout UInt64
+    ) {
+        var pos = start
+        var isAudio = false
+        var trackNum: UInt64 = 0
+
+        while pos < end {
+            guard let (id, idLen) = readElementID(bytes, at: pos) else { break }
+            pos += idLen
+            guard let (size, sizeLen) = readVINTValue(bytes, at: pos) else { break }
+            pos += sizeLen
+            let childEnd = min(pos + Int(size), end)
+
+            switch id {
+            case 0xD7: // TrackNumber
+                trackNum = readUInt(bytes, at: pos, count: Int(size))
+            case 0x83: // TrackType (2 = audio)
+                isAudio = readUInt(bytes, at: pos, count: Int(size)) == 2
+            case 0x63A2: // CodecPrivate (OpusHead)
+                if childEnd <= bytes.count {
+                    codecPrivate = Data(bytes[pos..<childEnd])
+                }
+            case 0xE1: // Audio sub-element
+                parseAudioElement(bytes, from: pos, to: childEnd,
+                                  sampleRate: &sampleRate, channels: &channels)
+            default:
+                break
+            }
+            pos = childEnd
+        }
+
+        if isAudio { audioTrackNum = trackNum }
+    }
+
+    private static func parseAudioElement(
+        _ bytes: [UInt8], from start: Int, to end: Int,
+        sampleRate: inout Double, channels: inout UInt32
+    ) {
+        var pos = start
+        while pos < end {
+            guard let (id, idLen) = readElementID(bytes, at: pos) else { break }
+            pos += idLen
+            guard let (size, sizeLen) = readVINTValue(bytes, at: pos) else { break }
+            pos += sizeLen
+            let childEnd = min(pos + Int(size), end)
+
+            switch id {
+            case 0xB5: // SamplingFrequency (float)
+                sampleRate = readFloat(bytes, at: pos, count: Int(size))
+            case 0x9F: // Channels
+                channels = UInt32(readUInt(bytes, at: pos, count: Int(size)))
+            default:
+                break
+            }
+            pos = childEnd
+        }
+    }
+
+    private static func parseCluster(
+        _ bytes: [UInt8], from start: Int, to end: Int,
+        audioTrackNum: UInt64, frames: inout [Data]
+    ) {
+        var pos = start
+        while pos < end {
+            guard let (id, idLen) = readElementID(bytes, at: pos) else { break }
+            pos += idLen
+            guard let (size, sizeLen) = readVINTValue(bytes, at: pos) else { break }
+            pos += sizeLen
+            let childEnd = min(pos + Int(size), end)
+
+            if id == 0xA3 { // SimpleBlock
+                guard let (blockTrack, trackLen) = readVINTValue(bytes, at: pos) else {
+                    pos = childEnd; continue
+                }
+                if blockTrack == audioTrackNum {
+                    let headerStart = pos + trackLen
+                    // 2 bytes timestamp + 1 byte flags
+                    guard headerStart + 3 <= childEnd else { pos = childEnd; continue }
+                    let flags = bytes[headerStart + 2]
+                    let lacingType = (flags >> 1) & 0x03 // bits 1-2 in Matroska spec
+                    let dataStart = headerStart + 3
+
+                    if lacingType == 0 {
+                        // No lacing — single frame
+                        if dataStart < childEnd {
+                            frames.append(Data(bytes[dataStart..<childEnd]))
+                        }
+                    } else {
+                        // Laced: first byte = number of frames - 1
+                        guard dataStart < childEnd else { pos = childEnd; continue }
+                        let numFrames = Int(bytes[dataStart]) + 1
+                        let lacedDataStart = dataStart + 1
+
+                        if lacingType == 0x02 { // Fixed-size lacing
+                            let totalData = childEnd - lacedDataStart
+                            let frameSize = totalData / numFrames
+                            for i in 0..<numFrames {
+                                let fStart = lacedDataStart + i * frameSize
+                                let fEnd = min(fStart + frameSize, childEnd)
+                                if fStart < fEnd { frames.append(Data(bytes[fStart..<fEnd])) }
+                            }
+                        } else if lacingType == 0x01 { // Xiph lacing
+                            var frameSizes = [Int]()
+                            var sizePos = lacedDataStart
+                            for _ in 0..<(numFrames - 1) {
+                                var frameSize = 0
+                                while sizePos < childEnd {
+                                    let val = Int(bytes[sizePos]); sizePos += 1
+                                    frameSize += val
+                                    if val < 255 { break }
+                                }
+                                frameSizes.append(frameSize)
+                            }
+                            // Last frame = remaining data
+                            let consumed = frameSizes.reduce(0, +)
+                            frameSizes.append(childEnd - sizePos - consumed)
+
+                            var fPos = sizePos
+                            for sz in frameSizes {
+                                let fEnd = min(fPos + sz, childEnd)
+                                if fPos < fEnd { frames.append(Data(bytes[fPos..<fEnd])) }
+                                fPos = fEnd
+                            }
+                        } else if lacingType == 0x03 { // EBML lacing
+                            var frameSizes = [Int]()
+                            var sizePos = lacedDataStart
+                            // First frame size: unsigned VINT
+                            if let (firstSize, firstLen) = readVINTValue(bytes, at: sizePos) {
+                                frameSizes.append(Int(firstSize))
+                                sizePos += firstLen
+                            }
+                            // Subsequent sizes: signed VINT deltas
+                            for _ in 1..<(numFrames - 1) {
+                                guard let (raw, len) = readVINTValue(bytes, at: sizePos) else { break }
+                                sizePos += len
+                                // Convert to signed: subtract midpoint
+                                let midpoint = (1 << (len * 7 - 1)) - 1
+                                let delta = Int(raw) - midpoint
+                                let prev = frameSizes.last ?? 0
+                                frameSizes.append(prev + delta)
+                            }
+                            // Last frame = remaining
+                            let consumed = frameSizes.reduce(0, +)
+                            frameSizes.append(childEnd - sizePos - consumed)
+
+                            var fPos = sizePos
+                            for sz in frameSizes {
+                                let fEnd = min(fPos + max(0, sz), childEnd)
+                                if fPos < fEnd { frames.append(Data(bytes[fPos..<fEnd])) }
+                                fPos = fEnd
+                            }
+                        }
+                    }
+                }
+            }
+            pos = childEnd
+        }
+    }
+
+    // MARK: - Opus → PCM → WAV (AudioToolbox AudioConverter)
+
+    /// Base frame duration in samples from Opus TOC byte (single frame, no count multiplier).
+    /// AudioConverter needs this for mFramesPerPacket (max 5760 = 120ms).
+    private static func opusBaseFrameSamples(_ packet: Data, sampleRate: Double) -> UInt32 {
+        guard !packet.isEmpty else { return 960 }
+        let config = Int((packet[0] >> 3) & 0x1F)
+        let durationMs: Int
+        switch config {
+        case 0...3:   durationMs = 10
+        case 4...7:   durationMs = 20
+        case 8...11:  durationMs = 40
+        case 12...15: durationMs = 60
+        case 16...19: durationMs = 10
+        case 20...23: durationMs = 20
+        case 24...27: durationMs = 10
+        case 28...31: durationMs = 20
+        default:      durationMs = 20
+        }
+        return UInt32(Int(sampleRate) * durationMs / 1000)
+    }
+
+    /// Total PCM samples this Opus packet will produce (base × frame count).
+    private static func opusTotalSamples(_ packet: Data, sampleRate: Double) -> UInt32 {
+        let base = opusBaseFrameSamples(packet, sampleRate: sampleRate)
+        guard !packet.isEmpty else { return base }
+        let code = packet[0] & 0x03
+        let count: UInt32
+        switch code {
+        case 0: count = 1
+        case 1, 2: count = 2
+        case 3: count = packet.count > 1 ? UInt32(packet[1] & 0x3F) : 1
+        default: count = 1
+        }
+        return base * count
+    }
+
+    private static func decodeOpusToWAV(_ parsed: ParsedWebM) -> Data? {
+        let sampleRate = parsed.sampleRate
+        let channels = parsed.channels
+
+        // Base frame size for ASBD (single sub-frame, max 5760 = 120ms)
+        let baseSamples = parsed.frames.first.map {
+            opusBaseFrameSamples($0, sampleRate: sampleRate)
+        } ?? 960
+
+        logger.info("Opus base frame: \(baseSamples) samples (\(Double(baseSamples) / sampleRate * 1000)ms)")
+
+        // Input: Opus compressed (base frame size for AudioConverter)
+        var inASBD = AudioStreamBasicDescription(
+            mSampleRate: sampleRate,
+            mFormatID: kAudioFormatOpus,
+            mFormatFlags: 0,
+            mBytesPerPacket: 0,
+            mFramesPerPacket: baseSamples,
+            mBytesPerFrame: 0,
+            mChannelsPerFrame: channels,
+            mBitsPerChannel: 0,
+            mReserved: 0
+        )
+
+        // Output: 16-bit PCM
+        let bytesPerSample: UInt32 = 2
+        var outASBD = AudioStreamBasicDescription(
+            mSampleRate: sampleRate,
+            mFormatID: kAudioFormatLinearPCM,
+            mFormatFlags: kAudioFormatFlagIsSignedInteger | kAudioFormatFlagIsPacked,
+            mBytesPerPacket: bytesPerSample * channels,
+            mFramesPerPacket: 1,
+            mBytesPerFrame: bytesPerSample * channels,
+            mChannelsPerFrame: channels,
+            mBitsPerChannel: bytesPerSample * 8,
+            mReserved: 0
+        )
+
+        var converter: AudioConverterRef?
+        var status = AudioConverterNew(&inASBD, &outASBD, &converter)
+        guard status == noErr, let converter else {
+            logger.warning("AudioConverterNew failed for Opus: \(status)")
+            return nil
+        }
+        defer { AudioConverterDispose(converter) }
+
+        // Set magic cookie (OpusHead) if available
+        if var cookie = parsed.codecPrivate.map({ [UInt8]($0) }), !cookie.isEmpty {
+            AudioConverterSetProperty(
+                converter,
+                kAudioConverterDecompressionMagicCookie,
+                UInt32(cookie.count),
+                &cookie
+            )
+        }
+
+        // Buffer sized for max multi-frame packet (6 × 60ms = 360ms = 17280 samples)
+        let maxOutputSamples: UInt32 = 17280
+        let pcmBufferSize = Int(maxOutputSamples * bytesPerSample * channels)
+        var allPCM = Data()
+        allPCM.reserveCapacity(parsed.frames.count * Int(baseSamples * bytesPerSample * channels))
+
+        for frame in parsed.frames {
+            // Per-packet total samples for output request
+            let totalSamples = opusTotalSamples(frame, sampleRate: sampleRate)
+
+            var pcmBuffer = [UInt8](repeating: 0, count: pcmBufferSize)
+            var outBufferList = AudioBufferList(
+                mNumberBuffers: 1,
+                mBuffers: AudioBuffer(
+                    mNumberChannels: channels,
+                    mDataByteSize: UInt32(pcmBufferSize),
+                    mData: &pcmBuffer
+                )
+            )
+            var ioOutputDataPacketSize: UInt32 = max(totalSamples, maxOutputSamples)
+
+            // Packet description for this Opus frame
+            var packetDesc = AudioStreamPacketDescription(
+                mStartOffset: 0,
+                mVariableFramesInPacket: 0,
+                mDataByteSize: UInt32(frame.count)
+            )
+
+            let inputContext = UnsafeMutablePointer<OpusInputContext>.allocate(capacity: 1)
+            inputContext.pointee = OpusInputContext(data: frame, packetDesc: packetDesc, consumed: false)
+            defer { inputContext.deallocate() }
+
+            status = AudioConverterFillComplexBuffer(
+                converter,
+                opusInputCallback,
+                inputContext,
+                &ioOutputDataPacketSize,
+                &outBufferList,
+                nil
+            )
+
+            if status == noErr || status == 100 /* done */ {
+                let produced = Int(outBufferList.mBuffers.mDataByteSize)
+                allPCM.append(contentsOf: pcmBuffer[0..<produced])
+            } else if status != noErr {
+                // Some frames may fail, continue with remaining
+                logger.warning("AudioConverter decode frame failed: \(status)")
+            }
+        }
+
+        guard !allPCM.isEmpty else { return nil }
+
+        // Build WAV header + PCM data
+        return buildWAV(pcmData: allPCM, sampleRate: UInt32(sampleRate),
+                        channels: UInt16(channels), bitsPerSample: UInt16(bytesPerSample * 8))
+    }
+
+    // Input callback for AudioConverterFillComplexBuffer
+    private struct OpusInputContext {
+        let data: Data
+        let packetDesc: AudioStreamPacketDescription
+        var consumed: Bool
+    }
+
+    private static let opusInputCallback: AudioConverterComplexInputDataProc = {
+        (converter, ioNumberDataPackets, ioData, outDataPacketDescription, inUserData) -> OSStatus in
+
+        guard let ctx = inUserData?.assumingMemoryBound(to: OpusInputContext.self) else {
+            ioNumberDataPackets.pointee = 0
+            return -50 // paramErr
+        }
+
+        if ctx.pointee.consumed {
+            ioNumberDataPackets.pointee = 0
+            return 100 // done
+        }
+
+        ctx.pointee.data.withUnsafeBytes { rawBuf in
+            ioData.pointee.mNumberBuffers = 1
+            ioData.pointee.mBuffers.mNumberChannels = 1
+            ioData.pointee.mBuffers.mDataByteSize = UInt32(rawBuf.count)
+            ioData.pointee.mBuffers.mData = UnsafeMutableRawPointer(mutating: rawBuf.baseAddress!)
+        }
+        ioNumberDataPackets.pointee = 1
+
+        if let descPtr = outDataPacketDescription {
+            descPtr.pointee = UnsafeMutablePointer<AudioStreamPacketDescription>.allocate(capacity: 1)
+            descPtr.pointee!.pointee = ctx.pointee.packetDesc
+        }
+
+        ctx.pointee.consumed = true
+        return noErr
+    }
+
+    /// Build minimal WAV file from raw PCM data.
+    private static func buildWAV(pcmData: Data, sampleRate: UInt32,
+                                  channels: UInt16, bitsPerSample: UInt16) -> Data {
+        let byteRate = sampleRate * UInt32(channels) * UInt32(bitsPerSample / 8)
+        let blockAlign = channels * (bitsPerSample / 8)
+        let dataSize = UInt32(pcmData.count)
+        let fileSize = 36 + dataSize
+
+        var wav = Data()
+        wav.reserveCapacity(Int(fileSize + 8))
+
+        // RIFF header
+        wav.append(contentsOf: [0x52, 0x49, 0x46, 0x46]) // "RIFF"
+        wav.appendLE32(fileSize)
+        wav.append(contentsOf: [0x57, 0x41, 0x56, 0x45]) // "WAVE"
+
+        // fmt chunk
+        wav.append(contentsOf: [0x66, 0x6D, 0x74, 0x20]) // "fmt "
+        wav.appendLE32(16)                                  // chunk size
+        wav.appendLE16(1)                                   // PCM format
+        wav.appendLE16(channels)
+        wav.appendLE32(sampleRate)
+        wav.appendLE32(byteRate)
+        wav.appendLE16(blockAlign)
+        wav.appendLE16(bitsPerSample)
+
+        // data chunk
+        wav.append(contentsOf: [0x64, 0x61, 0x74, 0x61]) // "data"
+        wav.appendLE32(dataSize)
+        wav.append(pcmData)
+
+        return wav
+    }
+
+    // MARK: - CAF Builder
+
+    private static func buildCAF(_ parsed: ParsedWebM) -> Data? {
+        var caf = Data()
+        let framesPerPacket: UInt32 = 960 // 20ms Opus frames at 48kHz
+
+        // -- File Header --
+        caf.appendBE32(0x63616666) // 'caff'
+        caf.appendBE16(1)          // version
+        caf.appendBE16(0)          // flags
+
+        // -- Audio Description ('desc') --
+        caf.appendBE32(0x64657363) // 'desc'
+        caf.appendBE64(32)         // chunk size = sizeof(AudioStreamBasicDescription)
+        caf.appendBEFloat64(parsed.sampleRate)
+        caf.appendBE32(0x6F707573) // 'opus' (kAudioFormatOpus)
+        caf.appendBE32(0)          // format flags
+        caf.appendBE32(0)          // bytes per packet (variable)
+        caf.appendBE32(framesPerPacket)
+        caf.appendBE32(parsed.channels)
+        caf.appendBE32(0)          // bits per channel (compressed)
+
+        // -- Magic Cookie ('kuki') — OpusHead from WebM CodecPrivate --
+        if let cp = parsed.codecPrivate, !cp.isEmpty {
+            caf.appendBE32(0x6B756B69) // 'kuki'
+            caf.appendBE64(Int64(cp.count))
+            caf.append(cp)
+        }
+
+        // -- Packet Table ('pakt') --
+        let numPackets = Int64(parsed.frames.count)
+        let numValidFrames = numPackets * Int64(framesPerPacket)
+
+        var paktBody = Data()
+        paktBody.appendBE64(numPackets)
+        paktBody.appendBE64(numValidFrames)
+        paktBody.appendBE32S(0)  // priming frames (Int32)
+        paktBody.appendBE32S(0)  // remainder frames (Int32)
+        for frame in parsed.frames {
+            paktBody.appendBERInt(frame.count)
+        }
+
+        caf.appendBE32(0x70616B74) // 'pakt'
+        caf.appendBE64(Int64(paktBody.count))
+        caf.append(paktBody)
+
+        // -- Audio Data ('data') --
+        var totalAudioSize = 0
+        for frame in parsed.frames { totalAudioSize += frame.count }
+
+        caf.appendBE32(0x64617461) // 'data'
+        caf.appendBE64(Int64(totalAudioSize + 4)) // +4 for editCount
+        caf.appendBE32(0)          // edit count
+        for frame in parsed.frames { caf.append(frame) }
+
+        return caf
+    }
+
+    // MARK: - CAF → M4A Transcoding
+
+    private static func transcodeCAFtoM4A(_ cafData: Data) -> Data? {
+        let tmp = FileManager.default.temporaryDirectory
+        let cafURL = tmp.appendingPathComponent("webm_conv_\(UUID().uuidString).caf")
+        let m4aURL = tmp.appendingPathComponent("webm_conv_\(UUID().uuidString).m4a")
+        defer {
+            try? FileManager.default.removeItem(at: cafURL)
+            try? FileManager.default.removeItem(at: m4aURL)
+        }
+
+        do {
+            try cafData.write(to: cafURL)
+
+            let inputFile = try AVAudioFile(forReading: cafURL)
+            let pcmFormat = inputFile.processingFormat
+
+            let outputSettings: [String: Any] = [
+                AVFormatIDKey: Int(kAudioFormatMPEG4AAC),
+                AVSampleRateKey: pcmFormat.sampleRate,
+                AVNumberOfChannelsKey: pcmFormat.channelCount,
+                AVEncoderBitRateKey: 64000
+            ]
+            let outputFile = try AVAudioFile(forWriting: m4aURL, settings: outputSettings)
+
+            let bufferCapacity: AVAudioFrameCount = 4096
+            guard let buffer = AVAudioPCMBuffer(pcmFormat: pcmFormat, frameCapacity: bufferCapacity) else {
+                return nil
+            }
+
+            while inputFile.framePosition < inputFile.length {
+                try inputFile.read(into: buffer)
+                try outputFile.write(from: buffer)
+            }
+
+            return try Data(contentsOf: m4aURL)
+        } catch {
+            logger.error("CAF→M4A transcode failed: \(error.localizedDescription)")
+            return nil
+        }
+    }
+
+    // MARK: - EBML Primitives
+
+    private static func readVINTValue(_ bytes: [UInt8], at pos: Int) -> (UInt64, Int)? {
+        guard pos < bytes.count, bytes[pos] != 0 else { return nil }
+        let length = bytes[pos].leadingZeroBitCount + 1
+        guard length <= 8, pos + length <= bytes.count else { return nil }
+
+        var value = UInt64(bytes[pos]) & ((1 << (8 - length)) - 1)
+        for i in 1..<length { value = (value << 8) | UInt64(bytes[pos + i]) }
+        return (value, length)
+    }
+
+    private static func readElementID(_ bytes: [UInt8], at pos: Int) -> (UInt32, Int)? {
+        guard pos < bytes.count, bytes[pos] != 0 else { return nil }
+        let length = bytes[pos].leadingZeroBitCount + 1
+        guard length <= 4, pos + length <= bytes.count else { return nil }
+
+        var id = UInt32(bytes[pos])
+        for i in 1..<length { id = (id << 8) | UInt32(bytes[pos + i]) }
+        return (id, length)
+    }
+
+    private static func readUInt(_ bytes: [UInt8], at pos: Int, count: Int) -> UInt64 {
+        var value: UInt64 = 0
+        for i in 0..<min(count, 8) where pos + i < bytes.count {
+            value = (value << 8) | UInt64(bytes[pos + i])
+        }
+        return value
+    }
+
+    private static func readFloat(_ bytes: [UInt8], at pos: Int, count: Int) -> Double {
+        if count == 4, pos + 4 <= bytes.count {
+            let bits = (UInt32(bytes[pos]) << 24) | (UInt32(bytes[pos + 1]) << 16) |
+                       (UInt32(bytes[pos + 2]) << 8) | UInt32(bytes[pos + 3])
+            return Double(Float(bitPattern: bits))
+        } else if count == 8, pos + 8 <= bytes.count {
+            var bits: UInt64 = 0
+            for i in 0..<8 { bits = (bits << 8) | UInt64(bytes[pos + i]) }
+            return Double(bitPattern: bits)
+        }
+        return 48000 // default Opus sample rate
+    }
+}
+
+// MARK: - Data Little-Endian Writers (WAV format)
+
+private extension Data {
+    mutating func appendLE16(_ v: UInt16) {
+        append(UInt8(v & 0xFF)); append(UInt8(v >> 8))
+    }
+    mutating func appendLE32(_ v: UInt32) {
+        append(UInt8(v & 0xFF)); append(UInt8((v >> 8) & 0xFF))
+        append(UInt8((v >> 16) & 0xFF)); append(UInt8((v >> 24) & 0xFF))
+    }
+}
+
+// MARK: - Data Big-Endian Writers (CAF format)
+
+private extension Data {
+    mutating func appendBE16(_ v: UInt16) {
+        append(UInt8(v >> 8)); append(UInt8(v & 0xFF))
+    }
+    mutating func appendBE32(_ v: UInt32) {
+        append(UInt8((v >> 24) & 0xFF)); append(UInt8((v >> 16) & 0xFF))
+        append(UInt8((v >> 8) & 0xFF));  append(UInt8(v & 0xFF))
+    }
+    mutating func appendBE32S(_ v: Int32) { appendBE32(UInt32(bitPattern: v)) }
+    mutating func appendBE64(_ v: Int64) {
+        let u = UInt64(bitPattern: v)
+        for i in stride(from: 56, through: 0, by: -8) { append(UInt8((u >> i) & 0xFF)) }
+    }
+    mutating func appendBEFloat64(_ v: Double) {
+        let bits = v.bitPattern
+        for i in stride(from: 56, through: 0, by: -8) { append(UInt8((bits >> i) & 0xFF)) }
+    }
+    /// Variable-length BER integer (CAF packet table).
+    mutating func appendBERInt(_ value: Int) {
+        var v = value
+        var encoded = [UInt8]()
+        encoded.append(UInt8(v & 0x7F))
+        v >>= 7
+        while v > 0 {
+            encoded.append(UInt8(0x80 | (v & 0x7F)))
+            v >>= 7
+        }
+        for byte in encoded.reversed() { append(byte) }
+    }
+}

Rosetta/Core/Utils/DarkMode+Helpers.swift

@@ -28,6 +28,7 @@ struct DarkModeWrapper<Content: View>: View {
                     if let windowScene = activeWindowScene {
                         let overlayWindow = UIWindow(windowScene: windowScene)
                         overlayWindow.tag = 0320
+                        overlayWindow.backgroundColor = .clear
                         overlayWindow.isHidden = false
                         overlayWindow.isUserInteractionEnabled = false
                         self.overlayWindow = overlayWindow
@@ -42,8 +43,15 @@ struct DarkModeWrapper<Content: View>: View {
                     case "system": style = .unspecified
                     default: style = .dark
                     }
+                    let bgColor: UIColor = (style == .light) ? .white : .black
                     for window in windowScene.windows {
                         window.overrideUserInterfaceStyle = style
+                        // Match window background to app background — prevents
+                        // systemBackground (dark gray) from showing as a line
+                        // in the bottom safe area.
+                        if window.tag != 0320 {
+                            window.backgroundColor = bgColor
+                        }
                     }
                 }
             }

Rosetta/Features/Chats/ChatDetail/MessageVoiceView.swift

@@ -11,6 +11,8 @@ final class MessageVoiceView: UIView {
 
     private let playButton = UIButton(type: .system)
     private let playPauseAnimationView = LottieAnimationView()
+    private let downloadArrowView = UIImageView()  // ↓ arrow before download
+    private let downloadRingView = VoiceDownloadRingView()
     private let waveformView = WaveformView()
     private let durationLabel = UILabel()
 
@@ -20,6 +22,7 @@ final class MessageVoiceView: UIView {
     private var attachmentId: String = ""
     private var isOutgoing = false
     private var isShowingPause = false  // tracks Lottie visual state
+    private var isDownloaded = false    // false = show ↓, true = show play/pause
     private var totalDuration: TimeInterval = 0  // original duration for label reset
     /// Center of play button in this view's coordinate space (for external blob positioning).
     var playButtonCenter: CGPoint { playButton.center }
@@ -50,6 +53,8 @@ final class MessageVoiceView: UIView {
 
     // Playback
     var onPlayTapped: (() -> Void)?
+    /// Download cancel callback (forwarded from VoiceDownloadRingView).
+    var onDownloadCancel: (() -> Void)?
 
     // MARK: - Setup
 
@@ -73,6 +78,17 @@ final class MessageVoiceView: UIView {
             playButton.setImage(UIImage(systemName: "play.fill", withConfiguration: config), for: .normal)
         }
 
+        // Download arrow (shown before voice is downloaded)
+        let arrowConfig = UIImage.SymbolConfiguration(pointSize: 16, weight: .bold)
+        downloadArrowView.image = UIImage(systemName: "arrow.down", withConfiguration: arrowConfig)
+        downloadArrowView.contentMode = .center
+        downloadArrowView.isUserInteractionEnabled = false
+        playButton.addSubview(downloadArrowView)
+
+        // Download progress ring (overlays play button during CDN download)
+        downloadRingView.onCancel = { [weak self] in self?.onDownloadCancel?() }
+        addSubview(downloadRingView)
+
         waveformView.peakHeight = 18    // Telegram AudioWaveformComponent peak
         waveformView.distance = 2.0    // Telegram AudioWaveformComponent (bubble context)
         waveformView.gravity = .bottom  // Telegram: bars grow upward from bottom
@@ -106,6 +122,11 @@ final class MessageVoiceView: UIView {
             width: playButtonSize - lottieInset * 2,
             height: playButtonSize - lottieInset * 2
         )
+        // Download arrow: same frame as button interior
+        downloadArrowView.frame = playButton.bounds
+
+        // Download ring: same frame as play button
+        downloadRingView.frame = playButton.frame
 
         // Waveform: from x=57 to near right edge, height=18, y=1
         let waveW = bounds.width - waveformX - 4
@@ -163,6 +184,8 @@ final class MessageVoiceView: UIView {
 
         playButton.backgroundColor = colors.playButtonBg
         playButton.tintColor = colors.playButtonFg
+        downloadArrowView.tintColor = colors.playButtonFg
+        downloadRingView.setRingColor(colors.playButtonFg)
         durationLabel.textColor = colors.durationText
         waveformView.foregroundColor_ = colors.waveformPlayed
         waveformView.backgroundColor_ = colors.waveformUnplayed
@@ -188,6 +211,32 @@ final class MessageVoiceView: UIView {
         }
     }
 
+    // MARK: - Download State
+
+    /// Set whether voice data is already downloaded (cached).
+    /// `false` → show ↓ arrow. `true` → show play/pause icon.
+    func setDownloaded(_ downloaded: Bool) {
+        isDownloaded = downloaded
+        downloadArrowView.isHidden = downloaded
+        playPauseAnimationView.isHidden = !downloaded
+    }
+
+    /// Show download progress ring overlaying the play button.
+    func showDownloadProgress(_ progress: CGFloat) {
+        downloadRingView.show()
+        downloadRingView.setProgress(progress)
+        downloadArrowView.isHidden = true
+        playPauseAnimationView.isHidden = true
+    }
+
+    /// Hide download progress ring and show play icon (download complete).
+    func hideDownloadProgress() {
+        downloadRingView.hide()
+        isDownloaded = true
+        downloadArrowView.isHidden = true
+        playPauseAnimationView.isHidden = false
+    }
+
     // MARK: - Play Action
 
     @objc private func playTapped() {
@@ -295,12 +344,14 @@ final class MessageVoiceView: UIView {
 
     // MARK: - Waveform Encoding
 
-    /// Encode waveform samples to 5-bit packed base64 string (for sending).
+    /// Encode waveform samples to comma-separated floats (Desktop parity).
+    /// Desktop DialogInput.tsx:217 sends `interpolateCompressWaves(35).join(",")`.
+    /// Desktop MessageVoice.tsx parses with `split(",").map(parseFloat)`.
     static func encodeWaveform(_ samples: [Float]) -> String {
         guard !samples.isEmpty else { return "" }
 
-        // Resample to ~63 bars (Telegram standard)
+        // Resample to 35 bars (Desktop standard: interpolateCompressWaves(35))
-        let targetCount = min(63, samples.count)
+        let targetCount = min(35, samples.count)
         let step = Float(samples.count) / Float(targetCount)
         var resampled = [Float](repeating: 0, count: targetCount)
         for i in 0..<targetCount {
@@ -310,28 +361,7 @@ final class MessageVoiceView: UIView {
             resampled[i] = samples[start..<end].max() ?? 0
         }
 
-        // Pack as 5-bit values
+        // Comma-separated floats (Desktop parity)
-        let bitCount = targetCount * 5
+        return resampled.map { String(format: "%.2f", $0) }.joined(separator: ",")
-        let byteCount = (bitCount + 7) / 8
-        var bytes = [UInt8](repeating: 0, count: byteCount)
-
-        for i in 0..<targetCount {
-            let value = UInt8(min(31, max(0, resampled[i] * 31)))
-            let bitOffset = i * 5
-            let byteIndex = bitOffset / 8
-            let bitIndex = bitOffset % 8
-
-            if bitIndex + 5 <= 8 {
-                bytes[byteIndex] |= value << (8 - bitIndex - 5)
-            } else {
-                let bitsInFirst = 8 - bitIndex
-                bytes[byteIndex] |= value >> (5 - bitsInFirst)
-                if byteIndex + 1 < bytes.count {
-                    bytes[byteIndex + 1] |= value << (8 - (5 - bitsInFirst))
-                }
-            }
-        }
-
-        return Data(bytes).base64EncodedString()
     }
 }

Rosetta/Features/Chats/ChatDetail/NativeMessageCell.swift

@@ -193,6 +193,7 @@ final class NativeMessageCell: UICollectionViewCell {
     // Voice message
     private let voiceView = MessageVoiceView()
     private var voiceBlobView: VoiceBlobView?
+    private var activeVoiceDownloadTask: Task<Void, Never>?
 
     // Avatar-specific
     private let avatarImageView = UIImageView()
@@ -881,30 +882,61 @@ final class NativeMessageCell: UICollectionViewCell {
                     duration: previewParts.duration,
                     isOutgoing: layout.isOutgoing
                 )
+                let voiceAttachment = voiceAtt
+                let storedPassword = message.attachmentPassword
+                let playbackDuration = previewParts.duration
+                let playbackMessageId = message.id
+
+                // Check cache: ↓ arrow if not downloaded, play icon if cached
+                let voiceFileName = "voice_\(Int(playbackDuration))s.m4a"
+                let isCached = Self.playableVoiceURLFromCache(
+                    attachmentId: voiceAttachment.id, fileName: voiceFileName
+                ) != nil
+                // Own outgoing voice = always "downloaded" (data came from local recording)
+                let isOwnVoice = layout.isOutgoing
+                voiceView.setDownloaded(isCached || isOwnVoice)
+
                 let isCurrentVoice = VoiceMessagePlayer.shared.currentMessageId == message.id
                 voiceView.updatePlaybackState(
                     isPlaying: isCurrentVoice && VoiceMessagePlayer.shared.isPlaying,
                     progress: isCurrentVoice ? CGFloat(VoiceMessagePlayer.shared.progress) : 0
                 )
-                let voiceAttachment = voiceAtt
+
-                let storedPassword = message.attachmentPassword
-                let playbackDuration = previewParts.duration
-                let playbackMessageId = message.id
                 voiceView.onPlayTapped = { [weak self] in
                     guard let self else { return }
-                    Task.detached(priority: .userInitiated) {
+
-                        guard let playableURL = await Self.resolvePlayableVoiceURL(
+                    // If already cached — play immediately
-                            attachment: voiceAttachment,
+                    if let cached = Self.playableVoiceURLFromCache(
-                            duration: playbackDuration,
+                        attachmentId: voiceAttachment.id, fileName: voiceFileName
-                            storedPassword: storedPassword
+                    ) {
-                        ) else {
+                        self.voiceView.setDownloaded(true)
+                        VoiceMessagePlayer.shared.play(messageId: playbackMessageId, fileURL: cached)
                         return
                     }
-                        await MainActor.run {
+
-                            guard self.message?.id == playbackMessageId else { return }
+                    // Show progress ring and start download
+                    self.voiceView.showDownloadProgress(0.027)
+                    let downloadTask = Task {
+                        let playableURL = await Self.resolvePlayableVoiceURL(
+                            attachment: voiceAttachment,
+                            duration: playbackDuration,
+                            storedPassword: storedPassword,
+                            onProgress: { [weak self] progress in
+                                self?.voiceView.showDownloadProgress(CGFloat(progress))
+                            }
+                        )
+                        guard !Task.isCancelled else { return }
+                        self.voiceView.hideDownloadProgress()
+                        if let playableURL, self.message?.id == playbackMessageId {
                             VoiceMessagePlayer.shared.play(messageId: playbackMessageId, fileURL: playableURL)
                         }
                     }
+                    self.activeVoiceDownloadTask = downloadTask
+                }
+                voiceView.onDownloadCancel = { [weak self] in
+                    self?.activeVoiceDownloadTask?.cancel()
+                    self?.voiceView.hideDownloadProgress()
+                    self?.activeVoiceDownloadTask = nil
                 }
                 fileIconView.isHidden = true
                 fileNameLabel.isHidden = true
@@ -1550,14 +1582,19 @@ final class NativeMessageCell: UICollectionViewCell {
     private static func resolvePlayableVoiceURL(
         attachment: MessageAttachment,
         duration: TimeInterval,
-        storedPassword: String?
+        storedPassword: String?,
+        onProgress: (@MainActor (Double) -> Void)? = nil
     ) async -> URL? {
         let fileName = "voice_\(Int(duration))s.m4a"
         if let cached = playableVoiceURLFromCache(attachmentId: attachment.id, fileName: fileName) {
             return cached
         }
 
-        guard let downloaded = await downloadVoiceData(attachment: attachment, storedPassword: storedPassword) else {
+        guard let downloaded = await downloadVoiceData(
+            attachment: attachment,
+            storedPassword: storedPassword,
+            onProgress: onProgress
+        ) else {
             return nil
         }
         _ = AttachmentCache.shared.saveFile(downloaded, forAttachmentId: attachment.id, fileName: fileName)
@@ -1591,7 +1628,11 @@ final class NativeMessageCell: UICollectionViewCell {
         }
     }
 
-    private static func downloadVoiceData(attachment: MessageAttachment, storedPassword: String?) async -> Data? {
+    private static func downloadVoiceData(
+        attachment: MessageAttachment,
+        storedPassword: String?,
+        onProgress: (@MainActor (Double) -> Void)? = nil
+    ) async -> Data? {
         let tag = attachment.effectiveDownloadTag
         guard !tag.isEmpty else { return nil }
         guard let storedPassword, !storedPassword.isEmpty else { return nil }
@@ -1599,14 +1640,24 @@ final class NativeMessageCell: UICollectionViewCell {
         do {
             let encryptedData = try await TransportManager.shared.downloadFile(
                 tag: tag,
-                server: attachment.transportServer
+                server: attachment.transportServer,
+                onProgress: onProgress
             )
             let encryptedString = String(decoding: encryptedData, as: UTF8.self)
             let passwords = MessageCrypto.attachmentPasswordCandidates(from: storedPassword)
             guard let decrypted = decryptAttachmentData(encryptedString: encryptedString, passwords: passwords) else {
                 return nil
             }
-            return parseAttachmentFileData(decrypted)
+            let rawData = parseAttachmentFileData(decrypted)
+
+            // Desktop sends WebM/Opus — convert to M4A for iOS playback.
+            // Transcoding (~200ms) runs off MainActor to avoid UI hitch.
+            if WebMOpusConverter.isWebM(rawData) {
+                return await Task.detached(priority: .userInitiated) {
+                    WebMOpusConverter.convertToPlayable(rawData)
+                }.value
+            }
+            return rawData
         } catch {
             return nil
         }
@@ -1632,12 +1683,20 @@ final class NativeMessageCell: UICollectionViewCell {
     }
 
     private static func parseAttachmentFileData(_ data: Data) -> Data {
+        // 1. Data URI format (iOS images, files, legacy voice)
         if let string = String(data: data, encoding: .utf8),
            string.hasPrefix("data:"),
            let comma = string.firstIndex(of: ",") {
             let payload = String(string[string.index(after: comma)...])
             return Data(base64Encoded: payload) ?? data
         }
+        // 2. Hex-encoded raw bytes (Desktop voice: Buffer.toString('hex'))
+        if let string = String(data: data, encoding: .utf8),
+           string.count >= 100,
+           string.allSatisfy({ $0.isHexDigit }) {
+            return Data(hexString: string)
+        }
+        // 3. Raw binary (fallback)
         return data
     }
 

Rosetta/Features/Chats/ChatDetail/VoiceDownloadRingView.swift

@@ -0,0 +1,169 @@
+import UIKit
+
+/// Telegram-parity circular progress ring for voice message downloads.
+/// Overlays the 44×44 play button during CDN download.
+///
+/// Reference: `SemanticStatusNodeProgressContext.swift` in Telegram-iOS.
+/// - Arc starts at 12 o'clock (-π/2), fills clockwise
+/// - Rounded line caps, ~2pt stroke, 2.5pt inset
+/// - Continuous rotation animation (4× speed)
+/// - Cancel ✕ in center (12pt, 1.8pt stroke)
+/// - Minimum visible progress: 2.7%
+final class VoiceDownloadRingView: UIView {
+
+    // MARK: - Telegram-exact constants
+
+    private let lineWidth: CGFloat = 2.0
+    private let inset: CGFloat = 2.5
+    private let startAngle: CGFloat = -.pi / 2
+    private let minProgress: CGFloat = 0.027
+    private let cancelCrossSize: CGFloat = 12.0
+    private let cancelLineWidth: CGFloat = 1.8
+
+    // MARK: - Callbacks
+
+    var onCancel: (() -> Void)?
+
+    // MARK: - Layers
+
+    private let progressLayer = CAShapeLayer()
+    private let cancelLayer = CAShapeLayer()
+
+    // MARK: - Init
+
+    override init(frame: CGRect) {
+        super.init(frame: frame)
+        setup()
+    }
+
+    required init?(coder: NSCoder) {
+        super.init(coder: coder)
+        setup()
+    }
+
+    private func setup() {
+        isUserInteractionEnabled = true
+        isHidden = true
+        backgroundColor = .clear
+
+        // Progress ring
+        progressLayer.fillColor = nil
+        progressLayer.lineCap = .round
+        progressLayer.lineWidth = lineWidth
+        progressLayer.strokeStart = 0
+        progressLayer.strokeEnd = minProgress
+        layer.addSublayer(progressLayer)
+
+        // Cancel ✕
+        cancelLayer.fillColor = nil
+        cancelLayer.lineCap = .round
+        cancelLayer.lineWidth = cancelLineWidth
+        layer.addSublayer(cancelLayer)
+
+        let tap = UITapGestureRecognizer(target: self, action: #selector(cancelTapped))
+        addGestureRecognizer(tap)
+    }
+
+    // MARK: - Layout
+
+    override func layoutSubviews() {
+        super.layoutSubviews()
+        let size = bounds.size
+        guard size.width > 0 else { return }
+
+        // Progress arc path (full circle — strokeEnd controls visible portion)
+        let pathDiameter = size.width - lineWidth - inset * 2
+        let radius = pathDiameter / 2
+        let center = CGPoint(x: size.width / 2, y: size.height / 2)
+        let circlePath = UIBezierPath(
+            arcCenter: center,
+            radius: radius,
+            startAngle: startAngle,
+            endAngle: startAngle + .pi * 2,
+            clockwise: true
+        )
+        progressLayer.path = circlePath.cgPath
+        progressLayer.frame = bounds
+
+        // Cancel ✕ centered
+        let half = cancelCrossSize / 2
+        let crossPath = UIBezierPath()
+        crossPath.move(to: CGPoint(x: center.x - half, y: center.y - half))
+        crossPath.addLine(to: CGPoint(x: center.x + half, y: center.y + half))
+        crossPath.move(to: CGPoint(x: center.x + half, y: center.y - half))
+        crossPath.addLine(to: CGPoint(x: center.x - half, y: center.y + half))
+        cancelLayer.path = crossPath.cgPath
+        cancelLayer.frame = bounds
+    }
+
+    // MARK: - Public API
+
+    /// Update ring color to match play button foreground.
+    func setRingColor(_ color: UIColor) {
+        progressLayer.strokeColor = color.cgColor
+        cancelLayer.strokeColor = color.cgColor
+    }
+
+    /// Set download progress (0.0–1.0). Values below 2.7% are clamped up.
+    func setProgress(_ value: CGFloat, animated: Bool = true) {
+        let clamped = max(minProgress, min(1.0, value))
+
+        if animated {
+            let anim = CABasicAnimation(keyPath: "strokeEnd")
+            anim.fromValue = progressLayer.presentation()?.strokeEnd ?? progressLayer.strokeEnd
+            anim.toValue = clamped
+            anim.duration = 0.2
+            anim.timingFunction = CAMediaTimingFunction(name: .easeInEaseOut)
+            anim.isRemovedOnCompletion = false
+            anim.fillMode = .forwards
+            progressLayer.add(anim, forKey: "progress")
+        } else {
+            progressLayer.removeAnimation(forKey: "progress")
+        }
+        progressLayer.strokeEnd = clamped
+    }
+
+    /// Show the ring and start rotation.
+    func show() {
+        guard isHidden else { return }
+        isHidden = false
+        alpha = 0
+        UIView.animate(withDuration: 0.18) { self.alpha = 1 }
+        startRotation()
+    }
+
+    /// Hide the ring and stop rotation.
+    func hide() {
+        guard !isHidden else { return }
+        UIView.animate(withDuration: 0.18) {
+            self.alpha = 0
+        } completion: { _ in
+            self.isHidden = true
+            self.stopRotation()
+            self.setProgress(self.minProgress, animated: false)
+        }
+    }
+
+    // MARK: - Rotation
+
+    private func startRotation() {
+        guard progressLayer.animation(forKey: "rotation") == nil else { return }
+        let rotation = CABasicAnimation(keyPath: "transform.rotation.z")
+        rotation.fromValue = 0
+        rotation.toValue = CGFloat.pi * 2
+        rotation.duration = 1.6  // ~4× per 2π normalized (Telegram: angle * 4.0)
+        rotation.repeatCount = .infinity
+        rotation.isRemovedOnCompletion = false
+        progressLayer.add(rotation, forKey: "rotation")
+    }
+
+    private func stopRotation() {
+        progressLayer.removeAnimation(forKey: "rotation")
+    }
+
+    // MARK: - Actions
+
+    @objc private func cancelTapped() {
+        onCancel?()
+    }
+}

Rosetta/RosettaApp.swift

@@ -801,7 +801,9 @@ struct RosettaApp: App {
             return
         }
 
-        UIWindow.appearance().backgroundColor = .systemBackground
+        UIWindow.appearance().backgroundColor = UIColor { traits in
+            traits.userInterfaceStyle == .dark ? .black : .white
+        }
 
         // Detect fresh install: UserDefaults are wiped on uninstall, Keychain is not.
         // If this is the first launch after install, clear any stale Keychain data.
@@ -862,6 +864,15 @@ struct RosettaApp: App {
                     if appState == nil {
                         appState = initialState()
                     }
+                    // Set hosting controller & window background to match app background.
+                    // Default .systemBackground (dark gray) leaks as a line in the safe area.
+                    DispatchQueue.main.async {
+                        guard let scene = UIApplication.shared.connectedScenes.first as? UIWindowScene,
+                              let window = scene.windows.first(where: { $0.tag != 0320 }) else { return }
+                        let bgColor: UIColor = window.traitCollection.userInterfaceStyle == .dark ? .black : .white
+                        window.backgroundColor = bgColor
+                        window.rootViewController?.view.backgroundColor = bgColor
+                    }
                 }
                 .onOpenURL { url in
                     handleDeepLink(url)