makefx.app
AI voice generation

Eleven v4 AI voice generator

Create expressive Eleven v4 speech with audio tags, selectable voices and clear per-generation credit estimates. Compare standard and Turbo models.

Create speech with Eleven v4 in MakeFX. Write or paste the line, choose a voice available to the paying account, add delivery cues such as [whispering] or [excited], and create an audio asset. The prompt, selected model, voice settings and generated word timings stay with the asset's recipe.

Choose Eleven v4 or Turbo

  • Eleven v4 is the default for expressive narration, character dialogue, audiobooks and voiceovers where delivery matters.
  • Eleven v4 Turbo uses the same v4 generation family tuned for lower latency, for faster iterations and responsive voice experiences.

Both models support more than 90 languages and expose Stability and Similarity controls. The provider accepts up to 10,000 characters, while MakeFX's request limit is 8,000. Audio tags are part of the text prompt. Speed and SSML controls are not available on v4. Read ElevenLabs' Eleven v4 guide for supported behavior and language details.

What it costs

MakeFX quotes each request from its selected model, text length, billing mode and minimum fee. The catalog's standard reference rates are $0.08 per 1,000 characters for Eleven v4 and $0.04 for v4 Turbo, before MakeFX's 25% managed markup and the per-asset minimum fee. Temporary provider promotions are not included in those standard rates. Check the quote before creating; your account's active ElevenLabs key changes the billing calculation.

See Make speech, music and sound for writing prompts and working with word timings, or compare Eleven v4 with Eleven v3.

Direct the performance in the prompt

Put spoken words and short delivery directions in the same prompt. For example, add [quiet, measured narration] before a sentence that should sound restrained, or [laughing] after a character's line. Split a long script into separate assets when you need to edit or time each passage independently.

Voice choice and model choice are separate: v4 may reproduce a source voice more faithfully than v3, and cross-language generations can use the target language's native accent. Test a short passage with the selected voice before building a long narration. Keep approved speech and rejected takes as separate assets so the edit remains reversible.