I'm left armless for a month or more and want to try dictating my work rather than typing it with one hand.

Please share your experiences in this area, especially WRT headset quality.

What type computer are you using?

The Macs that we use have microphones on the bezel around the screen, that work fine for dictation. Might not work to well in a noisy environment.

Lenovo laptop with Win 10.

A co-worker recently did a fairly comprehensive survey on this subject to assist a staffer who had been in an auto accident and broke her collarbone, making it difficult for her to take notes. We were not looking at headsets per se, but more at the transcription software, and this was for a Mac so perhaps not particularly applicable.

We found that most microphones designed for this purpose provided acceptable sound quality, output level, and ambient noise rejection. Note that even "quiet" environments often have significant levels of environmental noise, from the computer itself if not an external source. USB microphones, as a group, provided somewhat better audio quality than those that plug into an audio port. Microphones integrated with a USB webcam were as good as stand-alone USB microphones, and price/performance seemed to be pretty closely correlated.

Transcription software OTOH is junk. Voice to text technology does not appear to have made many (if any) improvements since the last time I looked at this, over 20 years ago, when I headed a well-funded study for a Federal Government research agency. Single-speaker+controlled environment+"sheep-voice"+limited (non-technical) dictionary is barely adequate, with the best software package performing at no better than 40% accuracy. Complicate any of those factors, and the accuracy level goes way down. Nor does "training" significantly improve accuracy. Bottom line, if you are interested in recording audio, e.g. for manual transcription at a later time, there are plenty of decent options, and even the least costly should be adequate. If you want automated speech->text, giving up before you start will spare you a lot of time and frustration.

For those readers curious about "sheep voice," people are, more or less, "sheep" or "goats" in their speech characteristics. Voice transcription software (and voice recognition as well) performs much better for a sheep than a goat. It is mostly rate of change of pitch and volume that differentiates a sheep from a goat, more than consistency. Most goats don't know they are such, and people can't tell one from the other as well as software can.

This is the on I use with Dragon speech recognition software. I use it to prepare PI reports. The Dragon program is surprisingly accurate compared to what was available a few years ago. You have to go back and do a few edits but it "learns" and becomes more accurate over time. I found the headset on Amazon

Monaural Voice Recognition USB Headset with Noise Cancelling boom Microphone for Dragon

I use a transcription service for my work. Human beings are much better than software. Just dictate into the phone, no arms required.

I don't know what kind of work you do, so unsure if it would be workable for you.

I imagine Google's voice typing solution would be the most robust, since it is learning from millions of user interactions per day. It's built into all of the Google apps, like Google Docs.

Just open up a Google doc, select the voice typing option, and give the app permission to access your microphone. It's free.

In the days of old, stenographers would come into your office take shorthand and hand you the typed copy for corrections. Some things were better before technology.