Text-to-Speech¶

The Text-to-Speech service converts text into natural-sounding speech.

v0.13.0 API Change

As of v0.13.0, TTS methods are accessed via client.TTS() instead of client.TextToSpeech().

Basic Usage¶

Simple Generation¶

audio, err := client.TTS().Simple(ctx, voiceID, "Your text here")
if err != nil {
    log.Fatal(err)
}

// audio is an io.Reader - save or stream it
f, _ := os.Create("output.mp3")
io.Copy(f, audio)

Full Control¶

import "github.com/plexusone/elevenlabs-go/tts"

resp, err := client.TTS().Generate(ctx, &tts.Request{
    VoiceID: "21m00Tcm4TlvDq8ikWAM",
    Text:    "Hello world!",
    ModelID: "eleven_multilingual_v2",
    VoiceSettings: &elevenlabs.VoiceSettings{
        Stability:       0.5,
        SimilarityBoost: 0.75,
        Style:           0.0,
        SpeakerBoost:    true,
    },
    OutputFormat: "mp3_44100_128",
})

Voice Settings¶

Setting	Range	Description
`Stability`	0.0-1.0	Higher = more consistent, lower = more expressive
`SimilarityBoost`	0.0-1.0	Higher = closer to original voice
`Style`	0.0-1.0	Style exaggeration (use sparingly)
`SpeakerBoost`	bool	Enhance speaker similarity
`Speed`	0.7-1.2	Speech speed multiplier

Default Settings¶

settings := elevenlabs.DefaultVoiceSettings()
// Stability: 0.5, SimilarityBoost: 0.75, Style: 0, SpeakerBoost: true

Output Formats¶

Format	Description
`mp3_44100_128`	MP3, 44.1kHz, 128kbps (default)
`mp3_44100_192`	MP3, 44.1kHz, 192kbps
`pcm_16000`	PCM, 16kHz
`pcm_22050`	PCM, 22.05kHz
`pcm_24000`	PCM, 24kHz
`pcm_44100`	PCM, 44.1kHz
`ulaw_8000`	u-law, 8kHz

Models¶

Model ID	Best For
`eleven_multilingual_v2`	Multiple languages, highest quality
`eleven_monolingual_v1`	English only, fast
`eleven_turbo_v2`	Low latency applications
`eleven_turbo_v2_5`	Lowest latency

Streaming¶

For real-time applications, use streaming:

resp, err := client.TTS().GenerateStream(ctx, &tts.Request{
    VoiceID: voiceID,
    Text:    "Long text to stream...",
})
if err != nil {
    log.Fatal(err)
}

// Stream audio chunks as they arrive
for chunk := range resp.Chunks {
    // Process chunk
}

Error Handling¶

audio, err := client.TTS().Simple(ctx, voiceID, text)
if err != nil {
    if elevenlabs.IsRateLimitError(err) {
        // Wait and retry
        time.Sleep(time.Minute)
    } else if elevenlabs.IsUnauthorizedError(err) {
        // Check API key
    }
    return err
}

Best Practices¶

Reuse the client - Create one client and reuse it
Check character limits - Monitor CharactersRemaining() before generation
Use appropriate models - turbo for speed, multilingual_v2 for quality
Batch when possible - Combine short texts to reduce API calls