TranscribeMe

Research Team Review of TranscribeMe

The world is full of data, but that data isn’t always available in text form. While video and audio are great ways to create engaging content, they’re not accessible to anyone who can’t hear them. They’re also hard to extract data from and take up lots of space on a computer, making them poor formats for textual analysis and archiving. TranscribeMe is a transcription product that helps solve these problems by extracting text data from audio and video files.

In addition to transcription, TranscribeMe offers language translation services. They also create and enrich specialized datasets for training AI and machine learning models.

TranscribeMe contracts with freelance transcribers and translators for many of their services. The company bills itself as an affordable, easy solution for any use case, from a single user with one video to a multinational enterprise that needs hours of video transcribed, translated, and annotated on a recurring basis.

How Does TranscribeMe Work?

Customers can use TranscribeMe in two ways: as a self-service web application, and as a contracted service provider. Individual users or businesses with one-off needs can use the web application to buy transcription services with minimal hassle. Businesses with recurring or complex needs can contact TranscribeMe for assistance and a customized price quote.

TranscribeMe Self-Service Transcription

TranscribeMe’s self-service portal is a straightforward web application that allows users to access their transcription services. To access their other services, such as translation, you’ll need to reach out to the vendor for a quote, as they cannot be accessed through the self-service application.

You can sign up for an account quickly and easily with just an email address. Once you’re logged in, you can start uploading audio or video files immediately.

Once you’ve uploaded your files, you can add notes for the transcribers and choose a quality level. TranscribeMe offers four levels of transcription quality:

• Machine Express is the cheapest, lowest-quality type of transcription offered by TranscribeMe. With this option, your files will immediately be analyzed by transcription software. TranscribeMe estimates that each minute of audio will take three minutes to transcribe at this level. You simply enter your payment information, and get a text file back a few moments later.

  Since the results are not reviewed by a human, this method works best with high-quality audio sources where the participants all speak clearly. Interference from crosstalk, background noise, or other auditory contaminants is likely to confuse the software and result in a poor transcription. However, for clean, high-quality audio, this can be a great cost-saving transcription solution.

  
  

• First Draft is a step up from machine transcription. At this tier, your audio will be given to human transcribers. If you have a lot of files or each file is long, the audio might be split into shorter pieces and assigned to many transcribers. These transcriptionists will perform a single pass on your audio, and you’ll get the results back without further human review.

  

  This tier only uses a single round of transcription. If a transcriptionist makes a mistake, there’s no review process to catch it. However, TranscribeMe is confident in the skills of the freelancers they use, and claims an average of 98% transcription accuracy for this tier.

  
  

• Standard is the second-highest quality tier. Similar to the First Draft tier, your audio will be given to human transcriptionists. However, TranscribeMe will send the first pass transcripts to a second round of transcribers, who will review the results and correct any mistakes. TranscribeMe guarantees 99% or higher accuracy for Standard transcriptions and offers to correct transcriptions that don’t meet this threshold.
  
  

• Verbatim is the highest-quality tier. At this level, your audio is given to human transcribers and passed through two rounds of review. Additionally, the transcribers will record every filler word and vocalization from the audio. If exact recreation of audio is what you need, this is the tier to go for.

For all of the human-transcribed tiers, you can specify additional factors that affect the transcription process. This includes industry-specific content (legal, healthcare, etc.) and heavily accented speech. You can also choose a turnaround window, which varies from next-day service to 8+ business days depending on the transcription tier, volume of content, and audio quality.

TranscribeMe offers a mobile app for their transcription services as well. This standalone application supports in-app recording, and is available for both iOS and Android devices. Don’t try to use TranscribeMe’s self-service website on a mobile device, though: the page is poorly optimized for small screens. Even at mid-sized window dimensions, the layout hides important UI elements, preventing you from using them properly. Use the mobile app instead!

TranscribeMe Custom Services

On top of transcription, TranscribeMe offers translation, custom datasets for training AI and machine learning models, and data annotation for specific needs.

Translation is exactly what it sounds like. TranscribeMe uses a network of freelance translators to take audio, video, or text in one language and translate it into text in another language.

TranscribeMe says that it translates “most major language pairs.” This implies that you can only choose a source language and a destination language, rather than translating one language into multiple others. It also implies that the product only supports specific language pairings.

TranscribeMe doesn’t publish any additional details about language pairings. However, they do offer an overall list of the following languages:

Machine learning (ML) models commonly use human-speech input as part of their training. TranscribeMe’s custom AI dataset service is designed to deliver text that is ready to feed into a ML model. These models require very specific structures for data input, and this format can vary between models. TranscribeMe works with clients to understand exactly what their data needs to look like, then gives the audio to freelancers trained specifically to transcribe for AI datasets. These datasets may then be used to train models that recreate text, analyze conversations for topic, or estimate customer sentiment.

Data annotation similarly focuses on AI and machine learning datasets, but from a different angle. Rather than simply providing transcribed audio in a raw dataset, TranscribeMe’s data annotation service adds context and detail to transcribed data. For instance, they can mark statements with their emotion or sentiment; they can identify grammatical parts of speech like nouns, verbs, and adjectives; or they can add tags to text that tells the computer to link that text to another piece of data, such as a name or an address.

TranscribeMe offers many kinds of data annotation with a variety of potential uses. One business might train their support chatbot to detect frustration and direct customers appropriately. Another might train software that makes sure their sales or support teams follow internal best practices. A third business might ask for visual annotation to help their software recognize specific objects.

All three specialized services require direct outreach to TranscribeMe. Similar to many enterprise software vendors, you’ll contact a TranscribeMe representative to explain the needs of your project, determine a turnaround time, and figure out a customized quote.

TranscribeMe Features

TranscribeMe offers several services and features:

• Audio and video transcription, by both humans and software

• Audio, video, and text translation

• Customized datasets for AI training

• Data annotation for advanced AI understanding

• Secure data encryption and transfer

• Data compliance standards, including HIPAA and GDPR

• A wide network of freelancers, supporting fast turnaround times and large-scale work

Audio and Video Transcription

As mentioned above, TranscribeMe’s core feature is transcription of audio and video content into text. You can choose software-based transcription or choose to have your files sent to TranscribeMe’s network of freelancers. The service supports any scale of use case, from single users with a short audio file to large enterprises with hours of content.

Audio, Video, and Text Translation

As mentioned before, you can work with TranscribeMe to translate between multiple languages. Their teams will translate from audio/video to text, or from text to text. You’ll have to contact TranscribeMe and work with them to arrange delivery of the files to be translated.

Customized AI Training Datasets

TranscribeMe works with clients to transcribe audio and video into specific formats to use in AI and ML model training. The requirements for these formats can be very strict and vary between ML models, so they work closely with customers to make sure the final deliverable meets the requirements.

Data Annotation

TranscribeMe’s data annotation service enriches conversational transcriptions with contextual information. This extra detail helps AI systems better understand and respond to specific situations. For example, data annotation could add emotional sentiment (angry, happy, sad) to a specific sentence.

This process is highly customized to your use case. You might want your data labeled on a broader scale, or you might want individual words tagged with grammatical data. TranscribeMe works with each customer to create the specifications for each project.

Data Security and Encryption

TranscribeMe is designed to be suitable for many levels of data sensitivity. Since your data will be seen by a network of freelancers, it’s understandable to be worried about sensitive data being used improperly. However, TranscribeMe guarantees that their internal distribution platform maintains information security, encrypts your data, and maintains it in secure formats.

Data Compliance

If your data is subject to legal compliance requirements, TranscribeMe is prepared to help. The company has options for HIPAA- and GDPR-compliant workflows. They can also customize their services to more specific needs. For instance, TranscribeMe offers workforce geofencing to make sure that only freelancers within certain geographic areas are able to work on your files.

Freelancer Network

Instead of a traditional workforce, TranscribeMe takes advantage of a large network of freelance translators, transcriptionists, and other roles worldwide. Using their distributed labor, TranscribeMe offers fast turnaround times for many project sizes. Their task distribution system also allows them to take on large-scale projects by disseminating bite-size tasks to many people.

Customer Support

TranscribeMe offers standard customer support for users of its web app and mobile app. You can contact their support team through a chat window and wait for a response via email. They also have a list of commonly asked questions to answer simple queries about turnaround times, refunds, and other simple situations.

For customers of their other services, TranscribeMe isn’t clear on what support they offer. However, you will already be in contact with a dedicated TranscribeMe representative to transfer files, discuss your needs, and provide a quote. It’s likely that your point of contact will also be able to answer any support issues you have or escalate them to the appropriate team.

Pricing

TranscribeMe uses unit-based pricing. They list the following starting prices:

• Transcription (Human): $0.79 per minute of audio/video
• Transcription (Machine): $0.07 per minute of audio/video
• Translation: $0.11 per word
• AI Datasets: $2.00 per minute of audio/video
• Data Annotation: $0.10 per task

Translation, AI dataset, and data annotation services all require a quote, so the rates listed above won’t necessarily be final. However, they’re a good way to estimate your costs.

Transcription Pricing

Since TranscribeMe’s web app serves as an online store for their transcription service, you can get a much better breakdown of their pricing through trial and error. We used a short sample file to find the lowest rates for all transcription tiers. Our tests showed that the starting rate ranges from $0.07 per minute at the Machine tier to $1.75 per minute at the Verbatim tier.

However, additional transcription options can greatly affect the rate. For example, a Standard-tier transcript of legal-industry Chinese-language audio with timestamps, IDs for 8+ speakers, and a one-day turnaround time would costs $5.72 per minute. That's over four times as much as the lowest rate.

Competitor Comparisons

TranscribeMe’s pricing is similar to competitors with a similar structure, like Scribie and WayWithWords. These services all portion out work to transcribers, so they can handle large projects with short turnaround times. To compare between these vendors, you’ll probably want to focus on the end price rate for your situation. The end product you receive will likely be similar for all three.

Some enterprise transcription products are fundamentally different in implementation. Amazon Transcribe and Azure Cognitive Speech Services are both cloud-based speech-to-text software. They use machine learning algorithms to transform speech to text either live or from a recording. If audio transcription is a core part of your business model, then dedicated software like this might be a more suitable option. These products are best used by large enterprises, especially ones who are already part of the AWS or Azure ecosystem.