Snynet Solution Logo
MON - SUN: 10 AM - 6 PM
+60 11 5624 8319

Blog

Speech-to-text apps: Microsoft vs Google - which is the best for dictation?

Image Description

Speech-to-text software has come a long way in recent years. Much of the gains in speed and accuracy are thanks to improvements in artificial intelligence, which undergirds these apps.

So, it should come as little surprise that two of the biggest names in AI—Microsoft and Google—are also major players in developing voice to text apps. Microsoft Azure Speech Service and Google Cloud Speech-to-Text are leading platforms for voice typing, transcription, and productivity.

But when push comes to shove and you have to choose one of these platforms over the other, which is better? In this guide, we’ll compare the Microsoft and Google speech-to-text apps to help you decide.

Features

Microsoft Azure Speech Service and Google Cloud Speech-to-Text overlap if you need basic audio transcription. But for more advanced voice dictation applications, the two platforms have different strengths. 

Google’s software stands out for its multi-language support. Speech-to-Text is capable of transcribing audio in any of 120 languages to text. By comparison, Microsoft’s speech to text software only supports 29 languages at this time. Google’s platform will even automatically detect the language of the recording and will recognize proper nouns so that you don’t have to worry about formatting and capitalization later on.

Google Cloud Speech-to-Text supports punctuation and recognizes multiple speakers in recordings. (Image credit: Google)

Microsoft Azure Speech Service is more feature-rich when it comes to getting your transcription exactly right. You can feed the software a custom speech model to help you improve accuracy for a single speaker or for speakers with a regional accent. Or, Speech Service supports acoustic models that you can use to cancel out noise in your recordings. This is especially helpful if you frequently experience audio noise in a conference room or over a headset.

Speech Service’s API also enables you to code real-time feedback. So, if the software is having trouble recognizing words, it could prompt the speaker to talk more slowly or clearly to achieve better results.

Both Microsoft and Googles’ platforms automatically detect when there are multiple speakers in a recording. So, you can easily use either of these speech-to-text apps for transcribing meetings and conference calls.

Performance

For straightforward audio transcription, Microsoft Azure Speech Service tends to perform better than Google Cloud Speech-to-Text. The difference is that Microsoft’s software uses AI to make sure that what it’s transcribing makes linguistic sense. Since this software can accept custom speech models, it also handles accents, lisps, and other speech impediments significantly better than Google’s Speech-to-Text platform.

Google largely sticks to recognizing words based on their audio signatures and stringing them together. This means that when the software is struggling with audio quality or interpreting an accent, the transcription quality can suffer quite a bit.

All that said, getting better results from Microsoft’s software is dependent on using high-quality speech and acoustic models. If you skip this step, you may find that the two platforms are much more comparable in their accuracy when transcribing difficult recordings. Feeding Speech Service poor models can also hurt your transcription and leave you with a less accurate result.

You can try Microsoft Azure Speech Services for free before committing to the app. (Image credit: Microsoft)

We found that the two apps are also very comparable when it comes to recognizing multiple speakers. This feature isn’t always perfectly accurate if you have two people with a similar tone and a less than crisp recording. But most of the time, both Speech Service and Speech-to-Text were each able to differentiate speakers on a conference call within the transcribed text.

Support

Google Cloud Speech-to-Text doesn’t come with much support by default. You’ll find some basic troubleshooting tips online, but otherwise Google directs you to ask the community for help on Stack Overflow or Slack. You can purchase a support plan from Google if you need to talk to a tech. Options start at $100 per user per month.

Microsoft offers more online documentation for its Speech Service software, including how-to videos and example code for the platform API. But, you’ll also need to pay extra if you want support from Microsoft techs. Email-only support plans start at $29 per user per month, while phone support plans start at $100 per user per month.

Support plans for Microsoft Azure Speech Service. (Image credit: Microsoft)

Pricing and plans

On its face, Microsoft Azure Speech Service is significantly cheaper than Google Cloud Speech-to-Text. Microsoft offers five hours of free transcription per month and then charges $1 per hour of audio after that. Google provides just one hour of free transcription, after which the service costs $1.44 per hour of audio.

Pricing for Google Cloud Speech-to-Text. (Image credit: Google)

That said, pricing with either of these services can be complex. Google offers a 30% discount if you allow the company to log your audio data on its servers. In that case, Speech-to-Text is slightly cheaper than Microsoft’s Speech Service. At the same time, Google charges $2.16 per hour if you want to use the ‘Enhanced’ speech model. Microsoft raises its price to $1.40 per hour of audio if you supply custom speech or acoustics models.

Verdict

For most cases in which you need to transcribe speech-to-text, we recommend Microsoft Azure Speech Service. It’s significantly cheaper than Google Cloud Speech-to-Text if you have many hours of audio. We also found that it can be much more accurate if you take the time to supply custom speech and acoustics models with your recordings.

That said, Microsoft’s language support is very limited compared to Google’s. So, if you want one app that can handle recordings in nearly any language, Google Cloud Speech-to-Text may be the better option.

Date

22 Mar 2021

Sources


Share


Other Blog

  • Splunk wants to help your business unlock new opportunities in the cloud

    As businesses of all sizes look to recover and thrive following the pandemic, getting the most out of their data will be critical. But in an enterprise world rapidly moving towards hybrid working and cloud-first approaches, finding the most effective way to do that can be a significant challenge.

    Splunk has long positioned itself as a key helper for businesses looking to streamline or upgrade their data capabilities, and revealed a number of new releases and updates at its recent .conf 21 event designed to do just that.

    But the company wants to do a whole lot more for its customers, so TechRadar Pro went to find out more.

    Cloud-first

    "The way we see it playing out is when our engineers build, everything we build, we're building in the cloud first,” Shawn Bice, President of Products and Technology at Splunk, told TechRadar Pro.

    "Our mental model is that you build for the cloud first, and then you're testing these bits out all the time, and then every six months or so you turn the crank, and you package up a corresponding set of bits that you deliver on-prem. That allows us to be constantly building in the cloud, whether it be a Google environment, an Amazon environment, or an Azure environment, because those environments are different. They're not all universally the same.”

    "Multi-cloud and hybrid is a real thing - I have customers that when they spend money, they're spending money across Google, Amazon and Azure, but they might not spend the equal amount of money.”

    “But they're leveraging big cloud providers in a way where they pick the best tool or they get the best pricing - and there's hybrid or on premise in the mix, too. So you know, so when you meet somebody, they'll tell you, 'Hey, I got data across multiple clouds, it's on prem, I'm now doing the edge' - and they want Splunk to be there with them, wherever their data is."

    Splunk Platform overview

    (Image credit: Splunk)

    One way Splunk is looking to not just keep current customers, but also lure in new ones, is through the introduction of workload pricing. This will mean customers can buy based on the exact infrastructure used to deliver services, meaning no one should pay over the limit.

    Bice notes that such an approach has been around for some time, but few companies are offering it.

    "When people learn about workload-based pricing, they love it,” he notes, “it's about storage, and compute, just like anything else that you would buy up in the cloud. So from that regard, I think, pretty much I would expect everybody to switch over to this because the economics are just better."

    “The more customers that are able to take advantage of that, they're just going to be saving, and that's a good thing to do for people. So I always like to remind people that it's not necessarily a new shiny feature, but it's going to be better economics for you and you're going to be able to get more out of the system, spend less and that's just a good thing for any business.”

    Working for the long term

    Splunk has launched some intriguing new features and additions to its portfolio, including new Ingest Actions that allow customers to action data in motion, and a boosted Federated Search feature that provides a unified search experience across all deployment types in a single search bar, allowing users to track down the exact data they want.

    Post-pandemic, Bice notes that, remote working boils down to one issue for most customers: “they just want everything to work”.

    "So if you have to build technology in a slightly different way, realizing that the way people are interacting with it is slightly different,” he adds. “You know, those are things that you just do. I don't think anybody here believes that the pandemic is a common goal kind of thing. I personally think its maybe changed the workforce for the long term."

    But Splunk is enthusiastic about what its new offerings can provide for customers, with Bice, who only joined the company recently after spells at AWS and Microsoft, noting the opportunities the new tools provide.

    "I would never assume just because you feel something, people will come,” he says, “I've made that mistake once or twice in my career...I definitely recognize a lot of these customers have their own businesses to run, they're not waking up each and every day wondering what we did. So I think part of our job is to make sure we understand... what customers are trying to do."

    Looking to up your data game? Here's our pick of the best Business Intelligence (BI) tools around

    Read More
  • Cisco vulnerability could cause your firewalls to fail

    A security researcher has discovered a vulnerability in Cisco's firewall products that could be exploited to achieve denial of service (DoS).

    The vulnerability, tracked as CVE-2021-34704 with a CVSSv3.0 score of 8.6, was found by Positive Technologies researcher Nikita Abramov in the networking giant's Cisco Adaptive Security Appliance (ASA) and Cisco Firepower Threat Defense (FTD) firewalls.

    According to Abramov, elevated privileges or special access are not required by an attacker to exploit the vulnerability. Instead, they can form a simple request in which one of the parts will be different in size than expected by the device to exploit it. From here, further parsing of the request will cause a buffer overflow and the system will be abruptly shut down and then restarted.

    Abramov provided further details on how business operations can be disrupted if this vulnerability is exploited in a press release, saying:

    "If hackers disrupt the operation of Cisco ASA and Cisco FTD, a company will be left without a firewall and remote access (VPN). If the attack is successful, remote employees or partners will not be able to access the internal network of the organization, and access from the outside will be restricted. At the same time, firewall failure will reduce the protection of the company. All this can negatively impact company processes, disrupt interactions between departments, and make the company vulnerable to targeted attacks." 

    Denial of service

    In a new security advisory, Cisco warns that multiple vulnerabilities were discovered in the web services interface of Cisco ASA and Cisco FTD that could allow an unauthenticated, remote attacker to trigger a denial of service (DoS) condition. 

    The company also explained that these vulnerabilities are “due to improper input validation when parsing HTTPS requests” and that an attacker could exploit them by sending a malicious HTTPS request to an affected device.

    Thankfully though, Cisco has released software updates that address these vulnerabilities but it's worth noting that there are no workarounds which means you will have to install the software update if you own a device running Cisco ASA or Cisco FTD with a vulnerable AnyConnect or WebVPN configuration.

    To determine if Cisco ASA or Cisco FTD has a vulnerable feature configured, users will need to use the show-running-config CLI and check to see if “crypto ikev2 enable client-service port” or “webvn enable” are configured. If this is the case, Cisco recommends that you install its updates as soon as possible to prevent falling victim to any potential attacks exploiting these vulnerabilities.

    We've also highlighted the best firewall and best cloud firewall

    Read More
  • Serious WordPress plugin security flaw puts thousands of sites at risk of attack

    Patches have been issued to fix three vulnerabilities in the Responsive Menu plugin but many site owners have yet to install them.

    Read More
  • AMD RX 6700 XT could arrive in March to challenge Nvidia’s mid-range GPUs

    Rumored launch date is close, but whether you’ll actually be able to buy one – that’s another question.

    Read More

Find Out More About Us

Want to hire best people for your project? Look no further you came to the right place!

Contact Us