# Audio AI Models for Game Studios

> Compare 45 audio and voice models on Layer — quality, capabilities, and per-generation pricing side by side. Generate production-ready assets for your game.

## Audio models (45)

- [ElevenLabs TTS V4 Turbo](https://layer.ai/models/elevenlabs-tts-v4-turbo) ([md](https://layer.ai/models/elevenlabs-tts-v4-turbo.md)) (aka Eleven v4 Turbo) by ElevenLabs: Lower-latency Eleven v4 Turbo speech with audio tags, stability, and similarity control.
- [ElevenLabs TTS V4](https://layer.ai/models/elevenlabs-tts-v4) ([md](https://layer.ai/models/elevenlabs-tts-v4.md)) (aka Eleven v4) by ElevenLabs: ElevenLabs' expressive v4 speech model with audio tags, stability, and similarity control.
- [Gemini 3.8 Flash TTS](https://layer.ai/models/gemini-3.8-flash-tts) ([md](https://layer.ai/models/gemini-3.8-flash-tts.md)) by Google: Google's expressive text-to-speech model with style direction and inline vocal events.
- [Qwen 3 TTS](https://layer.ai/models/qwen-3-tts) ([md](https://layer.ai/models/qwen-3-tts.md)) by Qwen 3 Tts: Multilingual TTS with zero-shot voice cloning and prompt-based voice design.
- [Gemini 3.8 Flash Lite TTS](https://layer.ai/models/gemini-3.8-flash-lite-tts) ([md](https://layer.ai/models/gemini-3.8-flash-lite-tts.md)) by Google: The lighter, lower-cost tier of Google's Gemini 3.8 text-to-speech model.
- [Gemini 3.1 Flash TTS](https://layer.ai/models/gemini-3.1-flash-tts) ([md](https://layer.ai/models/gemini-3.1-flash-tts.md)) by Google: Google's most controllable TTS model with 200+ audio tags for vocal style and delivery.
- [ElevenLabs TTS V3](https://layer.ai/models/elevenlabs-tts-v3) ([md](https://layer.ai/models/elevenlabs-tts-v3.md)) (aka Eleven v3) by ElevenLabs: ElevenLabs' most expressive TTS model with inline audio tags for emotion and delivery.
- [xAI TTS V1](https://layer.ai/models/xai-tts-v1) ([md](https://layer.ai/models/xai-tts-v1.md)) by xAI: xAI's text-to-speech model with speech tags for expressive delivery in 20+ languages.
- [ElevenLabs Multilingual V2](https://layer.ai/models/elevenlabs-multilingual-v2) ([md](https://layer.ai/models/elevenlabs-multilingual-v2.md)) by ElevenLabs: Multilingual text-to-speech with natural voice selection and stability controls.
- [ElevenLabs Sound Effects](https://layer.ai/models/elevenlabs-sound-effects) ([md](https://layer.ai/models/elevenlabs-sound-effects.md)) by ElevenLabs: Generate custom sound effects from text descriptions with duration and prompt control.
- [Sonilo Sound Effects 1.1](https://layer.ai/models/sonilo-v1-1-text-to-sound-effects) ([md](https://layer.ai/models/sonilo-v1-1-text-to-sound-effects.md)) by Sonilo: High-quality, commercial-use-safe sound effects from text with exact duration control.
- [Mirelo SFX 2.0 Video to Audio](https://layer.ai/models/mirelo-sfx-2-video-to-audio) ([md](https://layer.ai/models/mirelo-sfx-2-video-to-audio.md)) by Mirelo: Synchronized sound-effect stems timed to an input video.
- [Mirelo SFX 2.0](https://layer.ai/models/mirelo-sfx-2) ([md](https://layer.ai/models/mirelo-sfx-2.md)) by Mirelo: Text-to-sound-effects with duration, negative prompt, and format control.
- [Mureka Podcast 1.0](https://layer.ai/models/mureka-podcast) ([md](https://layer.ai/models/mureka-podcast.md)) by Mureka: Spoken audio from a script, split across built-in voices.
- [Mureka 9.5 Instrumental](https://layer.ai/models/mureka-9-5-instrumental) ([md](https://layer.ai/models/mureka-9-5-instrumental.md)) by Mureka: Vocal-free music from a text description.
- [Mureka 9.5 Song](https://layer.ai/models/mureka-9-5-song) ([md](https://layer.ai/models/mureka-9-5-song.md)) by Mureka: Complete songs with vocals from a description or supplied lyrics.
- [Stable Audio 3 Small SFX](https://layer.ai/models/stable-audio-3-small-sfx) ([md](https://layer.ai/models/stable-audio-3-small-sfx.md)) by Stability AI: Fast sound effects with restyle, extend and inpaint.
- [Stable Audio 3 Small Music](https://layer.ai/models/stable-audio-3-small-music) ([md](https://layer.ai/models/stable-audio-3-small-music.md)) by Stability AI: Fast, low-cost music up to two minutes, with restyle, extend and inpaint.
- [Stable Audio 3 Medium](https://layer.ai/models/stable-audio-3-medium) ([md](https://layer.ai/models/stable-audio-3-medium.md)) by Stability AI: Full songs and long scores up to six minutes, with restyle, extend and inpaint.
- [VEED Clean Audio 1.0](https://layer.ai/models/veed-clean-audio) ([md](https://layer.ai/models/veed-clean-audio.md)) by VEED: Remove background noise from speech while keeping quiet and distant words intact.
- [Lyria 3.5](https://layer.ai/models/lyria-3-5) ([md](https://layer.ai/models/lyria-3-5.md)) by Google: Google DeepMind's latest music model, writing full-length songs with vocals from one prompt.
- [MiniMax Music 3](https://layer.ai/models/minimax-music-3) ([md](https://layer.ai/models/minimax-music-3.md)) by MiniMax: High-performance MiniMax music model for complete songs up to five minutes with structure tags and seed control.
- [Sonilo Video to Music 1.1](https://layer.ai/models/sonilo-v1-1-video-to-music) ([md](https://layer.ai/models/sonilo-v1-1-video-to-music.md)) by Sonilo: Frame-synced, licensed music scored from video pacing, mood, and timing.
- [Sonilo Video to Sound Effects 1.1](https://layer.ai/models/sonilo-v1-1-video-to-sound-effects) ([md](https://layer.ai/models/sonilo-v1-1-video-to-sound-effects.md)) by Sonilo: Synchronized, royalty-free sound effects timed to actions visible in a video.
- [Sonilo Music 1.1](https://layer.ai/models/sonilo-v1-1-text-to-music) ([md](https://layer.ai/models/sonilo-v1-1-text-to-music.md)) by Sonilo: Licensed, commercial-use-safe music from a text prompt with exact duration control.
- [Seed Audio 1.0](https://layer.ai/models/seed-audio-1) ([md](https://layer.ai/models/seed-audio-1.md)) by ByteDance: Bytedance Seed Audio TTS with preset voices and zero-shot voice cloning.
- [Stable Audio](https://layer.ai/models/stable-audio) ([md](https://layer.ai/models/stable-audio.md)) by Stability AI: Open-source text-to-audio model for sound effects, field recordings, and instrument samples.
- [Minimax Music V2.6](https://layer.ai/models/minimax-music-v26) ([md](https://layer.ai/models/minimax-music-v26.md)) by MiniMax: Latest MiniMax music model with native audio-visual generation capabilities.
- [Minimax Music V2.5](https://layer.ai/models/minimax-music-v25) ([md](https://layer.ai/models/minimax-music-v25.md)) by MiniMax: Updated MiniMax music model with improved vocal quality and arrangement control.
- [Stable Audio 2.5](https://layer.ai/models/stable-audio-25) ([md](https://layer.ai/models/stable-audio-25.md)) by Stability AI: Enterprise-grade audio generation with multi-part compositions and audio inpainting.
- [CassetteAI Sound Effects](https://layer.ai/models/cassetteai-sfx) ([md](https://layer.ai/models/cassetteai-sfx.md)) by CassetteAI: Real-time sound effects generation in under 1 second, up to 30 seconds long.
- [SAM Audio Separate](https://layer.ai/models/sam-audio-separate) ([md](https://layer.ai/models/sam-audio-separate.md)) by Meta: Foundation model for audio source separation using text, visual, or temporal prompts.
- [Inworld TTS 1.5 Max](https://layer.ai/models/inworld-tts-1.5-max) ([md](https://layer.ai/models/inworld-tts-1.5-max.md)) by Inworld AI: Inworld's premium TTS model optimized for expressive game character voices.
- [Lyria 3](https://layer.ai/models/lyria-3) ([md](https://layer.ai/models/lyria-3.md)) by Google: Google DeepMind's next-generation music model with enhanced composition and vocal quality.
- [Lyria 3 Pro](https://layer.ai/models/lyria-3-pro) ([md](https://layer.ai/models/lyria-3-pro.md)) by Google: Premium tier of Google's Lyria 3 with highest fidelity and compositional complexity.
- [ElevenLabs Voice Changer](https://layer.ai/models/elevenlabs-voice-changer) ([md](https://layer.ai/models/elevenlabs-voice-changer.md)) by ElevenLabs: Transform the voice in an audio recording to a different target voice.
- [CassetteAI Music](https://layer.ai/models/cassetteai-music) ([md](https://layer.ai/models/cassetteai-music.md)) by CassetteAI: Ultra-fast music generation producing 3-minute tracks in under 10 seconds at 44.1 kHz.
- [Lyria 2](https://layer.ai/models/lyria-2) ([md](https://layer.ai/models/lyria-2.md)) by Google: Google DeepMind's high-fidelity 48 kHz music model with fine-grained creative control.
- [DeepFilterNet3](https://layer.ai/models/deepfilternet3) ([md](https://layer.ai/models/deepfilternet3.md)) by Rikorose: Real-time speech enhancement and noise suppression at 48 kHz full-band audio.
- [Kling Video to Audio](https://layer.ai/models/kling-video-to-audio) ([md](https://layer.ai/models/kling-video-to-audio.md)) by Kling: Generate synchronized sound effects, dialogue, and ambient audio from video content.
- [ElevenLabs Audio Isolation](https://layer.ai/models/elevenlabs-audio-isolation) ([md](https://layer.ai/models/elevenlabs-audio-isolation.md)) by ElevenLabs: Isolate clean voice from noisy recordings by removing background noise and music.
- [Mirelo SFX 1.6](https://layer.ai/models/mirelo-sfx) ([md](https://layer.ai/models/mirelo-sfx.md)) by Mirelo: Sound effects generation and editing with text-to-audio and audio inpainting.
- [ElevenLabs Music](https://layer.ai/models/elevenlabs-music) ([md](https://layer.ai/models/elevenlabs-music.md)) by ElevenLabs: Premium AI music generation from ElevenLabs with composition planning and section control.
- [ACE-Step](https://layer.ai/models/ace-step) ([md](https://layer.ai/models/ace-step.md)) by ACE Studio: Fast open-source music generation with lyrics alignment, remixing, and audio editing.
- [Gemini TTS](https://layer.ai/models/gemini-tts) ([md](https://layer.ai/models/gemini-tts.md)) by Google: Google's text-to-speech model with multi-speaker support and natural expressiveness.

## Frequently Asked Questions

### Which is the best AI audio model?

ElevenLabs TTS V4 Turbo by ElevenLabs leads Layer's audio models with an Artificial Analysis Arena Elo of 1330, ahead of 8 other scored audio models.

### What are the top audio models on Layer?

The top audio models by Elo rating are: 1. ElevenLabs TTS V4 Turbo (Elo 1330), 2. ElevenLabs TTS V4 (Elo 1329), 3. Gemini 3.8 Flash TTS (Elo 1275), 4. Qwen 3 TTS (Elo 1265), 5. Gemini 3.8 Flash Lite TTS (Elo 1248). Ratings come from blind votes in the Artificial Analysis Arena.

### How are audio models ranked on this page?

Audio models are ranked by their Artificial Analysis Arena Elo: people compare two generations made from the same prompt without being told which model made which, and a model preferred more often scores higher. An Elo is a standing inside one modality, so models are only ever ranked against others of their own kind. A model with no arena score yet sits below the scored ones, ordered by how much it is used on Layer and how recently it was released.

### How many audio models does Layer have?

Layer gives you 45 audio and voice models from 17 providers — ElevenLabs, Google and Stability AI among them — in one workspace, on one balance, behind one API.

### Which audio models can I train on my own artwork?

3 of them can be fine-tuned on a studio's own art, Qwen 3 TTS among them. Upload a reference set, train a style, and every generation from it carries your look rather than the model's default one.

### How much do audio models cost?

Nothing is charged per seat. Generating on Layer spends Creative Units, each model has its own rate, and you pay only for what you generate — every model page states its current price, and the pricing page lists them all.

---

[View original](https://layer.ai/models/audio) | [All models](https://layer.ai/models)