How do I turn an audio recording into text?

Audio transcription is the process of converting a spoken audio recording into text. There are several reasons why someone may need to transcribe an audio file:

• To create written notes or a transcript from a meeting, lecture, interview, or other spoken event.
• To produce text captions or subtitles for a video or podcast that contains speech.
• To create a written document from dictated audio notes or voice memos.
• For market research companies to analyze focus groups or customer service calls.
• For scientists to analyze verbal responses in psychological or medical studies.
• For lawyers or court reporters to produce legal transcripts from depositions, trials, or hearings.
• For journalists to obtain quotes and spoken remarks from interviews or speeches.
• For government agencies to generate accessible transcripts for public records or broadcasts.
• For law enforcement and security purposes such as transcribing wiretaps or interrogation recordings.

In summary, transcription allows spoken audio content to be efficiently searched, edited, shared or analyzed in written form.

Recording the Audio

Recording high-quality audio is crucial for getting an accurate transcription. Here are some tips for capturing clean audio that’s easy to transcribe:

Choose a quiet environment without background noise. Turn off any music, fans, or AC units. Close windows and doors to minimize external sounds. The clearer the speech, the better the transcription (Source).

Position microphones close to speakers and away from distracting sounds. Lavalier microphones clipped to clothing work well, as do external mics positioned 2-3 feet from the speaker’s mouth. Try different mic placements to find the optimal setup (Source).

Adjust recording levels to avoid peaking or distortion. Do a test recording and check levels in an audio editor. Make sure volume peaks don’t exceed -3 dB. Boost quiet voices while normalizing louder ones (Source).

Record in a high-quality format like WAV or AIFF at a 16-bit or 24-bit depth. The higher the audio resolution, the better. Sample rates of 44.1 kHz or 48 kHz are recommended for voice audio (Source).

Ask speakers to talk clearly at a steady pace and volume. Transcribing natural speech patterns with “umms” and “ahhs” is challenging. Pausing between thoughts helps. Provide water to avoid dry mouths (Source).

Transcription Options

You have two main options for transcribing an audio recording into text: doing it yourself or using a transcription service. Here’s an overview of the pros and cons of each approach:

Doing it yourself allows you to transcribe the audio manually and have full control over the process. However, transcribing audio is very time consuming. It takes about 4-6 hours to transcribe one hour of audio depending on your typing speed and accuracy. So for long recordings, it can become tedious and impractical to transcribe yourself.

Using a transcription service hands off the work to professionals and can save you significant time. Most services charge either an hourly rate or a per-minute rate. Rates vary but can be around $1-$3 per audio minute depending on the service, turnaround time, and formatting. While the cost adds up for long recordings, it’s often more affordable than investing hours of your own time (see Pros & Cons of Automated Transcription Service).

Transcription services offer two main options: automated transcription using AI/machine learning or human transcription using professional transcribers. Let’s look at both in more detail:

Transcribing Yourself

If you want full control over the transcription process or need to save money, you can transcribe the audio recording yourself. Here are the steps for manually transcribing an audio file:

1. Get the right equipment. You’ll need a computer, a good pair of headphones, and software that can play the audio and allow you to type. Transcription software like Descript, Microsoft Word, or Google Docs works well.

2. Import the audio file into your transcription software. Most programs allow you to upload an audio file directly. Make sure it’s a common format like WAV or MP3.

3. Start the audio and type out everything said verbatim. It’s tedious work, so you’ll need patience. Use a foot pedal if you have one to pause, rewind, and fast forward hands-free.

4. Add punctuation and paragraph breaks to make the transcript readable. Listen closely to the speaker’s tone, inflection, and pacing.

5. Double check the transcript while listening to spot any missed words. Fix spelling errors or typos.

6. Export the finished transcript as a text document. From there you can format it as needed.

Choosing a Transcription Service

When selecting a professional transcription service, there are several key factors to consider:

Accuracy: According to Six Factors To Consider When Choosing a Transcription Service Provider, accuracy is the most important factor. Look for a service that guarantees at least 99% accuracy.

Turnaround Time: Choose a service that offers a reasonable turnaround time for your needs. Turnaround can range from a few hours to multiple days depending on the audio length and service selected, as noted in Things to Consider While Choosing a Transcription Service.

Pricing: Transcription pricing may vary based on factors like audio quality, length, and turnaround time, according to How to choose a transcription service?. Compare prices between services for your specific project.

Security: Make sure any service you choose offers adequate security and confidentiality measures, as advised here.

Customer Support: Choose a service that provides reliable customer support in case any issues arise with your order.

Automated Transcription

Automated transcription services use speech recognition technology and artificial intelligence (AI) to automatically transcribe audio into text. Popular automated services include AssemblyAI, Otter.ai, Google Cloud Speech-to-Text, and Amazon Transcribe.

These services work by feeding the audio file into complex neural network models that have been trained on massive amounts of data to recognize speech. The AI listens to the audio, identifies the words and phrases, and types them out automatically without any human intervention.

Automated services are convenient because they allow you to upload an audio file and receive a transcription in minutes. However, they may not always be completely accurate. According to benchmark tests by AssemblyAI in 2022, the top automated services had word error rates between 5-10%. This means 5-10% of words were either incorrectly transcribed or missing from the transcript.

The accuracy of automated services depends on several factors like audio quality, speaker accents, background noise, and vocabulary. Transcribing conversational audio tends to be more difficult compared to a clear professional recording. Reviewing and editing the automated transcripts is advised to fix any errors.

For many basic transcription needs, automated services provide a good balance of convenience, speed, and accuracy. But for recordings requiring very high accuracy, human transcriptionists are still the gold standard.

Human Transcription

Human transcription services provide a number of benefits over automated services. Human transcribers can understand nuanced speech much better and will result in significantly higher accuracy, capturing every word with precision. Some top human transcription services include:

• Rev – Offers fast turnaround starting at 12¢/minute with 99% accuracy guaranteed. Great for interviews, meetings, and lectures. https://www.rev.com/
• GoTranscript – Affordable transcription starting at $0.90/minute with 24 hour turnaround. Specializes in legal, academic, and medical transcription. https://gotranscript.com/
• TranscribeMe – Offers general transcription starting at $0.79/minute with a 12 hour turnaround time. https://www.transcribeme.com/

The human touch allows for the highest accuracy possible when transcribing audio to text. Services that utilize trained professionals will result in clean verbatim transcripts without sacrificing quality.

Turnaround Time

The turnaround time for getting an audio recording transcribed can vary significantly depending on whether you choose automated or human transcription. Automated services using speech recognition technology can return transcripts within minutes or hours. However, accuracy rates tend to be lower with around 70-80% accuracy for automated services.

Human transcription turnaround time is longer but produces much higher accuracy, often 95% or greater. For a professional human transcriptionist, the typical turnaround time is about 4 hours of work for every 1 hour of clear audio. So an hour long recording would take around 4 hours to be transcribed. Turnaround time also depends on the number of speakers, audio quality, and subject matter complexity. For very fast turnaround, some services offer 12 hour, 24 hour or 48 hour delivery if you pay a premium rate.

According to sources like Gotranscript and Verbit Ai, turnaround time for human transcription of a 1 hour long clear audio file is typically 24-48 hours. For very fast turnaround, extra charges apply and prices may be double the base rate. Overall, plan for human transcription to take 4-5x the length of the audio recording to be completed.

Pricing

The cost of transcription services can vary significantly depending on the provider, turnaround time, and accuracy level. Here’s a breakdown of pricing for the main transcription options:

Automated transcription through services like Otter.ai offers plans starting at free for 60 minutes per month up to $20 per month for unlimited minutes. Accuracy levels tend to be around 80-90%.

Basic human transcription services charge in the range of $1 to $1.50 per audio minute, with a minimum fee of around $25-40 per file (source). Turnaround times are typically 1-2 days.

Human verbatim transcription with timestamps, speaker labeling, and 99% accuracy ranges from $3 to $5 per audio minute (source). Turnaround can be as fast as same day.

The higher the accuracy rate, faster turnaround time, and more formatting requested, the more transcription services tend to charge per audio minute. Factoring in large volumes or difficult audio can also increase pricing. Overall, automated services offer the lowest rates but have accuracy tradeoffs, while human verbatim transcription is the most expensive but highest quality option.

Improving Accuracy

There are several tips you can follow to get the most accurate transcript possible from your audio recording. First, make sure you are recording in a quiet environment without background noise, crosstalk or echo. As according to symbl.ai, clean audio improves accuracy.

You can also feed the transcription service custom vocabularies and word lists specific to your content, like product names or industry jargon. This helps the speech recognition engine better interpret your audio. As Amazon Transcribe explains, custom vocabularies improve accuracy.

Finally, review the transcript and correct any errors. Human-edited transcripts can be used to further train the service’s algorithms over time. As noted by Genesys, manual transcripts improve overall accuracy.