I mentioned above that you can charge real money for raw MT.
That’s because you can make your raw MT better than anybody else’s.
How? The same way you can deliver better work with the GAIT workflow on full human translation, by:
1. getting out in front of the AI with a robust anchor prompt and
2. taking command with a little opportunistic iteration.
Here’s what to do:
First, go ahead and human-translate that first “gold standard” content from the project, say, 20-30 segments or so, along with a short glossary of key terms (or just pull it from the TM if it’s good and relevant).
Next, put those into your anchor prompt, along with some stylistic instructions.
Then, start the process of sending content through the AI using this anchor prompt.
Don’t check the translation word for word. But keep your eye on a cycle or two of output. If any odd target-language terminology choices or stylistic issues pop out at you, fix them and update the anchor prompt.
Once you’ve gotten the low-hanging fruit fixed, turn up the gas… Crank the translation out with the anchor prompt at full throttle…. 20,000 words an hour easy… Maybe 30,000…. Watch that your CPU doesn’t overheat.
What does the client get for this?
A good (not perfect!) translation in context, following style and terminology guidelines, generally consistent from beginning to end… Formatted exactly like the source…. with data confidentiality.
Even if your client can do it “free” in ChatGPT, they’ll take four times as long and do much worse work. DeepL can’t touch this either.
Charge the client a $25-50 workflow setup fee (call it something really fancy!) and then a quarter to a half cent per word thereafter.
If you can’t make $60-100/hour at this, then you’re doing something wrong.
This is why upskilling on AI as a translator is critical for your career. Do this first, and you’ll superpower all of your follow-up efforts, including your marketing. Maybe you won’t even have to do marketing.