app/Services/EdgeTtsService.php

<?php

namespace App\Services;

use Illuminate\Support\Facades\Http;
use Illuminate\Support\Facades\Log;
use Illuminate\Support\Facades\Process;
use Illuminate\Support\Facades\Storage;
use Illuminate\Support\Str;
use RuntimeException;

class EdgeTtsService
{
    public function synthesize(string $text, ?string $language = null, ?string $gender = null, string $engine = 'edge'): string
    {
        $timeoutSec = max(30, (int) config('services.tts.timeout', 240));
        
        $directory = storage_path('app/public/tts');
        if (! is_dir($directory)) {
            mkdir($directory, 0777, true);
        }

        $filename = 'tts_'.now()->format('Ymd_His').'_'.Str::lower(Str::random(8)).'.mp3';
        $absolutePath = $directory.DIRECTORY_SEPARATOR.$filename;

        if ($engine === 'kokoro') {
            return $this->synthesizeHuggingFace($text, $language, $gender, $absolutePath, $filename, $timeoutSec);
        }

        return $this->synthesizeEdge($text, $language, $gender, $absolutePath, $filename, $timeoutSec);
    }

    private function synthesizeEdge(string $text, ?string $language, ?string $gender, string $absolutePath, string $filename, int $timeoutSec): string
    {
        $voice = $this->resolveVoice($language, $gender);
        $binary = $this->resolveBinary();

        $result = Process::timeout($timeoutSec)->run([
            $binary,
            '--voice', $voice,
            '--text', $text,
            '--write-media', $absolutePath,
        ]);

        if (! $result->successful()) {
            $error = trim($result->errorOutput() ?: $result->output());
            Log::error('Edge-TTS error: ' . $error);
            throw new RuntimeException('Edge-TTS gagal: ' . $error);
        }

        chmod($absolutePath, 0666);
        return Storage::url('tts/'.$filename);
    }

    private function synthesizeHuggingFace(string $text, ?string $language, ?string $gender, string $absolutePath, string $filename, int $timeoutSec): string
    {
        $token = config('services.huggingface.api_token');

        // Model candidates: Kokoro (multilingual, high quality) → MMS language-specific fallback
        $models = ($language === 'id')
            ? ['hexgrad/Kokoro-82M', 'facebook/mms-tts-ind']
            : ['hexgrad/Kokoro-82M', 'facebook/mms-tts-eng'];

        // Kokoro needs a voice hint via the inputs array; other models just need plain text
        $payloads = [
            'hexgrad/Kokoro-82M' => [
                'inputs' => $text,
                'parameters' => [
                    'voice' => ($language === 'id')
                        ? (($gender === 'male') ? 'im' : 'if')
                        : (($gender === 'male') ? 'am_michael' : 'af_heart'),
                ],
            ],
            'facebook/mms-tts-ind' => ['inputs' => $text],
            'facebook/mms-tts-eng' => ['inputs' => $text],
        ];

        $deadline = time() + $timeoutSec;
        $lastError = 'Unknown HF-TTS error';

        foreach ($models as $model) {
            $url     = "https://router.huggingface.co/hf-inference/models/{$model}";
            $payload = $payloads[$model] ?? ['inputs' => $text];

            while (time() < $deadline) {
                $response = Http::withToken($token)
                    ->timeout(min(60, $deadline - time()))
                    ->accept('audio/flac, audio/wav, audio/mpeg, application/octet-stream')
                    ->post($url, $payload);

                $status = $response->status();

                // Binary audio returned directly
                if ($status === 200) {
                    $body = $response->body();
                    if (strlen($body) > 1024) {
                        // Persist; extension stays .mp3 but content may be flac/wav — player handles it
                        file_put_contents($absolutePath, $body);
                        chmod($absolutePath, 0666);
                        return Storage::url('tts/' . $filename);
                    }
                    // Tiny body = likely JSON error wrapped in 200
                    $lastError = $response->json('error') ?? 'Empty audio body';
                    Log::warning("HF-TTS [{$model}] 200 but empty body: {$lastError}");
                    break; // try next model
                }

                // Model still loading — wait and retry
                if ($status === 503) {
                    $wait = (int) ($response->json('estimated_time') ?? 5);
                    $wait = max(3, min($wait, 20));
                    Log::info("HF-TTS [{$model}] loading, retrying in {$wait}s…");
                    sleep($wait);
                    continue;
                }

                // Rate limit — short back-off then retry current model
                if ($status === 429) {
                    Log::warning("HF-TTS [{$model}] rate limited, waiting 10s…");
                    sleep(10);
                    continue;
                }

                // Any other error: log and try next model
                $lastError = $response->json('error') ?? $response->body();
                Log::warning("HF-TTS [{$model}] HTTP {$status}: {$lastError}");
                break;
            }
        }

        // All HF models failed — fall back to Edge TTS
        Log::error("HF-TTS all models failed: {$lastError} — falling back to Edge TTS");
        return $this->synthesizeEdge($text, $language, $gender, $absolutePath, $filename, $timeoutSec);
    }

    public function languageOptions(): array
    {
        return [
            'id' => 'Indonesia',
            'en' => 'English (US)',
            'ja' => 'Japanese',
            'ko' => 'Korean',
            'es' => 'Spanish',
        ];
    }

    public function genderOptions(): array
    {
        return [
            'female' => 'Female',
            'male' => 'Male',
        ];
    }

    private function resolveVoice(?string $language, ?string $gender): string
    {
        $language = $language ?: 'id';
        $gender = $gender ?: 'female';

        $catalog = $this->voiceCatalog();

        if (isset($catalog[$language][$gender])) {
            return $catalog[$language][$gender];
        }

        return (string) config('services.edge_tts.voice', 'id-ID-GadisNeural');
    }

    private function voiceCatalog(): array
    {
        return [
            'id' => [
                'female' => 'id-ID-GadisNeural',
                'male' => 'id-ID-ArdiNeural',
            ],
            'en' => [
                'female' => 'en-US-JennyNeural',
                'male' => 'en-US-GuyNeural',
            ],
            'ja' => [
                'female' => 'ja-JP-NanamiNeural',
                'male' => 'ja-JP-KeitaNeural',
            ],
            'ko' => [
                'female' => 'ko-KR-SunHiNeural',
                'male' => 'ko-KR-InJoonNeural',
            ],
            'es' => [
                'female' => 'es-ES-ElviraNeural',
                'male' => 'es-ES-AlvaroNeural',
            ],
        ];
    }

    private function resolveBinary(): string
    {
        $configuredBinary = trim((string) config('services.edge_tts.binary', ''));
        $candidates = [
            $configuredBinary,
            'edge-tts',
            base_path('.venv-edge-tts/bin/edge-tts'),
            '/usr/local/bin/edge-tts',
            '/usr/bin/edge-tts',
            '/home/www-data/.local/bin/edge-tts',
        ];

        foreach ($candidates as $candidate) {
            if ($candidate === '') {
                continue;
            }

            $check = Process::timeout(5)->run([$candidate, '--version']);
            if ($check->successful()) {
                return $candidate;
            }
        }

        $find = Process::timeout(5)->run(['which', 'edge-tts']);
        if ($find->successful() && filled($find->output())) {
            return trim($find->output());
        }

        throw new RuntimeException('edge-tts belum terinstall atau tidak ditemukan di PATH.');
    }
}