Deepgram alternatives in 2027 for speech to text and voice
Deepgram alternatives compared for developers: speech to text APIs, text to speech voices, and voice agent engines, with 2027 price per minute. AssemblyAI adds audio intelligence, Speechmatics and Gladia cover more languages, ElevenLabs replaces Aura voices, and PyAI handles phone audio, voices and agents on one key.

On this page
Why developers look for Deepgram alternatives
Deepgram is respected for low latency streaming and has 482 G2 reviews at 4.6. Threads on Deepgram alternatives on Reddit focus on price, open models and reliability on specific audio. The common reasons to switch:
Streaming costs more than batch
Nova-3 streaming is $0.0077 a minute on pay as you go, against $0.0043 for pre-recorded audio, and Growth discounts of up to 20% need $4,000 a year prepaid.
Raw transcription only
Sentiment, PII redaction, entities and summaries need a separate LLM step, which is why many teams move to AssemblyAI for call analytics and compliance work.
Language coverage
G2 reviewers flag limited language support most often. Speechmatics, Gladia and Whisper cover more languages.
Conflicting accuracy claims
AssemblyAI and Deepgram each publish benchmarks that favor themselves, so teams test on their own audio before committing.
Deepgram alternatives at a glance
Each tool is ranked on price per minute, accuracy features, languages and fit for voice apps, from vendor pricing pages on 20 October 2026.
| Tool | Built for | Features | Free credit | Price (2027) | Type |
|---|---|---|---|---|---|
| PyAI | Phone audio, voices, agents | Hear for phone audio, Speak and Cast for synthesis, Omni for agents, all on an OpenAI compatible endpoint returning WAV, PCM or G.711 | $50, no credit cardSandbox keys never bill | Hear $0.001/min. Speak $0.04/min. Omni $0.05/min + $0.01 telephony | API |
| AssemblyAI | Teams wanting audio intelligence | Transcription plus sentiment analysis, PII redaction, entity detection and summaries | $50 | Universal-2 $0.15/hr, streaming $0.15/hr | API |
| Speechmatics | Teams needing multilingual accuracy | Wide language coverage with free speaker diarization across batch and real time | $100 | $0.24 to $0.80/hr | API |
| Gladia | Many languages in real time | Many languages, diarization and audio intelligence at a low hourly rate | EUR 50 | EUR 0.16/hr batch, EUR 0.68/hr real time | API |
| ElevenLabs | Expressive voices plus transcripts | Broad expressive voice library with Scribe transcription on the same account | 10,000 credits | Scribe v2 $0.22/hr. TTS $0.04 to $0.08 per 1K chars | API |
| OpenAI Whisper | Teams self hosting transcription | Open source model you run on your own GPUs with no license fee | Free, open source | Your own GPUs | Open model |
| Rev.ai | Batch or human transcripts | Machine and human transcription from one API for certified accuracy | 5 hours | $0.20/hr English | API |
| Deepgram (for comparison) | Low latency STT, TTS, agents | Nova transcription, Aura voices and a Voice Agent API over one connection | $200 | Nova-3 $0.0077/min streaming. Aura-2 $0.030 per 1K charsVoice Agent $0.075/min | API |
Deepgram pricing vs alternatives, per minute and per hour
A cheaper Deepgram alternative depends on whether you stream or batch. The chart prices 100,000 minutes of transcription a month, about 1,667 hours, from each vendor's published rate. The PyAI speech to text comparison shows the same rates side by side.
PyAI Hear vs Deepgram, 100,000 minutes a month
Save $670 a month on streaming
Monthly cost of 100,000 minutes of speech to text from published rates. Bars show the monthly cost. The right column compares each with Deepgram's pay as you go streaming rate.
- PyAI HearStreaming, phone audioSave $670
$100 a month at $0.001 a minute. Batch jobs cost $50 at $0.0005
- AssemblyAIStreaming UniversalSave $520
$250 a month at $0.15 an hour
- Rev.aiBatch EnglishSave $437
About $333 a month at $0.20 an hour, batch only
- Deepgram Nova-3The tool being replacedBaseline
$770 a month at $0.0077 a minute pay as you go streaming, or $430 for pre-recorded audio
- SpeechmaticsReal timeSave $270
From $500 a month at $0.30 an hour real time
- ElevenLabs ScribeReal timeSave $120
$650 a month at $0.39 an hour real time
- GladiaReal timeMore
About EUR 1,133 a month at EUR 0.68 an hour real time. Batch is far cheaper
- OpenAI WhisperSelf hostedNot comparable
No license fee. Cost is your GPU servers and engineering time
Rates checked 20 October 2026 on vendor pricing pages. Deepgram rates rechecked on deepgram.com/pricing 30 October 2026.
Deepgram competitors reviewed
The seven picks cover Deepgram's three jobs. PyAI ranks first for phone audio, voices and agents on one key. AssemblyAI, Speechmatics, Gladia and Rev.ai replace transcription, ElevenLabs replaces Aura voices, and Whisper is the free self hosted route.
PyAI
Built for phone audio transcription, natural voices and one engine phone agents
PyAI is voice infrastructure that covers the same three jobs as Deepgram on one key. Hear is speech to text tuned for 8 kHz phone audio, Speak is text to speech, and Omni is a speech to speech engine for phone agents. Hear uses an OpenAI compatible endpoint, so most code moves with a base URL change. PyAI is not a swap for every Deepgram use: its own comparison says Deepgram quotes a range that starts lower, 150 to 300 ms, and supports more languages.
Hear costs $0.001 per minute streaming, or $0.0005 in batch. Speak costs $0.04 a minute with 200 voices and free cloning. Omni runs $0.05 a minute, plus $0.01 for telephony, against $0.075 for Deepgram's Voice Agent API before telephony.
Hear launched recently, so G2 lists no PyAI reviews to weigh against Deepgram's.
PyAI features
- Hear
- Streaming and batch speech to text built for phone calls, with partial and final results.
- Speak
- Text to speech with first audio in about 53 ms in PyAI's tests.
- Omni
- Speech to speech phone agents, about 390 ms voice to voice.
- OpenAI compatible
- Swap the base URL to move code.
PyAI pricing
Hear
$0.001/min
Batch $0.0005
Speak
$0.04/min
200 voices
Omni
$0.05/min
Plus $0.01 phone line
Hear bills by the minute on no monthly plan, a sandbox key never charges you, and each new account opens with $50 of free credit, no credit card needed. See the Deepgram migration guide for field mapping.
PyAI pros and cons
Pros
- About an eighth of Deepgram's pay as you go streaming rate.
- Transcription, voices and agents on one account.
- Built for phone audio and call workflows.
Cons
- Fewer languages than Deepgram.
- Deepgram quotes a range that starts lower, 150 to 300 ms.
- Hear has no third party reviews on G2, and its 200ms first partial comes from PyAI's own tests.
PyAI reviews
G2 carries no PyAI listing today, so this tab quotes no user opinions. A sandbox key is free with $50 of credit and no credit card, enough to run your own call recordings through Hear and compare transcripts.
Choose PyAI if you transcribe phone calls at volume, or build phone agents, and want one key and a low per minute rate.
AssemblyAI
Built for teams that want transcription plus audio intelligence
AssemblyAI is the audio intelligence pick: beyond the transcript it adds sentiment analysis, PII redaction, entity detection, summaries and LLM analysis of transcripts, which Deepgram leaves to a separate model.
Universal-2 costs $0.15 an hour and Universal-3.5 Pro $0.21 an hour. Streaming starts at $0.15 an hour, and its Voice Agent API is $4.50 an hour. New accounts get $50 of credit. The AssemblyAI bundle suits products that need insight from calls, not just the words, and it costs more per hour than bare transcription, so it fits teams where the analysis earns its keep rather than high volume raw speech to text where the per minute price leads the decision.
AssemblyAI features
- Audio intelligence
- AssemblyAI runs sentiment analysis, PII redaction, entity detection and summaries on the transcript without a second model.
- LLM analysis
- Transcripts can be passed to LLM analysis inside AssemblyAI, the step Deepgram users wire up themselves.
- Batch models
- Universal-2 costs $0.15 an hour and Universal-3.5 Pro $0.21 an hour for recorded audio.
- Streaming
- AssemblyAI streaming starts at $0.15 an hour, about $0.0025 a minute.
- Voice Agent API
- A Voice Agent API at $4.50 an hour lets the same account handle live agents.
- Starter credit
- New AssemblyAI accounts receive $50 of credit for testing speech, agents and analysis together.
Against Deepgram's pain points: AssemblyAI
AssemblyAI is the switch for teams that want analysis on top of the transcript, at a lower streaming rate than Nova-3.
- Streaming cost: AssemblyAI streaming from $0.15 an hour works out near $0.0025 a minute, against $0.0077 for Nova-3 streaming on pay as you go
- Audio intelligence: AssemblyAI includes sentiment, PII redaction, entities and summaries, where Deepgram returns raw transcription only
- Language coverage: Language support is not detailed on the AssemblyAI card checked, so confirm your target languages before migrating
- Accuracy claims: AssemblyAI publishes benchmarks that favour its own models, the same pattern as Deepgram, so test both on your recordings
- Open models: AssemblyAI is a hosted API only, so teams weighing open weight models still pay per hour rather than per GPU
- Prepay terms: AssemblyAI gives $50 of credit with no prepay named on the card, compared with $4,000 a year for the Deepgram Growth discount
AssemblyAI pros and cons
Pros
- Audio intelligence is built in. Sentiment, PII redaction, entity detection, summaries and LLM analysis run on the transcript, so there is no second model to wire up the way Deepgram needs.
- Batch transcription is $0.15 an hour on Universal-2, below Deepgram's pre-recorded rate, which keeps large archives cheap to process.
- A Voice Agent API and $50 of starter credit let you test speech, agents and the analysis layer on one account.
Cons
- AssemblyAI and Deepgram each publish accuracy benchmarks that favour themselves, so the only honest comparison is your own audio.
- The advanced models cost more: Universal-3.5 Pro is $0.21 an hour and the Voice Agent API is $4.50 an hour, so insight-heavy workloads climb fast.
AssemblyAI reviews
AssemblyAI holds 4.6 on G2 from 125 reviews. AssemblyAI on G2 ↗
A reviewer highlights: "It was quick for our team to integrate, and it delivers accurate transcription results even with long audio files and conversations involving multiple speakers." (Verified user, Mid-Market, G2 Apr 2026)
Another flags: "Transcription accuracy occasionally drops with poor audio quality or strong background noise." (Gaurav R., Small-Business, G2 Jun 2026)
On pricing: "Pricing is fair, and they give you a generous free trial to test your integration without having to pay." (Verified user in film, Small-Business, G2 Aug 2026)
Choose AssemblyAI if your product needs insights from calls, sentiment, redaction and summaries, not just the text.
Speechmatics
Built for multilingual accuracy and speaker labels
Speechmatics pitches many languages and free speaker diarization.
Pre-recorded audio costs $0.24 to $0.75 an hour and real time $0.30 to $0.80 an hour by model. The Speechmatics TTS rate is $0.011 per 1,000 characters with the first million free. New accounts get $100 of credit. Speechmatics competes on language and accent coverage rather than price, so it fits global products where a caller can speak almost anything, and the free first million TTS characters soften entry cost. The tradeoff is published rates above the cheapest STT, which bites most at steady high volume, so test it on your own accents first.
Speechmatics features
- Broad languages
- Speechmatics pitches wide language coverage with accents handled in one model.
- Free diarization
- Speaker diarization comes at no extra charge, so multi speaker calls label cleanly by default.
- Model pricing
- Pre-recorded audio runs $0.24 to $0.75 an hour and real time $0.30 to $0.80 an hour, depending on model.
- Speech output
- Speechmatics text to speech costs $0.011 per 1,000 characters, with the first million characters free.
- Starter credit
- New Speechmatics accounts get $100 of credit, about 125 to 400 hours depending on model.
Against Deepgram's pain points: Speechmatics
Speechmatics answers the language complaint G2 reviewers raise most about Deepgram, with free diarization added.
- Streaming cost: Speechmatics real time runs $0.30 to $0.80 an hour, so the cheapest tier sits near $0.005 a minute and the top tier above Nova-3
- Audio intelligence: Diarization is included free, but sentiment, PII redaction and summaries are not described on the Speechmatics card checked
- Language coverage: Speechmatics leads its pitch with broad language and accent coverage, the gap G2 reviewers flag most in Deepgram
- Accuracy claims: Speechmatics publishes no independent accuracy figure, so benchmark on your own audio
- Open models: Speechmatics is a hosted service, so the open weight cost argument still applies at very large scale
- Prepay terms: Speechmatics opens with $100 of credit, half the $200 Deepgram gives, with no prepay commitment named on the card
Speechmatics pros and cons
Pros
- Speaker diarization is included at no extra charge, unlike vendors that bill it as an add-on, so multi-speaker calls label cleanly out of the box.
- New accounts get $100 of credit, more than most rivals here, enough to transcribe roughly a hundred hours before paying anything.
Cons
- Top-tier models reach $0.75 an hour pre-recorded and $0.80 real time, above the cheapest batch options, so cost matters once volume is high.
- The Speechmatics voice-agent and text-to-speech ecosystem is smaller than Deepgram's, so teams building live agents have fewer ready-made pieces to lean on.
Speechmatics reviews
Speechmatics holds 4.7 on G2 from 75 reviews. Speechmatics on G2 ↗
A reviewer highlights: "We appreciate the relative ease of integrating Speechmatics into our broadcast monitoring workflow. Customer support is reliable and provides quick responses." (Joe T., Small-Business, G2 Aug 2025)
Another flags: "It appears to fill in a whole lot of nonsense when it hears noise or background chatter, creating disconcerting false positive transcriptions." (Verified user, Mid-Market, G2 Sep 2026)
On pricing: "Functionalities and pricing tailored to your exact application and business domain, a good software stack and API, reliable transcoding." (Verified user in broadcast media, Enterprise, G2 Feb 2023)
Choose Speechmatics if you build global products where language and accent coverage, not price, is the deciding factor.
Gladia
Built for real time and batch transcription across many languages
Gladia pitches many languages, diarization and audio intelligence.
Batch Solaria costs EUR 0.16 an hour and real time EUR 0.68 an hour, with EUR 50 of free credit. Gladia leans on many languages, diarization and audio intelligence at a low hourly rate, so it suits batch heavy transcription across markets where per hour cost decides. Gladia is less proven for low latency streaming than Deepgram, and its rates are in euros, so confirm real time performance and convert euros to dollars carefully before building around it, since a low batch rate does not guarantee low streaming latency for live calls.
Gladia features
- Solaria batch
- Gladia batch transcription on Solaria costs EUR 0.16 an hour, a low hourly rate.
- Real time
- Gladia real time transcription costs EUR 0.68 an hour, far above the batch rate.
- Audio intelligence
- Diarization and audio intelligence ship with Gladia transcription, so one API covers multilingual batch work end to end.
- Free credit
- EUR 50 of free credit is enough to benchmark Gladia on your own recordings.
Against Deepgram's pain points: Gladia
Gladia is cheap for multilingual batch work, though live transcription loses that edge.
- Streaming cost: Gladia real time at EUR 0.68 an hour is about EUR 0.011 a minute, above the $0.0077 Nova-3 streaming rate
- Audio intelligence: Gladia bundles diarization and audio intelligence with the transcript, so no separate LLM step is needed for basic insight
- Language coverage: Gladia markets wide language coverage, which suits teams that hit the Deepgram language limits G2 reviewers mention
- Accuracy claims: Gladia publishes its own list of Deepgram alternatives, so treat its comparison as vendor content and test
- Open models: Gladia is hosted, though EUR 0.16 an hour for batch narrows the gap with running open weights yourself
- Prepay terms: Gladia bills in euros from EUR 50 of free credit, so a dollar budget moves with the exchange rate and no prepay is named
Gladia pros and cons
Pros
- Batch transcription is about EUR 0.16 an hour on Solaria, a low hourly rate in this list, which suits large archives in many languages.
- Coverage spans many languages with diarization and audio intelligence, so one API handles multilingual batch work end to end.
- EUR 50 of free credit lets you benchmark it on your own recordings before committing.
- Gladia holds 4.8 on G2 from 24 reviews, with reviewers praising smooth language switching and accuracy on accented speech.
Cons
- Real time is far dearer than batch at about EUR 0.68 an hour, so live use cases lose the price advantage.
- Pricing is quoted in euros, so a dollar budget moves with the exchange rate.
Gladia reviews
Gladia holds 4.8 on G2 from 24 reviews. Gladia on G2 ↗
A reviewer highlights: "Excellent multilingual real-time transcription with smooth language switching, superior accuracy on accented speech compared to competitors, and a clean API that is easy to integrate and deploy to production." (Yassine R., Small-Business, G2 January 2026)
Another flags: "The minimum maximum duration without endpointing is 5 seconds. For real-time dictation use cases, this makes the experience feel slightly slow. Competitors offer 4-second minimums, and that 1-second difference is noticeable in practice." (Yassine R., Small-Business, G2 January 2026)
On pricing: "The pricing model could be clearer for large volume enterprise use." (Faes W., User, G2 November 2025)
Choose Gladia if you run batch-heavy transcription across many languages where the per-hour rate decides.
ElevenLabs
Built for teams replacing Aura voices, with Scribe for transcription
ElevenLabs is the usual pick when Deepgram's Aura voices feel too plain, and its own voice library is broader and more expressive than Aura. Scribe adds transcription on the same account.
Scribe v2 costs $0.22 an hour and real time $0.39. Text to speech runs $0.04 to $0.08 per 1,000 characters, and Agents list at $0.08 a minute. G2 rates it 4.5 out of 5 across 1,260 reviews. ElevenLabs is the pick when Deepgram's Aura voices sound too flat and the voice itself is what matters, since its library and cloning lead on realism. Scribe covers transcription on the same account, but the tooling is built around speech synthesis first, so pure high volume STT is not where it is cheapest.
ElevenLabs features
- Scribe transcription
- ElevenLabs Scribe v2 costs $0.22 an hour for recorded audio and $0.39 an hour in real time.
- Natural voices
- ElevenLabs is the common pick when Deepgram Aura voices sound too plain for narration.
- Voice library
- Voice cloning and more than 1,000 library voices sit on the same account as Scribe.
- Speech pricing
- ElevenLabs text to speech runs $0.04 to $0.08 per 1,000 characters, above Aura-2 at $0.030.
- Agents
- ElevenLabs Agents list at $0.08 a minute for live calls.
Against Deepgram's pain points: ElevenLabs
ElevenLabs replaces Deepgram mainly on the voice side, with Scribe covering transcription on one account.
- Streaming cost: Scribe real time at $0.39 an hour is about $0.0065 a minute, slightly under the $0.0077 Nova-3 streaming rate
- Audio intelligence: Audio intelligence features are not described for Scribe in the sources checked, so plan an LLM step as with Deepgram
- Language coverage: Language support for Scribe is not detailed on the card checked, so confirm each target language before switching
- Accuracy claims: ElevenLabs publishes no Scribe accuracy benchmark, so run your own audio through the trial
- Open models: ElevenLabs is proprietary and credit billed, with G2 reviewers flagging cost, so open weights stay cheaper at large scale
- Prepay terms: ElevenLabs billing is credit based, and reviewers report credits draining quickly with failed generations still charged
ElevenLabs pros and cons
Pros
- The voices are the most natural in this list, the usual reason teams switch when Deepgram's Aura output feels flat for narration.
- Voice cloning and more than 1,000 library voices sit on the same account as Scribe transcription, so speaking and listening live together.
Cons
- Text to speech runs $0.04 to $0.08 per 1,000 characters, above Deepgram's Aura-2 at $0.030, so high-volume speech costs more.
- Billing is credit-based and reviewers say credits drain quickly at scale, with failed generations still charged.
- ElevenLabs voices can stumble on dense legal terminology and citations, a G2 reviewer reports, so specialist scripts need a listening pass.
ElevenLabs reviews
ElevenLabs holds 4.5 on G2 from 1,260 reviews. ElevenLabs on G2 ↗
A reviewer highlights: "The absolute best part is just how genuinely human the voices sound: the natural pauses, the breathing, and the way emphasis shifts." (Pradumya K., Enterprise, G2 Oct 2026)
Another flags: "The main pain point is how it handles dense legal terminology, citations and Latin maxims. The model frequently stumbles over citations." (Shreyansh P., Small-Business, G2 Oct 2026)
On pricing: "The best part about ElevenLabs is the free credit thing they have." (Rashi R., Small-Business, G2 Oct 2026)
Choose ElevenLabs if the voice itself, not the transcription price, is what your users notice.
OpenAI Whisper
Built for teams that want free, self hosted transcription
Whisper is open source and runs on your own hardware.
Whisper carries no license fee and runs on your own GPUs, so the cost is hardware and engineering rather than a per minute rate, which pays off at large scale and on privacy sensitive work where audio must stay in house. The catch is that streaming, scaling and uptime become your team's job, and for teams that would rather not run GPUs, OpenAI's own API offers hosted transcription at a per minute rate instead, which trades the self host savings for simplicity. Real time streaming with Whisper also needs extra engineering on top of the base model.
OpenAI Whisper features
- Open source
- Whisper is an open source model with no license fee, released by OpenAI.
- Self hosted
- Whisper weights run entirely on your own hardware, so sensitive audio never leaves your servers.
- Community forks
- A large community has built faster Whisper forks, so throughput improves without a vendor contract.
- Hosted option
- OpenAI's own API also offers hosted transcription for teams that do not want to run GPUs.
- Ops cost
- Whisper shifts spend to GPUs, scaling and uptime that your own team owns.
- Batch first
- Whisper is batch first, and real time streaming needs extra engineering on top.
Against Deepgram's pain points: OpenAI Whisper
Whisper is the open weight answer to Deepgram pricing, paid for in GPUs and engineering time.
- Streaming cost: Whisper has no per minute fee, but real time streaming needs extra engineering that Deepgram includes out of the box
- Audio intelligence: Whisper returns transcripts only, so sentiment, PII redaction and summaries still need a separate LLM, as with Deepgram
- Language coverage: Whisper offers broad language coverage in one model
- Accuracy claims: OpenAI publishes no accuracy comparison against paid APIs, so your own test on your own audio is the figure that counts
- Open models: Whisper is the open weight model teams most often self host, with no license fee at any scale
- Prepay terms: Whisper needs no prepay or credit at all, though GPU and operations costs replace the Deepgram invoice
OpenAI Whisper pros and cons
Pros
- The model is open source with no license fee, so at scale the only cost is the GPUs you already run.
- Language coverage is broad and the weights run entirely on your own hardware, which keeps sensitive audio in-house.
- A large community has built faster forks on top of it, so throughput keeps improving without a vendor contract.
Cons
- Whisper is batch-first. Real-time streaming needs extra engineering that a managed API like Deepgram gives you out of the box.
- You own the servers, scaling and uptime, so the free model still carries a real operations cost.
- Whisper covers speech to text only, so a team that also uses Deepgram Aura voices or Voice Agent still needs separate vendors for speech output and live calls.
- Teams that want Whisper without running GPUs pay a managed rate, such as hosted the large Whisper model at $0.005 a minute on Rev.ai, so the no license fee advantage ends once someone else runs it.
Choose OpenAI Whisper if you have spare GPUs and privacy rules that keep audio off third-party servers.
Rev.ai
Built for low cost batch transcription with a human fallback
Rev.ai offers machine and human transcription from one API, useful when some files need certified accuracy.
Reverb English costs $0.20 an hour and foreign languages $0.30, hosted the large Whisper model $0.005 a minute, and human transcription $1.99 a minute. New accounts get 5 free hours. Rev.ai pairs machine and human transcription behind one API, so it fits media, legal and research teams that mostly need fast machine output but occasionally a certified human pass. The human rate is far higher than the machine rate, so the value is having both on tap, and the 5 free hours are enough to test the machine model first.
Rev.ai features
- Machine and human
- Rev.ai serves machine and human transcription from one API, so certified files can be escalated without a second vendor.
- Reverb pricing
- Reverb English costs $0.20 an hour and foreign languages $0.30 an hour.
- Hosted Whisper
- Rev.ai hosts the large Whisper model at $0.005 a minute for teams that want open weights without GPUs.
- Human transcription
- Rev.ai human transcription costs $1.99 a minute for files that need certified accuracy.
Against Deepgram's pain points: Rev.ai
Rev.ai suits archives that sometimes need human certified transcripts, rather than live agents.
- Streaming cost: Rev.ai is batch focused and not built for real time voice agents, so streaming is not a reason to choose Rev.ai over Deepgram
- Audio intelligence: Audio intelligence features are not described on the Rev.ai card checked, so expect to add analysis yourself
- Language coverage: Rev.ai prices foreign language Reverb at $0.30 an hour, a third above the $0.20 English rate
- Accuracy claims: Rev.ai answers accuracy doubts with human transcription at $1.99 a minute, a certified fallback Deepgram lacks
- Open models: Rev.ai hosts the large Whisper model at $0.005 a minute, a managed way to use an open model without running GPUs
- Prepay terms: New Rev.ai accounts get 5 free hours to test both the machine and human paths, with no prepay named on the card
Rev.ai pros and cons
Pros
- Machine and human transcription share one API, so files that need certified accuracy can be escalated without bringing in a second vendor.
- Reverb English batch is about $0.20 an hour and hosted the large Whisper model $0.005 a minute, cheap for large English archives.
- New accounts get 5 free hours, enough to test both the machine and the human path on real files.
- At 100,000 minutes a month, Rev.ai Reverb English comes to about $333, roughly $437 under Deepgram's $770 streaming baseline.
Cons
- Rev.ai is batch-focused, not built for real-time voice agents the way Deepgram is.
- Human transcription costs $1.99 a minute, far above the machine rate, so certified files get expensive at volume.
- Rev.ai's Capterra rating covers Rev's transcription service rather than the API, so developers get no public score for the API itself.
Rev.ai reviews
Rev.ai has no public reviews for its API. The Capterra rating covers Rev's transcription service, not the API. Rev.ai on Capterra ↗
Choose Rev.ai if you are a media or legal team that mostly needs machine transcription but occasionally a certified human pass.
Deepgram alternatives by job and use case
Deepgram does three jobs. Pick a replacement for the one you use most.
Speech to text alternatives
| Need | Pick | Price |
|---|---|---|
| Phone calls at volume | PyAI Hear | $0.001 a minute |
| Audio intelligence | AssemblyAI | $0.15 an hour |
| Many languages | Speechmatics, Gladia | $0.24 an hour, EUR 0.16 an hour batch |
| Self hosted | Whisper | Free, your GPUs |
| Human option | Rev.ai | $0.20 an hour, human $1.99 a minute |
| AWS or Google shops | Amazon Transcribe, Google STT | Check cloud pricing |
Text to speech alternatives to Aura
| Voice | Price |
|---|---|
| Deepgram Aura-2 | $0.030 per 1K characters |
| PyAI Speak | $0.04 a minute, about $0.044 per 1K characters |
| ElevenLabs Flash | $0.04 per 1K characters |
| Cartesia | From $5 a month |
| Speechmatics TTS | $0.011 per 1K characters |
Voice agent alternatives
| Option | Per minute |
|---|---|
| Deepgram Voice Agent | $0.075 standard, plus telephony |
| PyAI Omni | $0.05/min + $0.01 telephony |
| AssemblyAI Voice Agent | $4.50 an hour, about $0.075 a minute |
| Retell AI | About $0.11 to $0.13 |
Free and open source Deepgram alternatives
Whisper is the main open source route, and open weight models can run far cheaper than proprietary APIs at scale. The trade off is servers, scaling and streaming work. Hosted free credit is the quickest way to compare: $200 on Deepgram, $100 on Speechmatics, $50 on AssemblyAI, EUR 50 on Gladia, 5 hours on Rev.ai, and $50 on PyAI with no credit card.
When Deepgram is still the right pick
Deepgram stays a strong choice for real time products that need fast interim results, broad language support and one vendor for speech to text, voices and agents. The Deepgram $200 credit with no expiry also makes testing cheap. Teams transcribing phone calls at volume, or wanting built in audio intelligence, gain the most from comparing.
How to switch from Deepgram in 4 steps
Vendor benchmarks conflict, so your own audio decides.
Step 1 of 4
Pick test audio
Choose 50 to 100 real files, including noisy calls and accents.
Step 2 of 4
Run every shortlisted API
Send the same files to each provider and save the transcripts.
Step 3 of 4
Score accuracy and latency
Compare word errors on names, numbers and jargon, and measure time to first partial.
Step 4 of 4
Switch with a fallback
Move traffic in stages and keep Deepgram as a fallback until error rates match.
Which Deepgram alternative to pick
The best Deepgram alternatives depend on the job. AssemblyAI adds audio intelligence, Speechmatics and Gladia cover more languages, ElevenLabs replaces Aura voices, and Whisper is free if you host it. PyAI covers phone audio transcription at $0.001 a minute, voices at $0.04, and one engine phone agents at $0.05/min + $0.01 telephony. A free sandbox key comes with $50 of starter credit, and no credit card is asked for.
More Deepgram comparisons
Deepgram alternatives FAQs
Common questions from developers replacing Deepgram.
Is AssemblyAI better than Deepgram?
For audio intelligence, often yes. AssemblyAI bundles sentiment, PII redaction, entities and LLM analysis, so you skip a separate model. AssemblyAI Universal-2 costs $0.15 an hour. Deepgram is known for low streaming latency and voice agents. Each vendor publishes benchmarks favoring itself, so test on your own audio.
How expensive is Deepgram?
Nova-3 streaming lists at $0.0077 a minute on pay as you go. Pre-recorded audio is $0.0043 a minute, Aura-2 voices cost $0.030 per 1,000 characters, and the Voice Agent API is $0.075 a minute on the standard tier. Sign up and Deepgram adds $200 of credit that does not expire. Growth plans cut both by up to 20%.
Is Deepgram better than Whisper?
For real time, usually yes. Deepgram streams with low latency as a managed API. Whisper is open source and free to self host, covers many languages, and suits batch work if you run your own GPUs. Self hosted Whisper trades a per minute rate for your own GPU and engineering cost on batch work.
What is a lower cost Deepgram alternative?
PyAI Hear and AssemblyAI cost less per minute than Deepgram. PyAI Hear costs $0.001 a minute, or $0.0005 in batch, and AssemblyAI Universal-2 is $0.15 an hour, about $0.0025 a minute. Self hosted Whisper has no license fee but needs your own servers.
What is a good free speech to text tool?
Whisper is a widely used free option. Whisper is open source and runs locally. Hosted free credit includes $200 on Deepgram, $100 on Speechmatics, $50 on AssemblyAI, 5 hours on Rev.ai, and $50 on PyAI with no credit card. The free credits cover a test of each tool on the same recordings.
Which Deepgram alternative suits voice agents?
PyAI Omni is the pick for voice agents because it runs the whole call on one speech to speech engine at $0.05/min + $0.01 telephony, against $0.075 for Deepgram's Voice Agent API before telephony. AssemblyAI and Speechmatics suit teams that prefer transcription to plug into their own stack.
Test PyAI on your own audio and calls.
Get a sandbox key and hear the output in under a minute. Sandbox keys are free, with $50 in starter credit and no credit card required. Speak is $0.04/min, Cast $0.02/min, Hear $0.001/min and Omni $0.05/min + $0.01 telephony.