Fish Speech

S2.1 Pro is a neural text to speech model that converts text into speech across 83 languages with automatic language detection. It supports inline bracket tags for emotion and paralinguistic control, multi-speaker dialogue, and instant voice cloning from a reference audio sample. It streams audio in real time with about 70 to 90 ms time to first audio and outputs mp3, wav, pcm, or opus.

Fish Audio Speech-to-Text is Fish Audio’s transcription model and tool for podcasts, interviews, and conversational recordings. It converts audio to text with speaker labels, timestamps, and inline emotion or paralanguage tags, supports 100+ languages, accepts 24 audio and video formats, and exports transcripts as SRT, VTT, or JSON without requiring code.

Fish Audio S2 is Fish Audio’s latest text-to-speech model designed for natural, emotionally rich speech generation with fine-grained prosody control and native multi-speaker dialogue.

Amazon Nova Sonic is a real-time, speech-first model in the Nova family. It streams speech-to-text, reasons while you’re talking, and speaks back instantly—enabling natural, interruptible voice agents on web and mobile.

CLOVA Voice is NAVER’s neural text-to-speech service for lifelike, multilingual voices. It offers natural prosody, style controls, SSML, real-time streaming, and brand-safe custom voice options—ready for apps, call centers, e-learning, and content creation.

Latency-optimized sibling of Speech-2.8-HD, trading a bit of ultimate fidelity for faster, cheaper generation while keeping multilingual, emotional, voice-cloning strengths.