License & Deployment Mix: 13 tools – 4 OSS, 8 SaaS. (OSS and SaaS counts can overlap when an open-source tool also offers a hosted edition.)
What Is Speech-to-Text Software?
Speech-to-text (STT), also called automatic speech recognition (ASR), converts spoken audio into written text. It powers transcription of meetings and support calls, live captioning, voice dictation, voice commands, and the input side of voice assistants.
The market splits cleanly along one axis that matters most:
- Local / self-hosted engines run the model on your own hardware. Audio never leaves the machine or network – important for privacy, compliance, and air-gapped sites. Examples: Whisper, whisper.cpp, Vosk, NVIDIA Riva.
- Cloud services send audio to a vendor API over the internet. They are the easiest to adopt and often the most accurate, but audio leaves your environment and usage is metered. Examples: Deepgram, AssemblyAI, Google Cloud Speech-to-Text, AWS Transcribe, Azure AI Speech, Otter.ai, Rev AI.
Core Concepts
| Concept | Description |
|---|---|
| ASR model | The neural model that maps audio to text |
| Batch vs streaming | Transcribe a finished file vs a live audio stream |
| WER | Word error rate – the standard accuracy metric (lower is better) |
| Diarization | Labeling who spoke each segment (speaker separation) |
| Timestamps | Word- or segment-level timing (for captions / subtitles) |
| Language / multilingual | Supported languages; auto language detection |
| On-device | Runs fully offline, no network call |
The information on these pages was researched by a combination of human review and large language models. To suggest an addition or correction, please contact us. Prepared by Rhodium Systems Inc., author of the ResorsIT platform — a unified IT operations management platform for IT teams and MSPs that integrates a curated suite of open-source, commercial, and SaaS applications into a single system with shared identity, single sign-on, access control, and a common audit trail. Use this catalogue only as a starting point for your own research, and review any tool carefully against your own requirements before relying on it. Catalogue data version 2026.197.
Comparison
This comparison covers speech-to-text engines for call and meeting transcription, video captioning, dictation, and accessibility. The guiding question is where the audio goes: local engines keep it in the customer environment; cloud services send it to a vendor over the internet.
Local vs Cloud (Where the Audio Goes)
This is the table that matters most: whether a tool runs entirely on your own hardware or needs an internet connection to a vendor service.
| Tool | Runs | Needs internet? | Notes |
|---|---|---|---|
| Whisper (OpenAI) | Local | No | Runs fully offline once the model is downloaded |
| whisper.cpp | Local | No | CPU-first C/C++ port; fully offline |
| Vosk | Local | No | Lightweight, offline; runs on small / embedded devices |
| NVIDIA Riva | Local (self-host) | No | Self-hosted server; needs an NVIDIA GPU |
| Handy | Local (desktop) | No | Push-to-talk dictation app; Whisper / Parakeet, fully offline |
| Deepgram | Cloud (+ on-prem) | Yes* | Cloud API; self-hosted container on enterprise |
| AssemblyAI | Cloud | Yes | Cloud API only |
| Speechmatics | Cloud (+ on-prem) | Yes* | Cloud API; on-prem container available |
| Google Cloud STT | Cloud | Yes | Cloud API only |
| AWS Transcribe | Cloud | Yes | Cloud API only |
| Azure AI Speech (STT) | Cloud (+ container) | Yes* | Cloud API; disconnected container available |
| Otter.ai | Cloud | Yes | Web / mobile app; SaaS only |
| Rev AI | Cloud | Yes | Cloud API (and human-transcription service) |
* = cloud by default; a self-hosted / on-prem option removes the internet dependency on paid enterprise tiers.
Platforms
Which operating systems / environments each tool runs on. Cloud APIs are marked “Any (HTTP)” because they are called over HTTPS from any platform; the row notes where a self-hostable component runs.
| Tool | Windows | macOS | Linux | Mobile / Embedded | Cloud |
|---|---|---|---|---|---|
| Whisper (OpenAI) | Yes | Yes | Yes | via ports | – |
| whisper.cpp | Yes | Yes | Yes | iOS / Android / ARM | – |
| Vosk | Yes | Yes | Yes | Android / iOS / RPi | – |
| NVIDIA Riva | No | No | Yes (Docker/K8s) | NVIDIA Jetson | – |
| Handy | Yes (x64/ARM) | Yes | Yes (x64/ARM) | – (desktop only) | – |
| Deepgram | Any (HTTP) | Any | Any | Any | Yes; on-prem = Linux/Docker |
| AssemblyAI | Any (HTTP) | Any | Any | Any | Yes |
| Speechmatics | Any (HTTP) | Any | Any | Any | Yes; on-prem = Linux/Docker |
| Google Cloud STT | Any (HTTP) | Any | Any | Any | Yes |
| AWS Transcribe | Any (HTTP) | Any | Any | Any | Yes |
| Azure AI Speech (STT) | Any (HTTP) | Any | Any | Any | Yes; container = Linux/Docker |
| Otter.ai | Browser | Browser | Browser | iOS / Android | Yes |
| Rev AI | Any (HTTP) | Any | Any | Any | Yes |
Licensing
| Tool | License | OSS | Cost Model |
|---|---|---|---|
| Whisper (OpenAI) | MIT | Yes | Free (self-host); compute only |
| whisper.cpp | MIT | Yes | Free (self-host) |
| Handy | MIT | Yes | Free (desktop app) |
| Vosk | Apache-2.0 | Yes | Free (self-host) |
| NVIDIA Riva | Proprietary | No | Free SDK; needs NVIDIA GPU / NGC |
| Deepgram | Proprietary | No | Per-minute; free credit; on-prem custom |
| AssemblyAI | Proprietary | No | Per-hour usage; free tier |
| Speechmatics | Proprietary | No | Per-hour; on-prem enterprise |
| Google Cloud STT | Proprietary | No | Per-minute usage |
| AWS Transcribe | Proprietary | No | Per-minute usage |
| Azure AI Speech (STT) | Proprietary | No | Per-hour usage; container tier |
| Otter.ai | Proprietary | No | Freemium; per-seat Business |
| Rev AI | Proprietary | No | Per-minute usage |
Feature Matrix
| Feature | Whisper | whisper.cpp | Vosk | Riva | Handy |
|---|---|---|---|---|---|
| Streaming (real-time) | No | Partial | Yes | Yes | Yes (dictation) |
| Batch (file) | Yes | Yes | Yes | Yes | No |
| Diarization | Add-on | Add-on | Basic | Yes | No |
| Word timestamps | Yes | Yes | Yes | Yes | No |
| Languages | ~100 | ~100 | 20+ | 10+ | ~100 |
| GPU needed | Optional | No | No | Yes | Optional |
| Feature | Deepgram | AssemblyAI | Speechmatics | AWS | Azure | Otter | Rev | |
|---|---|---|---|---|---|---|---|---|
| Streaming (real-time) | Yes | Yes | Yes | Yes | Yes | Yes | Yes | Yes |
| Batch (file) | Yes | Yes | Yes | Yes | Yes | Yes | Yes | Yes |
| Diarization | Yes | Yes | Yes | Yes | Yes | Yes | Yes | Yes |
| Word timestamps | Yes | Yes | Yes | Yes | Yes | Yes | Yes | Yes |
| Languages | 30+ | 15+ | 50+ | 125+ | 100+ | 100+ | few | few |
| Self-host option | Yes | No | Yes | No | No | Yes | No | No |
SSO / OIDC Comparison
The notes below assume Authentik as the external identity provider. Local engines are libraries/CLIs with no user login; cloud APIs authenticate with API keys, and end-user SSO (where present) is an enterprise feature.
| Tool | End-user SSO | Authentik Notes |
|---|---|---|
| Whisper / whisper.cpp / Vosk | N/A | Library / CLI; no login |
| Handy | N/A | Local desktop app; no login |
| NVIDIA Riva | N/A | Self-hosted service; API auth |
| Deepgram / AssemblyAI / Rev AI | None | API-key auth |
| Google / AWS / Azure | None | Cloud IAM, not Authentik OIDC |
| Speechmatics | Paid | SSO on enterprise |
| Otter.ai | Paid | SAML/OIDC on Business/Enterprise |
Tools
13 tools.
Amazon Transcribe
Amazon Transcribe is AWS’s cloud speech-to-text service with streaming and batch transcription, speaker diarization, custom vocabulary, and call-analytics and medical variants.
License: Proprietary (proprietary) · Kind: service · Deploy: saas · SSO: none · Tags: speech-to-text, asr, cloud, aws
AssemblyAI
AssemblyAI is a cloud speech-to-text API with real-time and batch transcription plus audio-intelligence models for diarization, summarization, and topic detection.
License: Proprietary (proprietary) · Kind: service · Deploy: saas · SSO: none · Tags: speech-to-text, asr, cloud, api
Azure AI Speech (Speech-to-Text)
Azure AI Speech provides cloud speech-to-text for 100+ languages with real-time and batch transcription, diarization, and custom models; a disconnected container is available.
License: Proprietary (proprietary) · Kind: service · Deploy: saas, docker · SSO: none · Tags: speech-to-text, asr, cloud, azure, container
Deepgram
Deepgram is a cloud speech-to-text API delivering fast, accurate real-time and batch transcription with diarization; a self-hosted container is available for enterprises.
License: Proprietary (proprietary) · Kind: service · Deploy: saas, docker · SSO: none · Tags: speech-to-text, asr, cloud, api
Google Cloud Speech-to-Text
Google Cloud Speech-to-Text is a cloud API transcribing 125+ languages with streaming and batch modes, speaker diarization, and model adaptation, billed per minute.
License: Proprietary (proprietary) · Kind: service · Deploy: saas · SSO: none · Tags: speech-to-text, asr, cloud, gcp
Handy
Handy is a free, open-source, cross-platform desktop app for offline voice dictation – press a shortcut, speak, and the transcribed text is typed into any application, with all speech recognition running locally.
License: MIT (OSS) · Kind: desktop · Deploy: native · SSO: none · Tags: speech-to-text, dictation, offline, desktop
NVIDIA Riva
NVIDIA Riva is a GPU-accelerated speech AI SDK for self-hosted, low-latency real-time speech recognition and synthesis, deployed as containers on NVIDIA hardware.
License: Proprietary (proprietary) · Kind: service · Deploy: docker, k8s · SSO: none · Tags: speech-to-text, asr, self-hosted, gpu
Otter.ai
Otter.ai is a cloud meeting-transcription app that records and transcribes conversations in real time with speaker labels, summaries, and search, via web and mobile.
License: Proprietary (proprietary) · Kind: web · Deploy: saas · SSO: OIDC, SAML · Tags: speech-to-text, transcription, meetings, cloud
Rev AI
Rev AI is a cloud speech-to-text API for asynchronous and streaming transcription; parent company Rev also offers human transcription for higher accuracy.
License: Proprietary (proprietary) · Kind: service · Deploy: saas · SSO: none · Tags: speech-to-text, asr, cloud, api
Speechmatics
Speechmatics is a speech-to-text API with strong accuracy across accents and 50+ languages, offered as a cloud service or a self-hosted on-premises container.
License: Proprietary (proprietary) · Kind: service · Deploy: saas, docker · SSO: OIDC, SAML · Tags: speech-to-text, asr, cloud, on-prem
Vosk
Vosk is an open-source offline speech recognition toolkit with lightweight models for 20+ languages, running on Linux, Windows, macOS, Android, iOS, and Raspberry Pi.
License: Apache-2.0 (OSS) · Kind: library · Deploy: native · SSO: none · Tags: speech-to-text, asr, offline, embedded
Whisper
OpenAI Whisper is an open-source automatic speech recognition model that transcribes and translates around 100 languages, running fully offline on local hardware.
License: MIT (OSS) · Kind: library · Deploy: native, docker · SSO: none · Tags: speech-to-text, asr, transcription, offline
whisper.cpp
whisper.cpp is a high-performance C/C++ port of OpenAI Whisper that runs speech recognition fully offline on CPU or GPU, from servers down to phones and ARM boards.
License: MIT (OSS) · Kind: cli · Deploy: native · SSO: none · Tags: speech-to-text, asr, offline, cpp
