Remy Reference/Models/Text to speech/ElevenLabs TTS

ElevenLabs TTS

ElevenLabs TTS, a text-to-speech model from ElevenLabs.

speechModelOverride: { model: "elevenlabs-tts" }

EElevenLabs·Released 2023

Pricing

List price: $0.04-$0.08 / 1k characters
Per 1M characters: $80.00
Per 1M characters: $80.00
Per 1M characters: $80.00
Per 1M characters: $40.00
Per 1M characters: $40.00
Per 1M characters: $40.00
Per 1M characters: $40.00
Per 1M characters: $80.00
Per 1M characters: $40.00
Per 1M characters: $80.00
Per 1M characters: $40.00
Per 1M characters: $240.00
Per 1M characters: $0

Specifications

Context window: 10,000 tokens
Released: 2023
Capabilities: Low latency

Parameters

Model options, passed in speechModelOverride.config.

Parameter	Type	Default	Range / options
`$mode`	Select	single	singledialogue
`model` Dialogue runs on Eleven v4 or v3. v4 and v3 take audio tags in the text, like [laughs] or [whispers].	Select	eleven_v4	eleven_v4eleven_v4_turboeleven_v3eleven_v3_conversationaleleven_multilingual_v2eleven_flash_v2_5+1
`voice` An ElevenLabs voice id. Browse and preview voices at https://elevenlabs.io/app/voice-library, or leave unset for the default (Elara).	Voice	WQP7cQUF5aAS6Axh5yaa	—
`speaker1_name` The name speaker 1's lines start with in the script ("Name: line").	Text	—	—
`speaker1_voice`	Voice	—	—
`speaker2_name` The name speaker 2's lines start with in the script ("Name: line").	Text	—	—
`speaker2_voice`	Voice	—	—
`speaker3_name` The name speaker 3's lines start with in the script ("Name: line").	Text	—	—
`speaker3_voice`	Voice	—	—
`speaker4_name` The name speaker 4's lines start with in the script ("Name: line").	Text	—	—
`speaker4_voice`	Voice	—	—
`stability` How steady the delivery stays. Lower is more expressive, higher more consistent.	Number	—	0–1, step 0.05
`similarity` How closely the speech holds to the voice. Not used by Eleven v3.	Number	—	0–1, step 0.05
`style` How much of the voice's own style to exaggerate. Multilingual v2 only.	Number	—	0–1, step 0.05
`speaker_boost` Strengthens likeness to the original speaker. Multilingual v2 and v3 Conversational.	Boolean	—
`speed` Speaking speed, 1.0 normal. Multilingual v2 and Flash only; v4 and v3 take pacing from the text.	Number	—	0.7–1.2, step 0.05
`output_format` 192 kbps MP3 needs an ElevenLabs Creator plan or above, WAV a Pro plan, on the workspace's own key.	Select	mp3_44100_128	mp3_44100_128mp3_44100_192wav_44100opus_48000_128
`language_code` Optional ISO 639-1 code (e.g. "es") to hold the model to one language. Not used by Multilingual v2.	Text	—	—
`apply_text_normalization` Whether numbers, dates and abbreviations are spelled out as spoken.	Select	—	autoonoff
`pronunciation_dictionary_id` Optional id of a pronunciation dictionary in the ElevenLabs account, for names and terms it should say a particular way.	Text	—	—
`seed` Optional, for repeatable output. ElevenLabs does not guarantee it.	Seed	—	—

Use this model

await mindstudio.textToSpeech({
  text: "...",
  speechModelOverride: {
    model: "elevenlabs-tts",
    config: {
      model: "eleven_v4",
      output_format: "mp3_44100_128",
    },
  },
});