Audio

Open-weight speech recognition supporting 50+ languages. Handles accents, noise, and technical language.

AI sound effect generation from text descriptions.

ElevenLabs speech recognition and transcription service.

Ultra-low latency TTS at 75ms TTFA. Best for real-time conversational voice agents.

MiniMax text-to-speech with HD and turbo variants. Supports voice cloning.

Ultra-realistic narration in 70+ languages with thousands of voice presets.

Latest OpenAI transcription with lower error rates than Whisper. Recommended over Whisper for API use.

OpenAI's fast text-to-speech model optimized for real-time use.

Higher quality TTS with improved naturalness and pronunciation accuracy.

Low-latency TTS at 90ms optimized TTFB for voice agent production use.

Google's latest speech recognition model with improved accuracy across 100+ languages.

Google's text-to-speech service supporting 75+ languages with WaveNet and Neural2 voices.

Multilingual TTS supporting 80+ languages with voice cloning capabilities.

Real-time STT specialist with sub-300ms latency, streaming WebSocket API, and domain-specific vocabulary.

Qwen 3 text-to-speech model with voice cloning support.

Speech-language model with multilingual streaming, safety guardrails, and LLM gateway integration.

Fastest production TTS at ~40ms TTFA. 15 languages, ~130 voices. One fifth the cost of ElevenLabs.

Benchmark-leading accuracy at ~8.4% WER with 30% fewer hallucinations than Whisper. Full audio intelligence suite.

Open-weight 4B TTS model. 9 languages, ~90ms TTFA, voice cloning from 3s reference. CC BY NC 4.0.