AI Sound Effects Generation for Mobile App

Implementing AI Sound Effects Generation in a Mobile App

Sound effects by description — separate task from music generation. Need not melody track, but short (0.5–5 sec) specific sound: sword hit, rain noise, button click. ElevenLabs Sound Effects and AudioCraft (EnCodec + AudioGen) — main tools.

ElevenLabs Sound Effects API

Best by quality and integration simplicity. Accepts text description, returns mp3 in 2–8 seconds:

POST https://api.elevenlabs.io/v1/sound-generation
xi-api-key: <key>
Content-Type: application/json

{
  "text": "A heavy metal sword hitting a stone floor with a sharp clang and short reverb",
  "duration_seconds": 2.0,
  "prompt_influence": 0.3
}

prompt_influence from 0 to 1: higher means literal prompt interpretation. For short effects (< 1 sec) set 0.7–0.9.

Response — binary mp3 in response body (not JSON with URL). On mobile:

// iOS: direct binary response download
func generateSoundEffect(description: String, duration: Double) async throws -> Data {
    var request = URLRequest(url: URL(string: "https://api.elevenlabs.io/v1/sound-generation")!)
    request.httpMethod = "POST"
    request.setValue("audio/mpeg", forHTTPHeaderField: "Accept")
    request.setValue("application/json", forHTTPHeaderField: "Content-Type")
    request.setValue(apiKey, forHTTPHeaderField: "xi-api-key")
    request.httpBody = try JSONEncoder().encode(SoundGenRequest(
        text: description, duration_seconds: duration, prompt_influence: 0.4
    ))

    let (data, response) = try await URLSession.shared.data(for: request)
    guard (response as? HTTPURLResponse)?.statusCode == 200 else {
        throw SoundGenError.apiError
    }
    return data // mp3 bytes
}

Generation time small — simple activity indicator sufficient, no complex async polling needed.

Cache generated sounds

Same sound effect user may use multiple times — regenerating each time expensive and slow. Cache by prompt hash + duration:

// Android
class SoundEffectCache(private val cacheDir: File) {
    private fun cacheKey(prompt: String, duration: Double): String =
        "${prompt.hashCode()}_${(duration * 10).toInt()}.mp3"

    fun getCached(prompt: String, duration: Double): File? {
        val file = File(cacheDir, "sfx/${cacheKey(prompt, duration)}")
        return if (file.exists()) file else null
    }

    fun saveToCache(prompt: String, duration: Double, data: ByteArray): File {
        val dir = File(cacheDir, "sfx").also { it.mkdirs() }
        val file = File(dir, cacheKey(prompt, duration))
        file.writeBytes(data)
        return file
    }
}

UI patterns: library vs inline generation

Two UX approaches:

1. Inline — user types description, generates immediately in-context (editor, game)
2. Library — pre-generated effects catalog, user browses and selects

Inline simpler but slower UX. Library requires backend catalog management but instant playback.

Timeline

Basic generation + playback UI — 2–3 days. With cache, library UI, batch generation — 1–2 weeks.