Limko

Voice · inworld/realtime-tts-2

Inworld TTS-2 — lifelike speech for characters, games and dialogue

Inworld TTS-2 is the most expressive model in the Inworld line, built for live conversation: its latency matches that of a real exchange. You set the delivery with a direction in square brackets before the text — a plain phrase, the way a director prompts an actor: [say excitedly], [whisper]. In limko you get four voices, 14 languages including Russian, auto-detection and smoothly adjustable speech speed, and voice-over costs 7.5 ₽ per 2,000 characters.

from 7.5 ₽ per voice-over up to 2,000 chars

What Inworld TTS-2 gives you

  • A direction before the line sets the emotion: [say excitedly] for excitement, [whisper] for a whisper.
  • Eight ready-made limko delivery styles — for when you would rather not write directions by hand.
  • Four voices for different roles: Ashley — warm, Darlene — soft, Alex — clear, Dennis — confident.
  • Speech speed from 0.5 to 1.5 changes smoothly, and the model itself adjusts the pacing.
  • 14 languages on the limko list, including Russian, plus auto-detection; the model itself has more than a hundred.
  • Up to 1,900 characters per run; a short line is usually ready in a few seconds.

Directions: how to set the tone of a line

A direction goes in square brackets before the text — a short phrase in English, like a note in a script. The line itself stays in your language.

What to writeHow it will sound
[say excitedly] We found a way out, let’s run. excited, with rising energy
[whisper] Quiet, he’s very close. in a whisper
[say sadly] I thought you would never come back. sad
[say calmly] First, open the settings. calm and even

How to voice a scene by roles

  1. Split the scene into lines

    Each line is a separate run of up to 1,900 characters. That way you can re-record one phrase without touching the rest.

  2. Assign voices to characters

    Ashley, Darlene, Alex and Dennis sound different — give each character their own voice for the whole story.

  3. Add directions and pace

    Before the line, a tone in brackets or a ready-made delivery style; speed from 0.5 to 1.5: slower for drama, faster for banter.

  4. Assemble the track

    Put the lines in a project folder and merge them into one recording right in limko — merging and trimming are free.

Price, languages and neighboring models

Voice-over costs 7.5 ₽ per 2,000 characters, and a character line shorter than 300 characters is billed as 300 — a small fraction of the price. Credits are charged only for a successful result.

The list has 14 languages, Russian among them, plus auto-detection. For Russian text there is a “Prepare for voice-over” button: for a small extra charge the service adds stress marks and the letter ё.

For a long monologue Grok TTS is more convenient: up to 14,500 characters per run. Gemini 3.8 TTS gives expressive delivery by instruction and 17 languages, MiniMax Speech Turbo gives fast, expressive voice-over of up to 9,500 characters, and a character voice based on your own sample can be made only in Qwen3-TTS.

Frequently asked about Inworld TTS-2

How do I set the emotion in Inworld TTS-2?

Write a direction in square brackets before the text: [say excitedly] for excitement, [whisper] for a whisper. Or choose one of limko’s eight ready-made delivery styles.

How is Inworld different from Gemini 3.8 TTS?

Inworld is steered by directions before the line, while Gemini 3.8 TTS takes its delivery from the style you pick and costs less: 5.1 ₽ versus 7.5 ₽ per 2,000 characters. Gemini 3.8 TTS also has more voices in limko — 12 versus four — and 17 languages on its list.

How much text can I voice at once?

Up to 1,900 characters — enough for a line, a scene or a short monologue. For long texts, Grok TTS fits: up to 14,500 characters per run.

Is Russian available?

Yes, Russian is on the list of 14 languages, and you can turn on auto-detection. The model itself has more than a hundred languages; limko offers a selection.

Can I voice a character in my own voice?

Not in Inworld TTS-2 — it has four preset voices. In limko, only Qwen3-TTS clones your own voice.