Voxtral alternatives in 2027 for speech to text
Voxtral alternatives compared for speech to text: hosted transcription APIs, open models and per minute pricing, with real 2027 numbers.

On this page
Why developers look for a Voxtral alternative
Voxtral is the provider's open weights speech model line, a transcription and audio model you can self host or call through the provider's API, launched as an open alternative to proprietary speech APIs. Voxtral is respected for accuracy. Developers look for an alternative when self hosting, the hosted rate or model churn gets in the way.
Self host burden
The open Voxtral models need your own GPUs, about 9.5 GB of GPU memory for the Mini model, so running them locally is real engineering before a single transcript.
Smaller model fragility
Guides note the Mini model has less tolerance for noise, overlapping speech and uneven audio, and works best on short, controlled clips.
Rates differ by model
The provider lists Voxtral Mini Transcribe 2 at $0.003 a minute at the standard rate. Live transcription uses a separate realtime model and rate.
Fast model churn
Voxtral is young, with Mini, Mini 2 and Transcribe 3 already and early versions deprecated, which makes it a moving target for integrators.
Voxtral alternatives at a glance
Ranked on coverage, hosting and price, from each vendor's own page and dated review data gathered 21 October 2026.
| Tool | Built for | Features | Covers | Hosting | Price (2026) |
|---|---|---|---|---|---|
| PyAI | Teams that want hosted transcription with a full voice stack | Hosted Hear speech to text, Speak voices and the Omni phone agent, OpenAI compatible | STT, TTS, speech to speech$50 credit, no credit card | Hosted | Hear $0.001/min, about $0.06/hr |
| OpenAI Whisper | Teams wanting a known transcription baseline | Open weights plus a hosted transcribe API, wide language spread | STT | Self host or hosted | Hosted about $0.006/min |
| Deepgram | Developers who want fast, low cost hosted speech to text | Nova-3 transcription, Aura voices and a Voice Agent API | STT, TTS | Hosted | About $0.0043/min batch |
| AssemblyAI | Teams building analytics on their transcripts | Transcription with summaries, sentiment and topic detection | STT | Hosted | Universal-2 $0.15/hr |
| Google Cloud | Teams whose data and access controls sit in Google Cloud | Managed transcription tied into Google Cloud, broad languages | STT | Hosted | Real time about $0.016/min |
| Speechmatics | Teams needing broad languages or on premise hosting | Accent accurate transcription, on premise and edge deployment | STT | Hosted or on premise | From about $0.13/hr |
| ElevenLabs | Teams that want transcription alongside expressive voices | Scribe transcription with natural text to speech and cloning | STT, TTS | Hosted | Credit based, STT and TTS |
| Voxtral (for comparison) | Teams that want to self host an open speech model | Open speech to text weights, with a non commercial voice model | STT, TTS | Self host or the Voxtral provider API | Free weights, API about $0.003/minStandard Transcribe 2 rate |
Voxtral pricing vs alternatives
Voxtral is free to self host on your own GPUs, or called through the provider's API near $0.003 a minute for Transcribe 2 at the standard rate, so compare it with other hosted file transcription rates. The chart lines up hosted transcription cost a minute, with PyAI Hear at $0.001. Self hosting the open weights has no per minute fee but needs your own hardware.
Hosted speech to text, cost per minute
Transcribe for a tenth of a cent a minute, hosted
Headline hosted transcription cost a minute, lower is cheaper. Each bar is that tool's hosted price per minute of audio. Voxtral and Whisper are also free to self host on your own GPUs.
- PyAI HearHosted, published$0.001 / min
$0.001 per minute, about $0.06 an hour, hosted and OpenAI compatible, with no GPUs to run
- AssemblyAIUniversal-2 asyncAbout $0.0025 / min
$0.15 an hour, about $0.0025 a minute, hosted with an audio intelligence suite
- Voxtralthe Voxtral provider hosted APIAbout $0.003 / min
About $0.003 a minute for Transcribe 2 on the provider's API, or free to self host the open weights on your own GPUs
- DeepgramNova-3 batch$0.0043 / min
$0.0043 a minute batch, with $200 of starting credit, hosted
- OpenAI WhisperHosted transcribe$0.006 / min
$0.006 a minute hosted, about $0.36 an hour, or free to self host
PyAI figure is the published pyai.com rate, 21 October 2026. Competitor rates come from vendor pages and dated third party trackers. The provider publishes a standard rate for Voxtral Mini Transcribe 2. Verify the price for your model and processing mode before relying on it.
Voxtral alternatives reviewed
The seven picks are hosted speech APIs and open models that cover what Voxtral does. PyAI ranks first for hosted per minute transcription plus a full voice stack. OpenAI Whisper is the open baseline, Deepgram, AssemblyAI, Google and Speechmatics are hosted, and ElevenLabs covers the voice side.
PyAI
Built for teams that want Voxtral style transcription hosted for them, plus TTS and a phone agent billed per minute
PyAI is voice infrastructure. Voxtral is one open model line you self host or call through the Voxtral provider. PyAI covers the same transcription job as a hosted API, with Hear transcribing at $0.001 a minute, then puts Speak text to speech and the Omni speech to speech phone agent on the same invoice. The API is OpenAI compatible, so code written for Whisper or the OpenAI transcription endpoint moves with a base URL change, with no GPUs to provision. PyAI is not an open model to self host, so a team that wants the open weights keeps Voxtral.
Hear is $0.001 a minute, about $0.06 an hour, matching original Voxtral's floor and below its Transcribe 2 hosted API near $0.003. Speak is $0.04 a minute, and Omni runs a phone agent at $0.05 for speech plus $0.01 telephony. A trial needs no contract: new accounts get $50 of credit with no credit card.
Being newly launched, PyAI shows no public rating on G2 or Capterra yet, so accuracy is easy to check on a sandbox key against your own audio.
PyAI features
- Hear
- Speech to text at $0.001 a minute, hosted, with a first partial near 200 ms.
- Speak and Cast
- Text to speech at $0.04 a minute, and long form at $0.02.
- Omni
- Speech to speech phone agents at $0.05 a minute, plus $0.01 for the line.
- OpenAI compatible
- Teams already on the OpenAI audio API point their client at PyAI and keep the same request shape.
What changes when moving from Voxtral to PyAI4 differences
| On Voxtral | On PyAI |
|---|---|
| Self host on GPUs, or the Voxtral provider API at about $0.003 | Hosted at $0.001 a minute, no infrastructure |
| Speech model only | Hear, Speak and the Omni agent together |
| the Voxtral provider SDK or self host | OpenAI compatible, base URL change |
| Fast model churn and deprecations | One stable per minute API |
PyAI pricing
Hear
$0.001/min
Speech to text
Speak
$0.04/min
Text to speech
Credit
$50
No credit card
Hear meters audio by the minute with no monthly plan, sandbox keys never bill, and every new account opens with $50 of free credit, no credit card needed. See common setups in PyAI solutions.
PyAI pros and cons
Pros
- Hosted Hear matches original Voxtral's $0.001 a minute floor and undercuts its $0.003 Transcribe 2 API, with no GPUs to run.
- OpenAI compatible, so code written for Whisper or the OpenAI transcription format moves across with a base URL change.
- The same account also runs Speak at $0.04 a minute and the Omni phone agent, which a speech only model like Voxtral does not bundle.
- Billed per minute with $50 of starter credit and no credit card, so a first transcription test costs cents.
Cons
- Not open weights, so a team that wants to run the model on its own GPUs stays with Voxtral.
- No public G2 or Capterra rating yet, as it is a recent launch.
- The 200ms first partial is from PyAI's own benchmarks rather than an outside test.
PyAI reviews
PyAI carries no G2 or Capterra score yet, so rather than quote a number we cannot source this tab stays empty. The fair test is a sandbox key, free with $50 of credit and no credit card, run against a batch of your own audio.
Choose PyAI if you want hosted, per minute transcription plus text to speech and a phone agent, and you would rather not run a model on your own hardware.
OpenAI Whisper
Built for the open and hosted transcription baseline
OpenAI Whisper is the transcription baseline most teams measure against. Like Voxtral, Whisper ships as open weights you can self host, but it is also offered as a hosted API and covers a wide spread of languages, which is why many developers pick it over Voxtral when they want a well understood model without standing up their own GPUs.
The hosted transcribe models cost about $0.006 a minute, with a cheaper mini model near $0.003, while the open Whisper weights are free under the permissive MIT license, so self hosting costs only the compute you provide. The catch is that plain Whisper has no streaming mode and no built in diarization, so live captions or speaker labels need extra tooling around the model, which keeps Whisper better as an accuracy reference than a finished real time service.
OpenAI Whisper features
- Open weights
- OpenAI publishes the Whisper weights under a permissive license, so self hosting costs only compute.
- Hosted API
- OpenAI's hosted transcribe models cost about $0.006 a minute, with a cheaper mini model near $0.003.
- Start managed, move later
- Whisper runs both hosted and self hosted, so a team can start on the API and bring the model in house later.
- Familiar format
- Most transcription tooling already supports Whisper output.
- Reviewer view
- G2 rates Whisper 4.4 from 52 reviews, citing accuracy, documentation and noise handling, with slow processing as the main con.
Against Voxtral's pain points: OpenAI Whisper
OpenAI Whisper is the baseline Voxtral gets measured against, with an easier hosted path.
- Self hosting: Whisper also needs your own compute when self hosted, but the hosted API at about $0.006 a minute removes the GPU work entirely
- Noisy audio: G2 reviewers name noise handling among Whisper's strengths, where Voxtral Mini works best on short, controlled clips
- Price visibility: OpenAI shows Whisper rates of about $0.006 and $0.003 a minute, and the standard rate for Voxtral Mini Transcribe 2 is $0.003 a minute
- Model churn: Whisper version history is not covered in the sources checked, though tooling support for its format is widespread
- Commercial TTS: Whisper is a transcription model in the sources checked, so commercial voice output needs a separate provider
- Language coverage: A transcription app developer reports the large v3 Whisper model Turbo covers far more languages than Voxtral and is easier to ship
OpenAI Whisper pros and cons
Pros
- Open weights to self host free, plus a hosted endpoint, so you can start managed and move in house later.
- Wide language spread and strong accuracy, the reasons reviewers reach for it most.
- A familiar format that most transcription tooling already supports.
Cons
- The hosted rate of about $0.006 a minute is dearer than Deepgram or AssemblyAI on batch.
- Reviewers flag slow processing on long files, and there are few built in analytics features.
OpenAI Whisper reviews
G2 rates the Whisper speech to text product 4.4 out of 5 from 52 reviews, praising accuracy and multilingual coverage while noting slow processing. OpenAI Whisper on G2 ↗
Choose OpenAI Whisper if you want a well known baseline you can run hosted now and self host later.
Deepgram
Built for fast, low cost hosted speech to text
Deepgram is a developer speech platform built around its Nova-3 model for transcription, with Aura text to speech and a Voice Agent API alongside, which positions it as a hosted, low cost route rather than a model like Voxtral you run yourself. Deepgram leans hard on real time speed, so streaming captions and live agents are a core use case rather than an afterthought.
Nova-3 batch costs about $0.0043 a minute while streaming runs $0.0077 on pay as you go, and new accounts start with $200 of credit, checked 21 October 2026. The tradeoff is that the $0.0043 rate applies only to batch work, so always on streaming costs closer to double, and the audio is cloud only, without the self host option Voxtral gives teams that have data residency rules.
Deepgram features
- Nova-3 transcription
- Deepgram Nova-3 batch costs about $0.0043 a minute, and streaming runs $0.0077 on pay as you go, checked 21 October 2026.
- Aura voices
- Deepgram sells its own Aura text to speech on the same platform as transcription.
- Voice Agent API
- Deepgram adds a Voice Agent API alongside speech to text and text to speech.
- Starting credit
- New Deepgram accounts get $200 of starting credit.
- Real time path
- Deepgram's strong real time path means no GPU cluster like a self hosted Voxtral deployment.
- Add on billing
- Deepgram bills diarization and other extras on top of the base per minute rate.
Against Voxtral's pain points: Deepgram
Deepgram is the hosted, low cost route away from running Voxtral, with voice output included.
- Self hosting: Deepgram is hosted, so the 9.5 GB of GPU memory Voxtral Mini needs never becomes your problem
- Noisy audio: Deepgram holds 4.6 on G2 from 482 reviews for accuracy and speed, while inaccuracy also appears among the cons
- Price visibility: Deepgram lists $0.0043 a minute batch and $0.0077 streaming on its pricing page, while Voxtral Mini Transcribe 2 is $0.003 a minute at the standard rate
- Model churn: Deepgram's model release pace is not covered in the sources checked, so confirm version support terms before integrating
- Commercial TTS: Aura text to speech is sold as a hosted Deepgram service, a route around the non commercial Voxtral voice license
- Language coverage: G2 reviewers flag narrower language coverage on Deepgram than the broadest engines
Deepgram pros and cons
Pros
- Hosted batch near $0.0043 a minute with a strong real time path, so no GPU cluster like self hosted Voxtral.
- Text to speech and a Voice Agent API sit on the same platform for both call directions.
Cons
- Diarization and other extras bill on top of the base per minute rate.
- G2 reviewers flag narrower language coverage than the broadest engines.
Deepgram reviews
Deepgram holds 4.6 on G2 from 482 reviews. Deepgram on G2 ↗
A reviewer highlights: "Deepgram is really a developer-friendly solution and is extremely fast if you compare it against other vendors like OpenAI Whisper and Speechmatics." (Diego P., Enterprise, G2 Aug 2026)
Another flags: "Although the accuracy is very good, it can sometimes be a bit difficult to adjust the models for very specific cases or when there is technical jargon." (Verified user, Small-Business, G2 Aug 2026)
On pricing: "It supports word recognition and speaker diarization, and it also has a very good price offering comparatively." (Akshay M., Mid-Market, G2 Aug 2026)
Choose Deepgram if you want fast, cheap hosted transcription with a matching text to speech side.
AssemblyAI
Built for transcripts enriched with summaries, sentiment and topic detection
AssemblyAI is a hosted transcription API with a broad audio intelligence layer on top, covering summarization, sentiment and topic detection, so it fits teams that build analytics on transcripts rather than run a model themselves. The AssemblyAI intelligence suite, more than the raw transcription rate, is the reason teams reach for AssemblyAI over a bare model like Voxtral.
Universal-2 async is $0.15 an hour and Universal-3.5 Pro $0.21, billed per hour of audio, with add ons for diarization and specialist models and roughly $50 of free credit to start. The tradeoff is that the per minute rate sits above the cheapest speech to text APIs, and the service is cloud only, so a team that chose Voxtral to self host for privacy would give that up here.
AssemblyAI features
- Audio intelligence
- AssemblyAI adds summarization, sentiment and topic detection on top of transcripts.
- Universal-2 rate
- AssemblyAI Universal-2 async costs $0.15 an hour of audio, about $0.0025 a minute.
- Pro model
- AssemblyAI Universal-3.5 Pro costs $0.21 an hour for higher accuracy needs.
- Free credit
- AssemblyAI gives roughly $50 of free credit to start.
- Specialist add ons
- AssemblyAI bills diarization and specialist models on top of the base rate.
Against Voxtral's pain points: AssemblyAI
AssemblyAI suits teams that build analytics on transcripts rather than run a model.
- Self hosting: AssemblyAI is a hosted API, so no GPUs or model serving sit with your engineers
- Noisy audio: AssemblyAI reviewers praise accuracy and clean transcripts, with 4.6 on G2 from 125 reviews
- Price visibility: AssemblyAI publishes $0.15 and $0.21 an hour rates, though reviewers note cost at scale once add ons stack
- Model churn: Two named AssemblyAI models, Universal-2 and Universal-3.5 Pro, are priced side by side, while deprecation terms are not covered in the sources checked
- Commercial TTS: AssemblyAI is transcription and audio intelligence only in the sources checked, so voice output needs another vendor
- Language coverage: Limited language support appears among the G2 cons for AssemblyAI, so test the languages you need
AssemblyAI pros and cons
Pros
- Summarization, sentiment and topic models turn a transcript into insight without a second vendor.
- Universal-2 async at $0.15 an hour, about $0.0025 a minute, is among the cheapest hosted rates here.
- A 4.6 G2 score from 125 reviews, with accuracy and clean transcripts the recurring praise.
Cons
- Streaming and the Pro model cost more than the async base rate.
- Diarization and specialist models are billed on top, and reviewers note cost at scale.
- AssemblyAI runs in the cloud only, so audio a Voxtral team once kept on its own servers would now be processed off site.
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 you are building analytics on top of transcripts and want summaries and sentiment built in.
Google Cloud Speech to Text
Built for managed cloud speech to text at scale
Google Cloud Speech to Text is the hyperscaler transcription service, with wide language support and deep ties into the rest of Google Cloud, which makes it a managed route for teams that would rather not run a model like Voxtral themselves. For shops already billing through Google Cloud, transcription lands on the same invoice and the same access controls, which is often the deciding factor.
Real time transcription is about $0.016 a minute, dropping at volume, with a cheaper dynamic batch tier near $0.003, a small free tier and Google Cloud credit for new accounts. The tradeoff is that the real time rate runs well above lean speech to text APIs, and the service is tightly coupled to Google Cloud, so moving it into another stack brings more setup than a single OpenAI compatible endpoint.
Google Cloud Speech to Text features
- Hyperscaler service
- Google Cloud Speech to Text integrates deeply with the rest of Google Cloud.
- Real time rate
- Google real time transcription is about $0.016 a minute, dropping at volume.
- Dynamic batch
- Google's dynamic batch option drops bulk transcription to roughly $0.003 a minute.
- Credits
- New Google Cloud accounts get starter credit plus a small free transcription allowance.
Against Voxtral's pain points: Google Cloud Speech to Text
Google Cloud Speech to Text is the managed route for teams already inside Google Cloud.
- Self hosting: Google runs the model as a managed service, so the Voxtral GPU setup disappears
- Noisy audio: Google scores 4.5 on G2 from 316 reviews for ease of use and accuracy, though accuracy issues also appear in the cons
- Price visibility: Google publishes per minute tiers, yet reviewers call the tiered, per channel billing complex
- Model churn: Google model versioning is not described in the sources checked for this page
- Commercial TTS: Google Cloud Speech to Text covers transcription only on this page, so voice output is a separate purchase
- Language coverage: Wide language support is one of the main reasons teams pick Google over a smaller Voxtral model
Google Cloud Speech to Text pros and cons
Pros
- Wide language support and tight integration with the rest of a Google Cloud stack.
- A dynamic batch tier near $0.003 a minute keeps bulk transcription cheap.
- A 4.5 G2 score from 316 reviews, with ease of use and accuracy the common praise.
- At 2 million minutes a month and up, one third party source puts Google real time near $0.004 a minute, a quarter of the $0.016 entry rate.
Cons
- Standard real time at $0.016 a minute is dearer than Deepgram or AssemblyAI.
- Reviewers call the tiered, per channel billing complex, and it fits best inside Google Cloud.
Google Cloud Speech to Text reviews
Google Cloud Speech to Text holds 4.5 on G2 from 316 reviews. Google Cloud Speech to Text on G2 ↗
A reviewer highlights: "The accuracy is genuinely impressive, especially when dealing with mixed-language audio. We run a logistics platform that operates across Arabic, English, and Urdu, and the transcription quality holds up well even with regional accents and background noise. The automatic punctuation and speaker diarization save a lot of manual cleanup time." (Muhammed A., Mid-Market, G2 July 2026)
Only one public G2 review was available to quote for Google Cloud Speech to Text here, so a single verbatim review is shown rather than inventing a second voice.
Choose Google Cloud Speech to Text if you already run your stack in Google Cloud and want a managed engine with a cheap batch tier.
Speechmatics
Built for broad languages and on premise
Speechmatics is a speech to text API known for wide language coverage and strong accuracy on accented speech, and it offers on premise and edge deployment, which speaks to the same teams that chose Voxtral for its run anywhere, open nature. Speechmatics deployment flexibility, more than price, is what separates Speechmatics from pure cloud APIs.
Rates run from about $0.13 an hour for batch multilingual up to higher real time tiers, with a free allowance and an opt in model training discount for teams willing to share data. The tradeoff is that the headline batch price does not cover the pricier real time and on premise tiers, so the real cost depends on deployment, and standing up an on premise install is heavier work than calling a hosted endpoint.
Speechmatics features
- Accent accuracy
- Speechmatics has built its name on accented speech accuracy across a broad language range.
- On premise and edge
- Speechmatics offers on premise and edge deployment, the run anywhere option Voxtral users value.
- Batch rate
- Speechmatics rates start at about $0.13 an hour for batch multilingual, with higher real time tiers.
- Discounts
- Speechmatics lowers entry cost with a free allowance and a discount for opting into model training.
Against Voxtral's pain points: Speechmatics
Speechmatics covers the self host use case without your team running Voxtral.
- Self hosting: Speechmatics on premise and edge deployment keeps audio local while the vendor supplies the model
- Noisy audio: Strong accuracy on accented speech is the Speechmatics specialty, with 4.7 on G2 from 75 reviews
- Price visibility: Speechmatics pricing is often quoted rather than listed, so budgets need a sales call, while Voxtral publishes model-specific rates
- Model churn: Speechmatics release history is not covered in the sources checked for this page
- Commercial TTS: Voice output from Speechmatics is not described in this page's sources, so pair it with a licensed text to speech provider
- Language coverage: Speechmatics covers one of the broadest language and accent ranges among these alternatives, beyond a young Voxtral line
Speechmatics pros and cons
Pros
- Among the widest language and accent support in this set, from about $0.13 an hour on batch multilingual.
- On premise and edge deployment covers the self host use case without you running Voxtral.
Cons
- Real time and higher tiers cost well above the batch floor.
- Pricing is often quoted rather than listed, so budgets need a sales call.
- A G2 reviewer from September 2026 reports false positive transcripts when Speechmatics hears noise or background chatter, so test it on noisy recordings.
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 need very broad language or accent coverage, or on premise hosting a cloud only API cannot give.
ElevenLabs
Built for transcription and expressive text to speech
ElevenLabs meets Voxtral on the voice side, pairing its Scribe speech to text with natural sounding text to speech and voice cloning, and the Voxtral provider itself benchmarks Voxtral Small against ElevenLabs Scribe on transcription. The draw is having both directions of audio, capture and generation, under one account rather than stitching two vendors together.
Plans run from Starter at $6 a month to Creator at $22, Scale at $299 and Business at $990, and Scribe transcription draws on the same credit pool as the text to speech rather than a separate per minute rate, so verify current credit costs on elevenlabs.io before committing. The tradeoff is that a shared credit pool makes heavy transcription harder to price than a flat per minute API, and the real strength here is generation, not low cost bulk speech to text.
ElevenLabs features
- Scribe transcription
- ElevenLabs Scribe handles speech to text, and the Voxtral provider compares Voxtral Small against Scribe.
- Natural voices
- G2 reviewers single out ElevenLabs for natural sounding text to speech and voice cloning.
- Shared credit pool
- Scribe transcription draws on the same ElevenLabs credit pool as text to speech, not a separate per minute rate.
- Plan ladder
- ElevenLabs tiers climb from $6 Starter and $22 Creator to $299 Scale and $990 Business each month.
- Review base
- ElevenLabs holds 4.5 on G2 from 1,260 reviews and 4.7 from 25 on Capterra, a review base of more than 1,200.
- Both sides of audio
- One ElevenLabs plan covers transcription and voices for teams that need both.
Against Voxtral's pain points: ElevenLabs
ElevenLabs matches Voxtral on both transcription and voice, billed from one credit pool.
- Self hosting: ElevenLabs is hosted only in the sources checked, so there are no GPUs to run and no weights to own
- Noisy audio: Scribe performance on noisy audio is not reported in the sources checked, so test it with your own recordings
- Price visibility: ElevenLabs credit based billing is harder to predict than a flat per minute API, so verify credit costs on elevenlabs.io
- Model churn: ElevenLabs model update cadence is not covered in the sources checked for this page
- Commercial TTS: ElevenLabs sells text to speech and cloning on paid plans from $6 a month, so check its license terms against the non commercial Voxtral voice model
- Language coverage: ElevenLabs language coverage is not detailed in the sources checked, so confirm the languages your audio uses
ElevenLabs pros and cons
Pros
- Natural sounding text to speech and cloning that G2 reviewers praise most, which is why the Voxtral provider benchmarks Voxtral Small against Scribe.
- Scribe transcription and the voices share one plan, handy for a team that needs both sides of audio.
- A 4.5 G2 score from 1,260 reviews and 4.7 from 25 on Capterra, a large review base in this set.
Cons
- Reviewers say monthly credits run out quickly, so heavy transcription gets expensive.
- Credit based billing makes cost harder to predict than a flat per minute API like Deepgram or PyAI.
- A G2 reviewer from October 2026 says the model stumbles over dense legal terminology and citations, so test it on specialist vocabulary first.
- Scale at $299 a month is about 13 times Creator at $22, so cost climbs sharply once usage outgrows the entry plans.
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 you want expressive text to speech and cloning alongside transcription, on one plan.
Voxtral alternatives by need and use case
Pick by whether you want a hosted API or an open model, and the job.
Pick by job
| Need | Pick | Why |
|---|---|---|
| Hosted full voice stack | PyAI | STT, TTS and agents |
| Open model to self host | OpenAI Whisper | Permissive license, broad languages |
| Low cost hosted STT | PyAI, AssemblyAI | $0.001 to $0.0025 a minute |
| Audio intelligence | AssemblyAI | Summaries and sentiment |
| On premise | Speechmatics, Whisper self host | On premise or self host |
| Whisper compatible | PyAI | Base URL change |
Run it or call it
| Hosted API (PyAI, Deepgram, AssemblyAI, Google) | Open model (Voxtral, Whisper weights) | |
|---|---|---|
| Infrastructure | None, the vendor runs it | You run GPUs |
| Cost | Per minute | Compute only, no license fee |
| Start | Sign up | Install and run |
| Best for | Production, no ops | Research and full control |
Hosted transcription per minute
| Tool | Per minute |
|---|---|
| PyAI Hear | $0.001, about $0.06 an hour |
| AssemblyAI | About $0.0025 |
| Voxtral API | About $0.003 |
| Deepgram | About $0.0043 batch |
| OpenAI Whisper | $0.006 hosted |
Hosted vs open source Voxtral alternatives
The biggest choice is whether you run the model. Open models such as Voxtral and the Whisper weights are free of a per minute fee, but you provision GPUs and run inference, which suits research and teams that want full control of the model. Hosted APIs such as PyAI, Deepgram, AssemblyAI and Google charge per minute and run the infrastructure for you, which suits a product that wants a clear rate, an SLA and no GPUs to manage.
Because PyAI is OpenAI compatible, teams often move the transcription calls with a base URL change, then add Speak and the Omni engine for the text to speech and phone agent Voxtral does not bundle.
When Voxtral is still the right pick
Voxtral stays a strong choice for teams that want an open weights speech model they can run on their own hardware, with Apache 2.0 licensing for the speech to text models and good accuracy, and who have the GPUs and engineering to deploy it. Teams that want a hosted per minute rate, a full voice stack, an OpenAI compatible path, or no infrastructure to manage gain the most from the alternatives above.
How to move off Voxtral in 4 steps
Moving the Voxtral transcription calls before anything else keeps the first change small and easy to test.
Step 1 of 4
Map your speech jobs
List where you use Voxtral for transcription, and whether you self host or call the Voxtral provider API.
Step 2 of 4
Point at PyAI
Move the transcription calls to PyAI with a base URL change, since Hear is OpenAI compatible.
Step 3 of 4
Check accuracy and cost
Compare accuracy and per minute cost on your own audio against your Voxtral setup.
Step 4 of 4
Add TTS and the agent
Once Voxtral transcription runs on Hear, add Speak for spoken replies or the Omni engine for a phone agent.
Which Voxtral alternative to pick
The best Voxtral alternative depends on whether you want hosted or self hosted. OpenAI Whisper is the open baseline, Deepgram and AssemblyAI lead on hosted transcription, Google fits its cloud, and Speechmatics adds on premise. For a hosted per minute rate with a full voice stack, PyAI Hear is $0.001 a minute, matching Voxtral's floor, and it is OpenAI compatible. Test keys never bill, and you can start without a credit card on $50 of credit.
More speech API comparisons
Voxtral alternatives FAQs
Common questions from developers comparing speech models and APIs.
What are the alternatives to Voxtral?
OpenAI Whisper, Deepgram, AssemblyAI, Google Cloud Speech to Text and Speechmatics are the main speech to text alternatives, ElevenLabs Scribe competes on transcription, and PyAI runs per minute speech to text plus a full voice stack. Voxtral is the provider's open weights speech model line.
How much does Voxtral cost?
Voxtral's open weights are free to self host under Apache 2.0, so you pay for your own GPUs. the provider's hosted API starts around $0.003 a minute for Transcribe 2 at the standard rate. Check the rates for your model and processing mode. PyAI Hear is $0.001 a minute hosted.
Is Voxtral free?
The Voxtral speech to text models are open weights under Apache 2.0, so free to self host, but you provision GPUs to run them, about 9.5 GB of GPU memory for the Mini model. the provider's hosted API is paid per minute, and the text to speech model is non commercial licensed.
What is a lower cost Voxtral alternative?
On hosted speech to text, PyAI Hear is $0.001 a minute, AssemblyAI about $0.0025 and Deepgram $0.0043, while Voxtral's hosted API is near $0.003. Self hosting the open Voxtral or Whisper weights has no per minute fee but needs your own GPUs.
Is there an OpenAI compatible Voxtral alternative?
Yes. PyAI is OpenAI compatible, so code written for Whisper or the OpenAI transcription endpoint works with a base URL change, hosted with no GPUs to run. OpenAI's own transcribe models and self hosted Whisper are other options for teams leaving Voxtral.
Can PyAI handle text to speech and phone agents too?
Yes. Alongside Hear for speech to text at $0.001 a minute, PyAI runs Speak for text to speech at $0.04 a minute and Omni, a speech to speech phone agent at $0.05 a minute plus $0.01 telephony, the line included. Voxtral is a speech model you call or self host, not a full voice 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.