In-house text-to-speech or managed service: the real cost
Wiring up a text-to-speech API looks cheaper than paying for a service. Once you count development, maintenance and text extraction, the maths often changes.
When a technical team discovers the per-character price of the big text-to-speech APIs, the conclusion seems obvious: just wire the API directly and avoid paying for a service on top. The reasoning is fair, and for some projects it is even correct. But it usually rests on a single line of the quote, the cost of synthesis, while ignoring everything around it. Here is an honest comparison between building text-to-speech in-house and going through a managed service, so the choice is made on the real cost rather than the illusion of a per-character rate.
What the per-character price leaves out
A text-to-speech API turns text into audio. That is all. Yet between a published article and a listenable audio file online, there is a whole chain the API does not handle.
First you have to extract the clean text from the page: strip the menus, the sidebars, the captions, the ad blocks, and keep only the body of the article. This step feels trivial until the day a template change breaks the extraction and every one of your audio files starts reading "Share on Facebook" in the middle of the text.
Then you have to host and serve the files, run a player on the site, track which articles have already been processed so you do not regenerate them, handle article updates, and store everything. None of these bricks come from the synthesis API. They are the parts that cost money, in development first and maintenance after.
The hidden cost of initial development
Building in-house means building everything that is missing. In practice, that is several weeks of a developer's time: text extraction, a processing queue, storage, a front-end player, re-run and correction handling, a tracking dashboard. At a realistic loaded cost, this initial project easily runs into thousands before the first article goes live.
This is not an argument against building in-house as such. If your volume is enormous and your team already staffed, that cost amortises. But it has to appear in the comparison, just like the per-character price. Comparing "the API at X per million characters" with "the service at Y per month" without counting development is comparing a spare part with a car.
Maintenance, the line everyone forgets
The real cost of an in-house system is not building it, it is keeping it alive. Voices evolve, APIs change their formats, your CMS updates, a strange pronunciation case comes back from support, a very long article knocks over the processing queue. Each of these incidents pulls in the developer who built the system, and often only them. This recurring cost is invisible on a quote but very real in an annual budget.
A managed service pools that maintenance across all its customers. When a voice improves or a pronunciation case is fixed, everyone benefits without thinking about it. That is precisely what you are not paying for: the time you will not have to spend.
When building in-house is still the right call
To be fair, building in-house makes sense in some cases. If you produce tens of thousands of hours of audio, if you have confidentiality constraints that forbid any external call, or if audio is the core of your product rather than a feature, building internally is defensible. You gain full control and, at very high volume, a lower marginal cost.
For the vast majority of publishers, though, audio is a useful feature but not the business. In that case the maths tilts toward the service: no initial development, no maintenance, an integration via API or RSS feed that fits in a few lines, and a predictable price.
Comparing like with like
The right instinct is not to ask "how much does text-to-speech cost" but "how much does an audio article online cost, extraction, player and maintenance included". Once the calculation is framed that way, the gap between the two options narrows a great deal, and the managed service often wins on time saved alone. For detailed figures on the service side, see our article on the cost of an audio version and the pricing page by article volume. And to judge quality before any commitment, the simplest path is to listen to the result on one of your own articles.
Give your articles a voice with WeDispatch
This blog is itself voiced by WeDispatch. Curious how it sounds on your content?
Book a demo