main

media-worker/Services/AudioCraftService.cs

@@ -0,0 +1,302 @@
+using System.Text;
+using System.Text.Json;
+using Microsoft.Extensions.Logging;
+using Microsoft.Extensions.Options;
+using SaasMediaWorker.Configuration;
+using SaasMediaWorker.Models;
+
+namespace SaasMediaWorker.Services;
+
+/// <summary>
+/// AudioCraft Service — Meta MusicGen + AudioGen entegrasyonu.
+/// 
+/// HuggingFace Inference API üzerinden çalışır (self-hosted model gerekmez).
+/// Raspberry Pi'da çalıştırılabilir — model inference cloud'da yapılır.
+///
+/// Yetenekler:
+///   MusicGen: Text-to-music (müzik üretimi)
+///     - Genre, BPM, enstrüman, mood bazlı prompt
+///     - Stereo çıktı desteği
+///     - 30 saniyeye kadar üretim
+///   AudioGen: Text-to-sound (ses efekti üretimi)
+///     - Ortam sesleri: yağmur, rüzgâr, deniz dalgaları
+///     - Foley sesleri: ayak sesleri, kapı gıcırtısı
+///     - Sahne bazlı ambient ses katmanları
+///
+/// multimodal-audiocraft skill'inden elde edilen bilgilerle tasarlandı.
+/// </summary>
+public class AudioCraftService
+{
+    private readonly HttpClient _httpClient;
+    private readonly ILogger<AudioCraftService> _logger;
+    private readonly ApiSettings _settings;
+
+    // HuggingFace Inference API endpoints
+    private const string MUSICGEN_MODEL = "facebook/musicgen-small";
+    private const string AUDIOGEN_MODEL = "facebook/audiogen-medium";
+    private const string HF_API_BASE = "https://api-inference.huggingface.co/models";
+
+    public AudioCraftService(
+        HttpClient httpClient,
+        ILogger<AudioCraftService> logger,
+        IOptions<ApiSettings> settings)
+    {
+        _httpClient = httpClient;
+        _logger = logger;
+        _settings = settings.Value;
+
+        _httpClient.DefaultRequestHeaders.Add(
+            "Authorization", $"Bearer {_settings.HuggingFaceApiKey}");
+        _httpClient.Timeout = TimeSpan.FromMinutes(5);
+    }
+
+    /// <summary>
+    /// MusicGen ile text-to-music üretimi.
+    /// Proje için background müzik oluşturur.
+    /// </summary>
+    public async Task<GeneratedMediaFile> GenerateMusicAsync(
+        string musicPrompt,
+        MusicTechnicalParams? technicalParams,
+        int targetDurationSeconds,
+        string outputDirectory,
+        CancellationToken ct)
+    {
+        // Prompt'u teknik parametrelerle zenginleştir
+        var enrichedPrompt = EnrichMusicPrompt(musicPrompt, technicalParams);
+
+        _logger.LogInformation(
+            "🎵 MusicGen müzik üretimi — Prompt: \"{Prompt}\", Süre: {Duration}s",
+            enrichedPrompt[..Math.Min(100, enrichedPrompt.Length)],
+            targetDurationSeconds);
+
+        var audioBytes = await CallHuggingFaceInference(
+            MUSICGEN_MODEL, enrichedPrompt, ct);
+
+        if (audioBytes == null || audioBytes.Length == 0)
+        {
+            throw new InvalidOperationException("MusicGen boş yanıt döndü");
+        }
+
+        // WAV dosyasını kaydet
+        var outputPath = Path.Combine(outputDirectory, "background_music_audiocraft.wav");
+        await File.WriteAllBytesAsync(outputPath, audioBytes, ct);
+
+        var fileInfo = new FileInfo(outputPath);
+        _logger.LogInformation(
+            "✅ MusicGen müzik üretildi: {Path} ({Size:N0} bytes)",
+            outputPath, fileInfo.Length);
+
+        return new GeneratedMediaFile
+        {
+            SceneId = string.Empty,
+            SceneOrder = 0,
+            Type = MediaFileType.AudioMusic,
+            LocalPath = outputPath,
+            FileSizeBytes = fileInfo.Length,
+            DurationSeconds = targetDurationSeconds,
+            MimeType = "audio/wav",
+            AiProvider = "audiocraft-musicgen"
+        };
+    }
+
+    /// <summary>
+    /// AudioGen ile sahne bazlı ambient ses efekti üretimi.
+    /// Her sahne için farklı bir ortam sesi oluşturulabilir.
+    /// </summary>
+    public async Task<GeneratedMediaFile?> GenerateAmbientSoundAsync(
+        string ambientPrompt,
+        int sceneOrder,
+        double durationSeconds,
+        string outputDirectory,
+        CancellationToken ct)
+    {
+        if (string.IsNullOrWhiteSpace(ambientPrompt))
+            return null;
+
+        _logger.LogInformation(
+            "🔊 AudioGen ses efekti — Sahne: {Order}, Prompt: \"{Prompt}\"",
+            sceneOrder, ambientPrompt[..Math.Min(80, ambientPrompt.Length)]);
+
+        try
+        {
+            var audioBytes = await CallHuggingFaceInference(
+                AUDIOGEN_MODEL, ambientPrompt, ct);
+
+            if (audioBytes == null || audioBytes.Length == 0)
+            {
+                _logger.LogWarning("AudioGen boş yanıt — sahne {Order} için ambient atlanıyor", sceneOrder);
+                return null;
+            }
+
+            var fileName = $"ambient_scene_{sceneOrder:D3}.wav";
+            var outputPath = Path.Combine(outputDirectory, fileName);
+            await File.WriteAllBytesAsync(outputPath, audioBytes, ct);
+
+            _logger.LogInformation(
+                "✅ Ambient ses üretildi: {FileName} ({Size:N0} bytes)",
+                fileName, audioBytes.Length);
+
+            return new GeneratedMediaFile
+            {
+                SceneId = string.Empty,
+                SceneOrder = sceneOrder,
+                Type = MediaFileType.AudioAmbient,
+                LocalPath = outputPath,
+                FileSizeBytes = audioBytes.Length,
+                DurationSeconds = durationSeconds,
+                MimeType = "audio/wav",
+                AiProvider = "audiocraft-audiogen"
+            };
+        }
+        catch (Exception ex)
+        {
+            _logger.LogWarning(ex,
+                "Ambient ses üretimi başarısız (sahne {Order}) — devam ediliyor", sceneOrder);
+            return null;
+        }
+    }
+
+    /// <summary>
+    /// Projenin tüm sahneleri için batch ambient ses üretimi.
+    /// Paralel çalışır — Raspberry Pi'da bellek dostu.
+    /// </summary>
+    public async Task<List<GeneratedMediaFile>> GenerateAllAmbientSoundsAsync(
+        List<ScenePayload> scenes,
+        string outputDirectory,
+        CancellationToken ct)
+    {
+        var results = new List<GeneratedMediaFile>();
+
+        // Sıralı üret (HuggingFace rate limit + Pi bellek optimizasyonu)
+        foreach (var scene in scenes)
+        {
+            var ambientPrompt = scene.AmbientSoundPrompt;
+            if (string.IsNullOrWhiteSpace(ambientPrompt))
+                continue;
+
+            var result = await GenerateAmbientSoundAsync(
+                ambientPrompt, scene.Order, scene.Duration,
+                outputDirectory, ct);
+
+            if (result != null)
+                results.Add(result);
+
+            // Rate limit koruma — 1 saniye bekle
+            await Task.Delay(1000, ct);
+        }
+
+        _logger.LogInformation(
+            "🔊 Toplam {Count} sahne için ambient ses üretildi", results.Count);
+
+        return results;
+    }
+
+    // ── Private: HuggingFace Inference API çağrısı ──────────────────
+
+    private async Task<byte[]?> CallHuggingFaceInference(
+        string modelId, string prompt, CancellationToken ct)
+    {
+        var url = $"{HF_API_BASE}/{modelId}";
+
+        var payload = new { inputs = prompt };
+        var content = new StringContent(
+            JsonSerializer.Serialize(payload),
+            Encoding.UTF8,
+            "application/json");
+
+        var maxRetries = 3;
+        for (var attempt = 1; attempt <= maxRetries; attempt++)
+        {
+            try
+            {
+                var response = await _httpClient.PostAsync(url, content, ct);
+
+                // Model yükleniyor (cold start)
+                if (response.StatusCode == System.Net.HttpStatusCode.ServiceUnavailable ||
+                    (int)response.StatusCode == 503)
+                {
+                    var json = await response.Content.ReadAsStringAsync(ct);
+                    _logger.LogInformation(
+                        "Model yükleniyor ({Model}), deneme {Attempt}/{Max}...",
+                        modelId, attempt, maxRetries);
+
+                    // Model yüklenme süresi bekleme
+                    var waitTime = ExtractEstimatedTime(json);
+                    await Task.Delay(TimeSpan.FromSeconds(waitTime), ct);
+                    continue;
+                }
+
+                response.EnsureSuccessStatusCode();
+
+                // Audio binary yanıt
+                return await response.Content.ReadAsByteArrayAsync(ct);
+            }
+            catch (HttpRequestException ex) when (attempt < maxRetries)
+            {
+                _logger.LogWarning(ex,
+                    "HuggingFace API hatası, deneme {Attempt}/{Max}", attempt, maxRetries);
+                await Task.Delay(3000 * attempt, ct);
+            }
+        }
+
+        throw new InvalidOperationException(
+            $"HuggingFace API {maxRetries} deneme sonrası başarısız — Model: {modelId}");
+    }
+
+    /// <summary>
+    /// MusicGen prompt'unu teknik parametrelerle zenginleştirir.
+    /// AudioCraft skill'den öğrenilen best practice'lere göre optimize eder.
+    /// </summary>
+    private string EnrichMusicPrompt(string basePrompt, MusicTechnicalParams? technical)
+    {
+        if (technical == null)
+            return basePrompt;
+
+        var enriched = basePrompt;
+
+        // BPM ekle (prompt'ta yoksa)
+        if (!enriched.Contains("BPM", StringComparison.OrdinalIgnoreCase) && technical.Bpm > 0)
+        {
+            enriched += $", {technical.Bpm} BPM";
+        }
+
+        // Key ekle
+        if (!string.IsNullOrEmpty(technical.Key) &&
+            !enriched.Contains(technical.Key, StringComparison.OrdinalIgnoreCase))
+        {
+            enriched += $", {technical.Key}";
+        }
+
+        // Emotional arc ekle
+        if (!string.IsNullOrEmpty(technical.EmotionalArc))
+        {
+            enriched += $", {technical.EmotionalArc.Replace("-", " ")} energy";
+        }
+
+        return enriched;
+    }
+
+    private int ExtractEstimatedTime(string json)
+    {
+        try
+        {
+            var doc = JsonSerializer.Deserialize<JsonElement>(json);
+            if (doc.TryGetProperty("estimated_time", out var time))
+                return Math.Max(10, (int)time.GetDouble());
+        }
+        catch { }
+
+        return 20; // Default: 20 saniye bekle
+    }
+}
+
+/// <summary>
+/// MusicGen teknik parametreleri — AI senaryo çıktısından parse edilir.
+/// </summary>
+public class MusicTechnicalParams
+{
+    public int Bpm { get; set; }
+    public string? Key { get; set; }
+    public List<string> Instruments { get; set; } = new();
+    public string EmotionalArc { get; set; } = string.Empty;
+}