Missed me? I bet you did.
I missed you too and I can’t even hear you, so hey, I’ll conclude you missed me a TON!
It’s only been 11 days since my dramatic departure, but I thought I’d share something I did at work looking at synthetic voice services and their “impossible to compare”-pricing. They all price pr. minute, character, token, hour or words. COME ON.
Most of the services offer discounts on yearly plans, most of them can give better prices when you have heavier usage, so this is just indicative.
Take a look.
5M characters? WTF.
The reason why I calculate with 5M characters pr month is that it connects with AWDIOs business. It’s not how much we actually use (we use more), but to compare prices.
A way to understand characters is to see it in a simple equation. Average words are about 6 characters long. 1 minute of narrated text is about 150 words (normal reading tempo).
So with 1 hr of narration, you end up with:
150 words/minute x 60 minutes = 9000 words / hour
9000 words x 6 characters/word = 54000 characters / hour
5M characters = 92,5 hours of narrated contentConclusion
Average price is $743,25 and that is including cheap Amazon ($80) and expensive Resemble ($2000).
So why is Amazon so cheap? They don’t care about making money “here”. With their 1000s of services in AWS, it is all part of the bigger play. Their quality is currently below the competition, but I’m certain they’ll get there (they have the $ to build or buy)
Resemble? It’s their a la carte pricing. If you talk to them it will get a lot cheaper (I suspect) and their quality is really good.
Quality is pretty comparable right now, except for Amazon, but prices vary a lot, as you can see.
My favourite? Jury is still out, but I love the quality of ElevenLabs and Play.ht
What lies ahead for the synthetic voice market?
I suspect this market will consolidate soon. These small and curious players will get acquired by the bigger ones as the prices go into a race for the bottom. Non of the startups have enough “moat” or cash to fight with Amazon, Google or Microsoft.
Oh, and Stability AI, Hugging Face and OpenAI probably will take a stab at synthetic voices too.
Eat or be eaten. So much drama coming up. Love it!
I’m off again. Have a great summer! I’ll be back in [I have no idea]!