B2B Review LabSoftware and business finance research

Categories › AI Transcription Software

Best Speech-to-Text APIs for Developers 2026

AssemblyAI is the pick for a general-purpose API: Universal-2 costs $0.15 per hour for 99 languages, with speaker diarization at $0.02 per hour; Speechmatics is the pick for mixed-language audio, with 56 languages at $0.12 per hour. Deepgram is the pick for streaming and pre-recorded Nova models in 60+ languages, while OpenAI is the pick for teams already building on the OpenAI API at $0.0045 per minute. Amazon Transcribe and Google Cloud Speech-to-Text are the picks for teams already on AWS or Google Cloud; Azure AI Speech is the pick for teams already on Azure.

Updated

8 productsbuyer-specific decision factors

Ranked shortlist

  1. Best for: Providers bill by audio duration: per hour (AssemblyAI, Speechmatics, Rev AI) or per minute (OpenAI, Deepgram, Amazon, Google).

    Documented buyer fit
    General-purpose batch and streaming API
  2. Rank2

    OpenAI

    Best for: Providers bill by audio duration: per hour (AssemblyAI, Speechmatics, Rev AI) or per minute (OpenAI, Deepgram, Amazon, Google).

    Documented buyer fit
    Teams already building on the OpenAI API
  3. Best for: Providers bill by audio duration: per hour (AssemblyAI, Speechmatics, Rev AI) or per minute (OpenAI, Deepgram, Amazon, Google).

    Documented buyer fit
    Streaming and pre-recorded Nova models
  4. Best for: Providers bill by audio duration: per hour (AssemblyAI, Speechmatics, Rev AI) or per minute (OpenAI, Deepgram, Amazon, Google).

    Documented buyer fit
    Mixed-language audio
  5. Best for: Google Cloud Speech-to-Text lists $0.016 per minute from the 60th minute each month, with the first 60 minutes free.

    Documented buyer fit
    Teams on Google Cloud
  6. Best for: Microsoft Azure AI Speech is the pick for teams already on Azure.

    Documented buyer fit
    Microsoft Azure AI Speech is the pick for teams already on Azure. Speech to Text is billed per second. Standard batch transcription is $0.18 per hour and real-time transcription is $1 per hour. Custom Transcription batch is $0.225 per hour and real-time is $1.20 per hour; endpoint hosting is $0.0538 per model per hour. These are pay-as-you-go list prices in USD on Microsoft’s pricing page.
  7. Rank8

    Rev AI

    Best for: Providers bill by audio duration: per hour (AssemblyAI, Speechmatics, Rev AI) or per minute (OpenAI, Deepgram, Amazon, Google).

    Documented buyer fit
    English-first transcription from Rev's developer API

Compare the deciding features

Category-specific workflows and plan requirements
Best Speech-to-Text APIs for Developers 2026: documented features
ProductDocumented buyer fit
AssemblyAIGeneral-purpose batch and streaming API[1]
OpenAITeams already building on the OpenAI API[2]
DeepgramStreaming and pre-recorded Nova models[3]
SpeechmaticsMixed-language audio[4]
Amazon TranscribeTeams on AWS[5]
Google Cloud Speech-to-TextTeams on Google Cloud[6]
Azure AI SpeechMicrosoft Azure AI Speech is the pick for teams already on Azure. Speech to Text is billed per second. Standard batch transcription is $0.18 per hour and real-time transcription is $1 per hour. Custom Transcription batch is $0.225 per hour and real-time is $1.20 per hour; endpoint hosting is $0.0538 per model per hour. These are pay-as-you-go list prices in USD on Microsoft’s pricing page.[7]
Rev AIEnglish-first transcription from Rev's developer API[8]

Evidence and sources

  1. AssemblyAI vendor pageRetrieved 2026-11-09T16:21:58Z

    Universal-2 Our highly accurate speech-to-text model trained on over 12.5 million hours of audio data. Supports 99 languages. Exceptional accuracy at a lower price. $0.15 /hr

  2. OpenAI vendor pageRetrieved 2026-11-09T16:22:12Z

    gpt-transcribe Transcription - - $0.0045 / minute

  3. Deepgram vendor pageRetrieved 2026-11-09T16:22:00Z

    Nova models support 60+ languages, with capabilities including speaker diarization, Smart Formatting, and Keyterm Prompting.

  4. Speechmatics vendor pageRetrieved 2026-11-09T16:22:10Z

    Our next-generation multilingual model. Transcribes 56 languages without choosing one up front, switches language mid-sentence.

  5. Amazon Transcribe vendor pageRetrieved 2026-11-09T16:22:02Z

    Batch transcription is priced at $0.006 per minute.

  6. Google Cloud Speech-to-Text vendor pageRetrieved 2026-11-09T16:22:02Z

    60 minute and above $0.016 / 1 minute, per 1 month / account

  7. Azure Speech in Foundry Tools pricingRetrieved 2026-11-09T19:05:00Z

    Speech to Text (per second billing) Standard Transcription Real-time Transcription: $1 per hour Batch Transcription: $0.18 per hour Custom Transcription Real-time Transcription: $1.20 per hour Batch Transcription: $0.225 per hour Endpoint hosting: $0.0538 per model per hour

  8. Rev AI vendor pageRetrieved 2026-11-09T16:22:14Z

    Reverb Transcription $0.20 per hour Languages: English

AssemblyAI is the pick for a general-purpose API: Universal-2 costs $0.15 per hour for 99 languages, with speaker diarization at $0.02 per hour; Speechmatics is the pick for mixed-language audio, with 56 languages at $0.12 per hour. Deepgram is the pick for streaming and pre-recorded Nova models in 60+ languages, while OpenAI is the pick for teams already building on the OpenAI API at $0.0045 per minute. Amazon Transcribe and Google Cloud Speech-to-Text are the picks for teams already on AWS or Google Cloud; Azure AI Speech is the pick for teams already on Azure.

At a glance

Best Speech-to-Text APIs for Developers 2026: documented pricing, limits and fit
ProductBest forPublished terms and capabilities
AssemblyAIGeneral-purpose batch and streaming APIPricing: Universal-2 $0.15/hr (99 languages); Universal-3.5 Pro $0.21/hr (18 languages) · Extras: Speaker diarization $0.02/hr
Vendor source
OpenAITeams already building on the OpenAI APIPricing: gpt-transcribe $0.0045/min; gpt-4o-transcribe $0.006/min; live $0.017/min
Vendor source
DeepgramStreaming and pre-recorded Nova modelsPricing: Nova-3 monolingual $0.0077/min pay-as-you-go; $0.0065 per minute on the Growth plan (pre-paid from $4,000 a year) · Languages: 60+
Vendor source
SpeechmaticsMixed-language audioPricing: Multilingual model $0.12/hr; Enhanced model $0.38/hr · Languages: 56 without choosing one up front
Vendor source
Amazon TranscribeTeams on AWSPricing: Batch $0.006/min; 60 free minutes a month for 12 months
Vendor source
Google Cloud Speech-to-TextTeams on Google CloudPricing: $0.016/min after 60 free minutes a month
Vendor source
Azure AI SpeechTeams already on Microsoft AzurePricing: Standard batch $0.18/hour; real-time $1/hour; Custom batch $0.225/hour; real-time $1.20/hour; endpoint hosting $0.0538 per model per hour
Vendor source
Rev AIEnglish-first transcription from Rev's developer APIPricing: Reverb $0.20/hr English; foreign language $0.30/hr · Languages: English plus 57 more
Vendor source

How it works

A speech-to-text API accepts audio from your application, either as files (batch or pre-recorded) or as a live stream, and returns text with timestamps and, optionally, speaker labels.

Providers bill by audio duration: per hour (AssemblyAI, Speechmatics, Rev AI) or per minute (OpenAI, Deepgram, Amazon, Google). Some features, such as AssemblyAI's diarization, are priced separately.

1. AssemblyAI: best general-purpose speech-to-text API

AssemblyAI offers pay-as-you-go with a free start and no commitments. Universal-2 costs $0.15 per hour and supports 99 languages; Universal-3.5 Pro costs $0.21 per hour and works across 18 languages with native code switching. Speaker diarization is priced at $0.02 per hour, and word-level timestamps are listed. Both a pre-recorded and a realtime API are offered, and Universal-3.6 Pro Realtime is its realtime model for voice agents.

Primary vendor pages: AssemblyAI vendor page.

2. OpenAI: best for teams already building on the OpenAI API

OpenAI lists gpt-transcribe at $0.0045 per minute, gpt-4o-transcribe at $0.006 per minute, and gpt-realtime-whisper for live transcription at $0.017 per minute.

Primary vendor pages: OpenAI vendor page.

3. Deepgram: best for streaming and pre-recorded Nova models

Deepgram's Nova models support 60+ languages with speaker diarization, smart formatting and keyterm prompting, and are offered for streaming and pre-recorded audio. Nova-3 monolingual is listed at $0.0077 per minute pay-as-you-go, with $0.0065 per minute on the Growth plan, which is pre-paid from $4,000 a year. New accounts start with a $200 credit, with no minimums and no card required.

Primary vendor pages: Deepgram vendor page.

4. Speechmatics: best for audio that switches languages

Speechmatics' multilingual model transcribes 56 languages without choosing one up front and switches language mid-sentence, at $0.12 per hour. The Enhanced model, at $0.38 per hour, targets the highest accuracy; you pick the language for each transcription, and it supports a custom dictionary and advanced punctuation. Real-time models are also offered, and new accounts start with $100 in credits without a card.

Primary vendor pages: Speechmatics vendor page.

5. Amazon Transcribe: best for teams already on AWS

Amazon Transcribe bills per second of audio. Batch transcription is $0.006 per minute, tiered pricing applies to streaming and batch, and the free tier gives up to 60 audio minutes a month for the first 12 months. Automatic Content Redaction can redact sensitive information such as personally identifiable information.

Primary vendor pages: Amazon Transcribe vendor page.

6. Google Cloud Speech-to-Text: best for teams already on Google Cloud

Google Cloud Speech-to-Text lists $0.016 per minute from the 60th minute each month, with the first 60 minutes free. The chirp model is available in Speech-to-Text V2 only. Pricing is listed separately for requests with and without data logging.

Primary vendor pages: Google Cloud Speech-to-Text vendor page.

7. Azure AI Speech: best for teams already on Microsoft Azure

Microsoft Azure AI Speech is the pick for teams already on Azure. Speech to Text is billed per second. Standard batch transcription is $0.18 per hour and real-time transcription is $1 per hour. Custom Transcription batch is $0.225 per hour and real-time is $1.20 per hour; endpoint hosting is $0.0538 per model per hour. These are pay-as-you-go list prices in USD on Microsoft’s pricing page.

Primary vendor pages: Azure Speech in Foundry Tools pricing.

8. Rev AI: best for English-first transcription from Rev's developer API

Rev AI is the developer API from Rev. Reverb transcription costs $0.20 per hour for English, and Reverb Foreign Language transcription costs $0.30 per hour for Spanish, French, Chinese, Portuguese and 53 more languages. Rates round up to the nearest second with a 15-second minimum. A streaming API is offered, and new accounts receive free credits equal to 5 hours of Reverb ASR.

Primary vendor pages: Rev AI vendor page.

How to choose

  • Decide batch versus real-time first: AssemblyAI, Deepgram and Speechmatics each list separate streaming or real-time and pre-recorded options.
  • Convert every price to the same unit before comparing: $0.15 per hour is $0.0025 per minute.
  • Check the language list for the model you will use, since AssemblyAI's Universal-2 (99 languages) and Universal-3.5 Pro (18 languages) carry different prices.
  • Price the features you need (diarization, redaction, custom vocabulary) and use the free credits to test your own audio.
  • Live-meeting note-taking belongs in the AI meeting assistant category; this page compares uploaded recordings or developer APIs.

Frequently asked questions

Which speech-to-text APIs offer speaker diarization?

AssemblyAI lists speaker diarization at $0.02 per hour, and Deepgram lists it among its Nova model capabilities.

How is speech-to-text API pricing structured?

AssemblyAI, Speechmatics and Rev AI price per hour; OpenAI, Deepgram, Amazon and Google price per minute; Amazon bills per second of audio. AssemblyAI prices diarization separately.

Which API has the lowest published price?

Speechmatics lists its multilingual model at $0.12 per hour, the lowest rate here. AssemblyAI Universal-2 is $0.15 per hour, and OpenAI gpt-transcribe is $0.0045 per minute, which is $0.27 per hour.

Do any of these APIs have a free tier?

AssemblyAI starts free, Deepgram gives a $200 credit, Speechmatics gives $100 in credits, Amazon Transcribe gives up to 60 minutes a month for 12 months, Google gives 60 free minutes a month, and Rev AI gives 5 hours of free credits.

Which API handles mixed-language audio?

Speechmatics' multilingual model transcribes 56 languages without choosing one up front and switches language mid-sentence; AssemblyAI Universal-3.5 Pro supports native code switching across 18 languages.

Methodology

Every price, limit, language count and feature comes from the vendor's own pages, opened on 9 November 2026 (Europe/Helsinki), with the exact URL and a verbatim excerpt recorded for each fact. Prices are shown in the currency the vendor prints; where a vendor lists annual and monthly billing, the billing basis is named.

Vendor families are counted once.

Picks follow documented fit for the buying situation and published facts, not paid placement or review-site scores. Accuracy figures are shown only as the vendor states them.

How this comparison is made

Ranked on documented capabilities and verified pricing from official vendor sources. Every fact links to its source and check date, and the criteria are specific to this buying situation.

Read the methodology →

How we assessed this shortlist

This is an unordered, evidence-led comparison for this buyer situation. Recommendations follow workflows and boundaries documented on primary vendor pages. Ranked on documented capabilities and verified pricing from official vendor sources.

Want your software in this comparison? A vendor profile listing costs $9.90 via Stripe and goes live within 24 hours of payment. Comparison-list placements cost a fee, and links are dofollow. Factual corrections are free.

Submit your software →

Spot an error or a missing product?

Send a correction with a source link and we review it against the vendor's own documentation.

Suggest a correction