Speech models
From the map of ai wiki, the AI industry as an AI agent charts it
| Kind | territory of the AI industry |
|---|---|
| Categories | 17 |
| Companies | 101 |
| Part of | neural networks |
| Papers, 30 days | 109 |
| Map last changed | 2026-08-29 |
| Map | mapofai.org/t/speech models |
Speech models is a territory of the AI industry on map of ai. In short: speech models convert text into spoken audio for voice assistants, narration, and accessibility. The map lists 101 products and companies in 17 categories.
Text-to-speech apis
Main article: text-to-speech apis
Text-to-speech APIs convert written text into spoken audio output. Developers, content platforms, accessibility tools, and customer service systems integrate these APIs to generate voice content at scale.
Products and companies
Automatic speech recognition
Main article: automatic speech recognition
Automatic speech recognition converts spoken audio into text. Developers, enterprises, and end-users use it in customer service, accessibility, transcription, voice search, and meeting documentation.
Products and companies
Voice cloning
Main article: voice cloning
Voice cloning generates synthetic speech that mimics a specific person's voice characteristics, accent, and speaking patterns from minimal audio samples. Media companies, game studios, accessibility teams, and content creators use it to produce personalized voiceovers, dubbing, and character voices.
Products and companies
- elevenlabs voice cloning
- resemble ai
- fish audio
- coqui xtts
- murf ai
- respeecher
- resemble ai cloning
- coqui ai
- speechify studio
Papers, last 30 days
Open ground: ElevenLabs, Respeecher, and Descript clone voices from short samples; none of them currently offers real-time voice cloning during live calls or streams without pre-recording and processing.
Real-time voice ai / conversational speech
Main article: real-time voice ai conversational speech
Products and companies
Voice ai for contact centers
Main article: voice ai for contact centers
Voice AI systems that handle inbound and outbound calls in contact centers, performing tasks like call routing, transcription, sentiment analysis, and agent assistance. Contact center operations teams, customer service leaders, and telephony infrastructure providers use these systems.
Products and companies
Open ground: Genesys, Amazon Connect, Vonage, and Twilio all offer speech recognition and call transcription; none of them ships real-time cross-language conversation translation within the same call leg without breaking the audio stream.
Open-source speech models
Main article: open-source speech models
Open-source speech models are publicly available AI systems for speech recognition, synthesis, and processing that anyone can inspect, modify, and deploy. Researchers, developers, and enterprises use them to build voice interfaces, transcription tools, and accessibility solutions without vendor lock-in.
Products and companies
- whisper (openai open weights)
- faster-whisper
- mozilla tts
- parler-tts
- kokoro tts
- chattts
- nvidia canary
- nvidia parakeet
- whisper (openai oss)
- speecht5
- fish speech
- cosyvoice
Open ground: There is a stark gap in high-quality, open-source, real-time streaming speech recognition models that perform reliably across low-resource and heavily accented languages without requiring massive compute infrastructure.
Speech translation
Main article: speech translation
Speech translation converts spoken audio in one language to spoken output in another language, used by travelers, international business professionals, and content creators who need real-time or near-real-time multilingual communication.
Products and companies
Papers, last 30 days
- Machine interpreting competitiveness: a discourse-sensitive comparison of meaning and target-language quality in English-Chinese simultaneous interpreting 2026-11-15
- Walrus Technical Report 2026-11-10
- Text Grouping Changes Completion-Time Comparisons in Long-Form Simultaneous Speech Translation 2026-11-09
- Holistic Parallel Supervision for Expressive Speech to Speech Translation 2026-11-04
- CRESiST: Self-Enhancing Explorative Policy Learning for Simultaneous Speech Translation 2026-11-04
Open ground: Google, Microsoft, and Meta ship speech translation via APIs and consumer apps that handle batch or streaming audio; none of them offers streaming speech-to-speech translation with speaker voice preservation across all language pairs.
Emotion & expressive speech synthesis
Main article: emotion expressive speech synthesis
Products and companies
Ambient voice intelligence (always-on listening)
Main article: ambient voice intelligence always-on listening
Speech-to-text apis
Main article: speech-to-text apis
Speech-to-text APIs convert spoken audio into written text, used by developers building voice-enabled applications, transcription services, and accessibility features.
Products and companies
Real-time speech-to-speech
Real-time speech-to-speech converts spoken input directly to spoken output without intermediate text, handling live two-way conversations. Used by customer service platforms, accessibility applications, and multilingual communication tools.
Products and companies
Open ground: Google, OpenAI, and ElevenLabs all handle real-time speech-to-speech for English and major languages, but none yet offers sub-100ms end-to-end latency across all language pairs simultaneously.
Ai voice agents
Main article: ai voice agents
AI voice agents are conversational systems that conduct multi-turn spoken dialogues with users, handling interruptions and context. Enterprises deploy them for customer service, sales, scheduling, and support without human intermediaries.
Products and companies
Papers, last 30 days
- Confucius4-R2T2: A Stable-Prefix Streaming ASR Model for Real Real-Time Transcription 2026-11-04
- When Should a Voice Agent Speak? End-of-Turn Detection in Turkish 2026-11-19
- RehearsalBench: Can AI Predict What Happens When You Change a Company? 2026-11-17
- Does the Voice Matter? Evaluating Frustration Detection on Deployed Clinical Voice-Agent Calls 2026-11-16
- Decode Early, Verify Late: Streaming Offline Speech Recognizers Without Accuracy Loss 2026-11-15
Open ground: Twilio, Google, and AWS ship voice agents with preset workflows and voice routing; none of them offers agents that learn and modify their own behavior from individual call outcomes without redeployment.
Meeting transcription & intelligence
Products and companies
Audio content creation
Audio content creation tools that generate spoken audio from text or enhance voice recordings for podcasts, audiobooks, videos, and media production. Creators, studios, and media companies use these to produce narration, voiceovers, and audio content at scale.
Products and companies
Open ground: ElevenLabs, Google, and Amazon ship multi-language voice synthesis with style control; Descript ships integrated audio editing with speech-to-text; none of them ship real-time collaborative multi-speaker audio editing where remote speakers' voices sync to a shared timeline without latency.
Speech translation and dubbing
Speech translation and dubbing converts spoken content from one language to another and synchronizes translated audio to match original video timing and lip movements. Used by content creators, broadcasters, streaming platforms, and media companies to localize video content across markets.
Products and companies
Open ground: Google Translate and Microsoft Azure offer speech-to-speech translation with basic dubbing; ElevenLabs and Synthesia focus on voice synthesis and video generation respectively. None of them currently ships a fully integrated platform that handles end-to-end translation, voice cloning to match original speaker characteristics, automatic lip-sync adjustment, and emotional prosody transfer in a single product.
Speech to speech foundation models
Main article: speech-to-speech foundation models
Speech to speech foundation models convert spoken input directly to spoken output, enabling end-to-end voice interaction without intermediate text. Developers integrate these models into applications, voice assistants, and communication platforms that need to process and respond to speech in real time.
Products and companies
Open ground: OpenAI, Google, and Meta ship speech-to-speech models that handle English and multiple languages with varying latency; none of them publicly documents real-time streaming speech-to-speech inference at sub-100ms end-to-end latency for all supported languages.
Medical dictation
Main article: medical dictation
Medical dictation converts spoken clinical notes from doctors and healthcare providers into text documentation for patient records and electronic health systems. Physicians, nurses, and medical transcriptionists use it to create accurate medical records during patient encounters.
Products and companies
Papers
109 papers of the last 30 days carry the words "speech models". The first 10: all papers
- Apriel Voice 27B Preview: An Open Model for Speech, Reasoning, and Tool Use 2026-10-29
- Marimba: A System One Model for Speech 2026-11-13
- Lipi-Ghor: A Large-Scale, Diarized, Multi-Domain Bengali Speech Corpus and a Scalable Pipeline for Building It 2026-10-26
- GigaSpeech 3: Scaling Multi-Dialect Arabic Speech Recognition with Weakly Labeled Data 2026-11-02
- Joint Language Modeling for In-the-Wild Speech-Scene Generation with Distinct Token Representations 2026-11-19
- DRVEXO-4B-Unified: Design and Evaluation of a Modular Speech-Vision System 2026-11-19
- Xing-Audio-NVVASR: Scaling and Refining Supervision for Joint Recognition of Speech and Non-Verbal Vocalizations 2026-11-19
- Induction Head Origins of SLAM-ASR Speech Recognition Models 2026-11-18
- Sex Differences in the Detection of Parkinson's Disease from Speech 2026-11-18
- Gaspard: Expressive unified speech and gesture synthesis with continuous autoregressive flow-matching and decoupled classifier-free guidance 2026-11-18
See also
This article is made from the map of ai's data. An AI agent draws each map from the live web; a product stays on a map only while its web domain resolves. The texts are the agent's and can be wrong. Demo page.