- Speech to text APIs side by side
- Eleven speech to text APIs to compare
- Where to start for your app
- A transcript isn't automatically a caption
- Real-time versus batch transcription
- Questions to ask before you choose a transcription API
- Signs you need a better transcription setup
- Four more speech to text options worth a look
- My plain advice on choosing a speech to text API
- Frequently asked questions
The best speech to text APIs turn recorded or live audio into text inside your own app, website or system. Some are part of big cloud platforms, others are specialists in live conversations, many languages or on-device privacy. Eleven options follow in no particular order.
Every call, meeting, podcast and video holds words you could search, quote or caption. A speech to text API is how software gets at them without someone typing everything out.
The eleven APIs below are listed alphabetically and in no particular order, and every description comes from the provider's own website, read this week. If you're adding captions for video, my notes on turning on captions in Zoom show how a finished caption experience looks.
Speech to text APIs side by side
I checked the free plan column on each tool's own website, as of Nov 2026. Yes means the site names a free plan, a free account, a free tier or an open source version. Trial only means a time-limited trial. Not stated means the pages I read named neither, even where there's a button inviting you to start for free, so check the vendor's page before you sign up.
| Tool | Type | Good for | Free plan |
|---|---|---|---|
| Amazon Transcribe | Speech to text service from AWS | Apps already built on AWS | Yes |
| AssemblyAI | AI models to transcribe and understand speech | Clean, ready-to-use transcripts | Yes, as of Nov 2026 (up to 185 hours pre-recorded, no card) |
| Azure AI Speech | Microsoft Azure speech services | Microsoft-based businesses | Yes |
| Deepgram | Voice AI with STT, TTS and agent APIs | Live conversations and voice agents | $200 free credit, as of Nov 2026 |
| ElevenLabs | Speech to text model and voice AI | Creators already using voice AI | Yes |
| Gladia | Audio infrastructure for voice products | Real-time and batch transcription | Not stated |
| Google Cloud Speech-to-Text | Speech to text from Google Cloud | Apps on Google Cloud | Not stated |
| Picovoice Leopard | On-device speech to text engine | Transcribing without sending audio | Not stated |
| Rev AI | Speech to text API | Accuracy across many languages | Not stated |
| Soniox | Multilingual speech AI platform | Multilingual real-time apps | Not stated |
| Speechmatics | Speech APIs for voice AI | Accents and many languages | $100 free credit, as of Nov 2026 |
Eleven speech to text APIs to compare
Amazon Transcribe
Best for: Companies already on AWS who want speech to text that plugs into their existing cloud setup, with a free tier to start.
AssemblyAI
Best for: Product teams that want transcripts tidied up for them, with filler words removed and names spelled correctly.
Azure AI Speech
Best for: Businesses built on Microsoft tools that want speech to text inside the Azure and Foundry ecosystem.
Deepgram
Best for: Teams building voice agents and live call tools that need speech recognition built for real-time conversation.
ElevenLabs
Best for: Creators and developers who already use voice AI and want speech to text from the same provider, with a free plan.
Gladia
Best for: Startups building voice products that want both real-time and asynchronous speech to text APIs.
Google Cloud Speech-to-Text
Best for: Google Cloud customers who want speech recognition and transcription that integrates with their other Google services.
Picovoice Leopard
Best for: Privacy-sensitive apps that need transcription on the device itself, without sending audio to any server.
Rev AI
Best for: Developers who want a speech to text API with translation options for customers in many languages.
Soniox
Best for: Global products that need real-time speech to text, translation and voice in many languages.
Speechmatics
Best for: Businesses with customers who speak many languages and accents that want multilingual speech to text APIs.
Where to start for your app
- Apps already built on AWS: Amazon Transcribe.
- Live calls and voice agents: Deepgram.
- Transcribing on the device itself: Picovoice Leopard.
- Many languages in real time: Soniox or Speechmatics.
- Tidy transcripts ready to use: AssemblyAI.
A transcript isn't automatically a caption
The W3C Web Accessibility Initiative's guidance on WCAG Success Criterion 1.2.2 explains that captions don't only include dialogue. They identify who is speaking and include non-speech information conveyed through sound, such as meaningful sound effects.
So if you're using an API to caption videos, accuracy is only part of the job. Look for speaker labels and timing, and plan to add sound cues where they matter to the meaning.
Next on the reading list: text to speech APIs.
Real-time versus batch transcription
Real-time transcription processes speech as it happens, which suits live captions, call assistants and voice agents. Speed and low delay matter most.
Batch transcription processes a recording after it ends. It usually suits podcasts, interviews and meeting recordings, where you can wait a little longer for the most accurate result.
Audio isn't the only thing that needs turning into text, and our guide to OCR software lines up tools that read scans and photos.
Want AI doing the heavy lifting in your marketing?
I build the systems that handle the boring 80 percent, so you get your week back. Done properly, with the human kept in.
Questions to ask before you choose a transcription API
- Does it support the languages and accents I need?
- Is there real-time streaming as well as batch?
- Can it label different speakers?
- Does it give word-level timestamps for captions?
- Where is the audio processed and stored?
- Can I add custom words, such as product names?
Transcripts often need checking by hand, and this list of data annotation tools weighs up tools built for labelling audio and text.
Signs you need a better transcription setup
- Someone types up recordings by hand. Automate the first draft.
- Names come out wrong every time. Add custom vocabulary.
- Captions don't show who's talking. Use speaker labels.
- Live captions lag badly. Try a real-time engine.
Do you make a speech to text API? Tool makers can ask to be considered for a featured place on this list.
Four more speech to text options worth a look
Updated 6 November 2026. These four aren't in the list above. I've set them out as the questions I'd want answered before trying each one.
Which one has open-weight models I can run myself?
Mistral Voxtral. Mistral's blog post dated 15 July 2025 introduced Voxtral in a 24B size for production-scale apps and a 3B size for local and edge use, released under an open licence.
Which one suits streaming and voice agents?
Cartesia Ink, Cartesia's streaming speech to text model, sold alongside its text to speech and voice agent products. As of Nov 2026 the pricing page shows a Free plan at $0 a month with 20K credits a month, speech to text on every plan, and Pro at $5 a month.
Is there a low-cost Whisper API?
Lemonfox.ai runs Whisper large-v3 with speaker recognition and 100+ languages. As of Nov 2026 its homepage says less than $0.17 per hour of speech, with the first month free, then $5 a month for 30 hours of speech to text. EU-based processing is available.
What if I only need to know who said what?
pyannoteAI isn't a transcription engine. It's a speaker diarization, identification and separation API that sits beside whichever speech to text API you pick. As of Nov 2026 Developer is €19.00 a month including €19 of usage credit, Starter is €99.00 a month, and there's a 30-day trial with no credit card required.
My verdict: Voxtral if you want to self-host, Cartesia Ink for live voice, Lemonfox.ai if you're watching every cent, and pyannoteAI as an add-on when speaker labels matter.
Want your tool on this list? Featured spots and reviews here are paid partnerships. Here's how to partner with me.
My plain advice on choosing a speech to text API
Test with your own audio. Every provider sounds brilliant on clean studio samples, so send each one a few of your real recordings, with your accents, background noise and product names, and compare the results side by side.
Then think about where the audio goes. If recordings include customer details, choose a provider whose data handling you're happy to explain to those customers. For a quick look at style choices once the words are right, my video caption style cheat sheet helps.
Related: Best AI Kitchen Design Tools.
Frequently asked questions
What is a speech to text API?
A speech to text API is a service your software calls to turn audio into written text, either live or from a recording.
Which speech to text APIs have a free tier?
Some offer a free tier, free plan or free credits. The free plan column above shows what each provider's own site says.
Can speech to text APIs identify speakers?
Many offer speaker labels, sometimes called diarization, which mark who said each part of the transcript.
Is a transcript the same as captions?
Not quite. W3C explains that captions also identify speakers and include important non-speech sounds.
Can transcription run without the cloud?
Some engines run on the device itself, so audio never leaves the phone, computer or server.
How do I improve transcription accuracy?
Use clear audio, add custom vocabulary for names and jargon, and test several providers with your own recordings.
