I have been reading up on a variety of topics lately one of which was organizations making the segway
from actual voice-over actors to the use of artificial intelligence (AI). Before I get too deep into this
discussion let me designate what the actual voice-over actor costs on an annual basis according to a
According to ZipRecruiter, industry averages include:
Entry-level voice actors: $13,500–$31,999 per year.
Mid-level voice actors: $69,000–$87,499 per year.
Experienced voice actors: $111,500–$199,000 per year.
The article I am referencing here is based on the use of AI for an organization called Snowflake which
was founded in 2012. Snowflake is a publicly traded (since 2020) company on the NYSE (New York Stock
Exchange). They are a data cloud company.
Snowflake was the first cloud platform in the world to house and unite siloed data that would be easily
discovered and shared across multiple clouds. Today they boast over 6,000 customers and have
dramatically changed the way their clients use data that helps them to make better business decisions.
Snowflake offers a variety of products and services. A lot of what they offer is customer support courses
that last between twenty and thirty minutes in duration. All are filled with videos and voice-overs to
ease the learning experience for the consumer. Naturally, these are updated on a very regular basis to
keep pace with the marketplace.
Nick Goodman is their (Snowflake’s) Educational Program Director. Goodman said that they create about ten videos per quarter. All of which require regular updates and/or addition of new materials. He said that he realized that in the production of his organization’s vast library of educational materials, they needed to find a better avenue for providing high-quality, consistent, and easily produced video voice-overs.
Snowflake was using people from a variety of locales to create their voice-overs. Most worked from their home studios. This voice-over process was oftentimes hit-and-miss and required several takes to create a seamless end product. Splicing into existing voices to add updates was a mess, time-consuming, and quite inconsistent. What could they do?
Goodman said the revision time was lengthy. It began with the instructional designer creating the recorded module along with their subject matter experts (SMEs) input. Depending on the feedback they (the instructional designer) would go back to rerecord the entire session and this was often done in multiple takes. Each revision opened them up to more inefficiency and inconsistency.
Goodman had an idea to try artificial intelligence (AI) instead. He devised an experiment whereby her created a single lesson with three voices. One voice was an actual voice-over person. Two were created via AI. He sent these recordings off to his team for feedback telling them it was a simple evaluation of a new voice-over person. Nobody could tell that the AI was not an actual person. Goodman’s test was a huge success, but was it going to be a better return on their investment than how they were doing things? Goodman realized that updates were made for each module of learning at least annually. However, product updates sometimes came as often as weekly. This presented an issue because as the instructional designers and SMEs were in the process of creating modules updates needed to be incorporated. Using regular voice-over people created more time and less flexibility than the AI voices. Some of the voice-over actors were not always available on short notice and if they wanted seamless modules Snowflake had to wait for that person to free up their schedules for revisions. This was not always practical.
His selection for this new venture was a product called WellSaid. WellSaid offers several avatars for voice-over and they are always available. Avatars never get sick or have conflicting schedules. They offer consistent quality and availability. There are many other organizations that offer similar products to WellSaid. Goodman was able to shorten the timelines for completion of these projects significantly by using the WellSaid software for voice-overs. Instead of days to redo the voice-over material it now can be done in
a matter of minutes. He now says that the visual elements of these lessons are the most time-consuming part of production.
Goodman said “We used to have to produce the visuals, then record the audio to sync to it. Now, one of those (the voiceover) is just a quick thing that we do, and the other one (the visual) takes up most of the production.”
Requiring less focus on the voice-over portion for lesson production has freed up the instructional design team to concentrate more on creating the right material for their clients. Editing the voice-over is so much easier and faster. The instructional designers are working more efficiently than ever before. Thus, saving Snowflake time and money.
I offer a personal example below:
I worked for Quadmark out of Chandler Arizona as an Instructional Designer for around seven months back in 2021 and they used actual voice-over actors. Several were in their stable of content developers for their instructional materials. I am entirely unsure how they were compensated but know that they were compensated beyond their content development work for these sessions. They had an in-house green room and state-of-the-art equipment to assist in these sessions. Their video and audio staff were
very very good too. Naturally, these costs, whatever they were, were passed along to their clients. I can recall some sessions being revised and others not specifically revised, but requiring multiple takes to ensure a perfect product was presented to their clients. This takes time and money to produce. Could they possibly reap more financial rewards from something like what Snowflake is doing with the use of AI for their instructional products?
I would say probably. However, it also depends on cost. Snowflake is a large publicly traded company and Quadmark was a relatively small company with assets around the globe.