Best Audio

MiniMax Music 2.6

MiniMax Music 2.6 is MiniMax’s latest music generation model for full vocal songs and instrumentals from text prompts. It supports natural-language prompts or detailed production-style instructions, follows specified BPM and key with high reliability, and exposes fine-grained song structure control through section tags. The same Music API also supports instrumental generation, lyrics-assisted workflows, and synchronous or streaming delivery.

MiniMax Music Cover

MiniMax Music Cover is MiniMax’s song-to-song transformation model for reimagining an existing track in a new style. It preserves the original vocal melody while changing voice timbre, instrumentation, genre, and arrangement through a text prompt. It supports one-step generation from reference audio or a two-step workflow with preprocessing and optional lyric editing.

ACE-Step v1.5 XL Base is the 4B DiT variant of ACE-Step 1.5 for high-quality music generation and editing. It supports text-to-music, cover generation, repaint, extract, lego, and complete workflows, uses 50 inference steps with CFG, and is designed for longer-form audio generation up to 10 minutes with broad multilingual prompt support.

ACE-Step v1.5 XL Turbo is the accelerated 4B DiT variant of ACE-Step 1.5. It is optimized for faster music generation with 8-step distilled inference while retaining the higher-capacity XL architecture. It supports text-to-music, cover generation, and repaint workflows, making it suitable for rapid iteration when the 2B turbo model is not enough in audio quality.

xAI Text-to-Speech

xAI Text-to-Speech converts text into natural-sounding spoken audio with a single API call. It offers five expressive voices (Eve, Ara, Leo, Rex, and Sal), inline speech tags for fine-grained control over pauses, laughter, whispers, and emphasis, and supports over 20 auto-detected languages.

MiniMax Speech 2.8

MiniMax Speech 2.8 is an advanced text-to-speech model that turns text into natural, expressive audio in multiple languages. It delivers broadcast-ready speech with rich prosody, emotional control, and a diverse voice library. The model supports up to large input lengths and can be used for voiceovers, narration, accessibility tools, and interactive voice applications.

Inworld TTS-1.5 Max

Inworld TTS-1.5 Max is a high-fidelity text-to-speech model engineered for expressive voice synthesis with rich prosody, nuanced emotional range, and broadcast-ready audio quality. It supports a wide set of languages and delivers more natural pronunciation and expressive variation suitable for narration, content creation, and immersive character voices. The Max variant prioritizes audio quality and expressiveness while still supporting responsive generation.

Inworld TTS-1.5 Mini

Inworld TTS-1.5 Mini is a lightweight text-to-speech model designed for real-time voice experiences with ultra-low latency and efficient performance. It delivers natural, expressive audio suitable for interactive agents, voice assistants, and conversational applications where responsiveness is critical. The Mini variant balances speed and quality, enabling responsive speech output even under constrained compute conditions.

Qwen3-TTS 1.7B Base

Qwen3-TTS 1.7B Base is the foundation text-to-speech model from Alibaba's Qwen3-TTS family. It generates human-like speech across 10+ languages including Chinese, English, Japanese, Korean, and European languages. It supports voice cloning from a 3-second audio sample and achieves latency as low as 97ms for real-time applications.

Qwen3-TTS 1.7B CustomVoice

Qwen3-TTS 1.7B CustomVoice is a text-to-speech model from Alibaba that offers nine premium preset timbres across various combinations of gender, age, language, and dialect. It provides precise style control over target voices through user instructions, supports voice cloning from a 3-second sample, and generates speech in 10+ languages with latency as low as 97ms.

Qwen3-TTS 1.7B VoiceDesign

Qwen3-TTS 1.7B VoiceDesign is a text-to-speech model from Alibaba that creates custom voices from natural language descriptions specifying emotion, tone, and prosody. It supports voice cloning from a 3-second audio sample, generates speech in 10+ languages including Chinese, English, Japanese, Korean, and European languages, and achieves latency as low as 97ms.

ACE-Step v1.5 Base is an open-source music generation foundation model built on a hybrid LLM planner and Diffusion Transformer architecture. It generates full tracks from text prompts with support for voice cloning, lyric editing, remixing, cover generation, and compositions up to 10 minutes. It supports over 50 languages and runs on consumer hardware with under 4GB VRAM.

Dia2 2B is a 2 billion parameter streaming text-to-speech model from Nari Labs designed for real-time conversational AI. It begins generating audio immediately from partial text input, supports multi-speaker dialogue via speaker tags, voice cloning from a few seconds of reference audio, and non-verbal cues like laughter, sighs, and coughs. Released under Apache 2.0 for commercial use.

lipsync-2-pro

lipsync-2-pro extends lipsync-2 with diffusion-based enhancement for studio-grade lip synchronization. It preserves fine facial details such as teeth, facial hair, and micro-expressions while supporting high-resolution output suitable for professional post-production workflows.

Eleven Music v1

Eleven Music v1 is a text to music model for high quality multilingual tracks. Control structure, genre, and style at section level. Generate instrumentals or vocal songs from natural language prompts. Integrate through API for automated soundtrack and content workflows.

Eleven v3

Eleven v3 is a premium text to speech model for production audio. It supports 70+ languages with studio grade quality and precise expressive control using inline audio tags. Ideal for narration, podcasts, dialogue, audiobooks, and game voiceover where stable prosody matters.

Eleven Monolingual v1

Eleven Monolingual v1 is an English only text to speech model from ElevenLabs. It focuses on simple natural delivery and stable output. Ideal for lightweight applications, legacy integrations, or projects that need predictable English voice synthesis with low complexity.

Eleven Flash v2.5

Eleven Flash v2.5 is a real time text to speech model for voice agents and interactive apps. It delivers natural speech in about 75 ms latency across 32 languages. Use it for low latency conversational AI, games, live tools, and large scale TTS workloads.

Eleven Multilingual v2

Eleven Multilingual v2 is a high fidelity multilingual text to speech model for 29 languages. It supports expressive prosody with emotional nuance. Ideal for audiobooks, localization pipelines, customer support and international applications that require natural neural voices.

Eleven Multilingual v1

Eleven Multilingual v1 is an early multilingual text to speech model from ElevenLabs. It converts text to natural speech across major languages. It suits legacy integrations, experimentation, and non critical production flows that do not need the quality of v2.

Eleven Flash v2

Eleven Flash v2 is an earlier English speech model that delivers very low latency and clear audio. It is built for live streaming use cases. It also fits real time gaming and interactive tools where rapid voice feedback is critical.

Eleven Turbo v2.5

Eleven Turbo v2.5 delivers fast text to speech for production apps. It targets low latency flows with rich voice quality in 32 languages. Use it to power interactive agents, games, and voice enabled tools that need natural speech with rapid response.

Eleven Turbo v2

Eleven Turbo v2 is an English text to speech model tuned for low latency and low cost. It generates smooth natural speech for chatbots, IVR flows, and automated announcements. Ideal for production systems that need rapid responses and predictable pricing.