AI tools for Speech AI

Ming-3.0-TTS is an open-weight speech-generation family for narration, audio stories, podcasts, and conversational products. A lightweight dense model and a mixture-of-experts v

Dessa is Block's full-duplex conversation model, trained directly on human speech to hold expressive real-time conversations. It learns from audio without a text-first transcrip

Drama 3 is Fish Audio's expressive voice-generation model for directing how a line is performed, not just which words are spoken. Natural-language instructions and inline contro

Underdog Bark 3.0 is a compact English speech recognition family for private dictation and meeting transcription. The model runs locally on a Mac, with small downloadable varian

Eleven v4 is ElevenLabs' expressive speech-model family, including a Turbo variant for real-time voice applications. The model interprets script context and inline audio tags to

Whistle is Cactus's compact speech-recognition model for phones, wearables, and other CPU-based devices. A 16.9 MB quantized model runs in the same C++ engine as Needle and can

Phonon-2 is a compact open English speech-recognition model for local transcription. Its encoder stores weights in five learned levels at approximately 2.1 bits, compressing a P

Gemini 3.8 TTS is Google's expressive speech-generation family, including Flash TTS and Flash-Lite TTS. Natural-language directions guide character voices and line-by-line deliv

AuK is a speech foundation model that generates and edits audio through natural-language instructions. It combines speech synthesis, spoken-content changes, enhancement, separat

Sophea AI Meet is a meeting intelligence platform that records conversations and turns them into searchable transcripts, summaries, decisions, and follow-up tasks. Its transcrip

Gemini 3.5 Transcribe turns live or recorded speech into formatted text with support for vocabulary adaptation and conversational cleanup. It is designed for voice interfaces, c

VoiceMem adds persistent memory to voice agents by organizing factual information alongside preferences, personality, and emotional context. It can ingest speech and retrieve re

IndexTTS 2.5 is a zero-shot neural text-to-speech foundation model for expressive multilingual voice synthesis. It builds on a transformer-based Text-to-Semantic module and a no

MiDashengLM-Gen is an end-to-end framework for unified audio scene generation. Given a structured text description, it can generate a coherent scene that blends speech, music, s

Audio8 TTS Preview 0.1B is a compact zero-shot text-to-speech model for speech generation and voice cloning. Its main generative model is approximately 170 million parameters, w

CrisperWhisper 2.0 is Nyra Labs' verbatim speech recognition model for transcribing how people actually speak. It supports controllable output styles, multilingual transcription

Fish Audio S2.1 Pro is a production-grade neural text-to-speech model available through the Fish Audio API with the model string s2.1-pro-free. It is designed for developers who

Fish Audio S2 is an expressive open-source text-to-speech model release centered on S2 Pro, Fish Audio's 4B-parameter multilingual voice model. It is built for natural, realisti

dots.tts is a 2B-parameter fully continuous, end-to-end autoregressive text-to-speech system. Its backbone combines a semantic encoder, an LLM, and an autoregressive flow-matchi

Gemini Live Translate is Google's Gemini-powered near real-time speech translation experience for natural spoken conversations. The announcement describes availability through G