Models
Audio and speech models
The Clipia models that voice text and make music — developers, languages, voices, prices in credits and tips on choosing a model and a voice.
Clipia's audio tools run on models from several developers: Google and ElevenLabs for speech, Suno for music. This page covers what each model does, which languages it supports, what it costs and how to pick the right model and voice. For the modes themselves and step-by-step workflows, see Audio Studio.
All models
| Model | Developer | Purpose | Languages | Price |
|---|---|---|---|---|
| Gemini 3.1 Flash TTS | Fast, controllable text-to-speech | Russian, English | 4 credits / 1,000 characters | |
| Gemini 2.5 Pro TTS | Text-to-speech for the final cut | Russian, English | 4 credits / 1,000 characters | |
| ElevenLabs v4 | ElevenLabs | Expressive speech and dialogue | Russian, English | 7 credits / 1,000 characters |
| ElevenLabs v4 Turbo | ElevenLabs | Fast speech and dialogue | Russian, English | 4 credits / 1,000 characters |
| Suno Music | Suno | Music and songs from a description | — | 12 credits per run |
Prices
Prices on this page come from the service pricing and are refreshed together with the documentation. The exact cost is always shown on the start button before you are charged.
Text-to-speech and music work in Audio Studio: voice-over at /create-audio, music at /create-audio/ai-song-generator. Transcription, speech translation, recording cleanup and personal voices are still in development.
Speech models: which one to choose
All four speech models read Russian and English. The language is detected from the text: if it contains Cyrillic, it is read in Russian, otherwise in English.
Gemini 3.1 Flash TTS
The studio default. Fast and controllable: 30 voices plus style, pace, accent and narrator profile. A good choice for short videos, ads and drafts.
Gemini 2.5 Pro TTS
The same 30 voices and direction controls, with finer control over delivery. Use it for the final voice-over once a Flash draft is approved.
ElevenLabs v4
The most expressive delivery: emotion, cues such as whispering and laughing, multi-voice dialogue, up to 10,000 characters per request.
ElevenLabs v4 Turbo
The same 21 voices and dialogue, but faster and almost half the price of v4. Suits large volumes where speed matters.
| Task | Recommended model |
|---|---|
| Voice-over for a short video or ad | Gemini 3.1 Flash TTS |
| Final voice-over with precise direction | Gemini 2.5 Pro TTS |
| Dialogue between several characters | ElevenLabs v4 or v4 Turbo |
| Large volumes of text at the lowest price | ElevenLabs v4 Turbo or Gemini |
Text per request: Gemini accepts 10 to 500 characters, ElevenLabs 10 to 10,000. On the Audio page long scripts are split into fragments automatically, so the per-request limit does not get in the way.
Price example: the price is per 1,000 characters, so a 10,000-character text costs ten times as much: with ElevenLabs v4 that is 10 × 7 credits.
How to choose a voice
- Listen to samples. Every voice has a sample, and listening costs no credits. For Gemini you can switch the sample language between Russian and English, although the model itself takes the language from your text.
- Filter. The voice library can be filtered by gender (male, female, neutral) and character, or searched by name.
- Match the genre. For ads, pick an energetic, bright voice; for lessons and audiobooks, a calm, even voice with clear diction; for news, a confident announcer.
- Test on your own text. Voice a short passage of one or two sentences before running the whole script.
Voices by model:
- Gemini — 30 voices: 16 male and 14 female. Delivery is shaped by direction controls: style, pace, accent, variability, narrator profile and scene.
- ElevenLabs v4 and v4 Turbo — 21 voices, such as George, Bella, Brian, Sarah, Alice, Daniel, Laura and Matilda. Delivery can be Expressive, Balanced or Even, plus a voice similarity setting.
Music
Suno Music (Suno) creates a track from a description: an instrumental backing track or a song with vocals. The model chooses the track length. A run costs 12 credits and includes every variant the model returns. Descriptions work best when they name the genre, mood, tempo, instruments and scene: "calm piano, warm strings, soft rhythm, no vocals".
See also
Video models
Passports of Clipia video models — Veo, Kling, Seedance, Wan, HappyHorse, Hailuo, Grok and more: modes, duration, quality, sound, prices and picks by task.
Quickstart
Run your first image generation with the Clipia API in four steps, from creating a key to fetching the result, with cURL, TypeScript and Python examples.