Remote transcription in Australia is not what it was ten years ago. Most of the volume now sits with AI training work rather than legal or medical typing, which changes how it is priced, how it is reviewed, and what makes one applicant worth more than another.
I run the contributor side of a speech data company, so this is the work I commission. This page covers what the tasks actually are, what they pay, how the ATO treats the money, and why an Australian accent is an asset in this particular market rather than a neutral fact.
What remote transcription and AI audio work involves
The core task is still typing what you hear, but the surrounding work has grown. Projects ask for transcripts to a specific written convention, for checks of an existing transcript against the recording, for speaker turns to be marked in conversation, for tags on accent, emotion or background noise, and for prompts to be recorded in your own voice.
The conventions are the part people underestimate. Every project has house rules about hesitations, overlapping speech, numbers and unclear audio, and getting those wrong is the usual reason a promising start turns into a rejected batch. Read the brief twice and check your first few files against the examples before doing fifty more.
What the work pays
Tasks are priced individually and shown before you accept. Across current projects that works out to roughly 12 to 50 USD per hour, depending on the task type and how scarce your language is. It is an observed effective rate on per-task work rather than a wage, and no platform can promise a given week, because work arrives when clients commission projects.
Transcription is usually priced per audio minute or per file rather than per item, which matters more than it sounds: clean single-speaker audio and noisy multi-speaker audio pay differently for good reason, and your real rate depends on audio quality as much as on typing speed. Payouts are commonly set in US dollars, so check the conversion and any transfer fee into an Australian account.
Australian accents and multilingual households
Existing speech training data is dominated by American and, to a lesser degree, British English. Australian English is comparatively thin on the ground, which is precisely why an Australian contributor has something to sell. Broad Australian accents, regional variation, and rural speech are harder for projects to source than the general Australian voice people assume is wanted.
The bigger opportunity is multilingual. Australian households speaking Mandarin, Arabic, Vietnamese, Cantonese, Punjabi, Greek, Italian or Tagalog alongside English are exactly what projects run short of, especially where someone code-switches naturally mid-sentence. List every language you have, at the level you actually have it, rather than only the ones you think a company would want.
How the ATO treats this income
This is income you declare yourself. Nothing is withheld at source, so keep records of what you were paid and when, and set money aside as you go rather than meeting the position at tax time.
Whether you need an ABN depends on whether you are carrying on a business rather than doing occasional work, and GST registration only becomes a question above the registration turnover threshold, which part-time contributors will not usually reach. Both of those tests have detail to them, so confirm your position with the ATO rather than with a platform. No data company, this one included, is a source of tax advice.
Voice recording and the adjacent work
If transcription queues are quiet, the neighbouring work is worth knowing. Voice recording tasks pay for reading prompts aloud in your own accent and often work fine on a recent phone. Transcript review pays for checking someone else's work against the audio, which suits people who are precise but not fast typists. Annotation tasks extend to marking speaker turns, tagging noise and emotion, and judging model output.
These use the same profile and the same account, so qualifying once tends to open more than one queue. That matters in a market this size, where any single project may not have Australian-specific volume open at a given moment.
Checking a platform before you give it your time
One test catches almost every scheme in this field: which way does the money move? A real platform pays you for approved work and never asks you for a registration fee, paid training, a starter kit, an equipment charge, or a deposit to unlock a withdrawal. If money is asked to flow from you to them, it is not work, however polished the site looks.
The pattern reaching Australian phones most often is unsolicited recruitment through WhatsApp or Telegram, promising a fixed daily income for trivial tasks and then requiring a top-up before a balance can be withdrawn. It is not a version of this work. Never hand a recruiter a banking password or a one-time code, and walk away from anything whose real earning mechanism is recruiting other people.
How to start from Australia
Build the profile before a project opens, because matching runs against details already on file. Name your accent and every language in the household, say what you can record with, and be honest about the room. A quiet room with soft furnishings and a recent phone beats a good microphone in a tiled kitchen. For transcription, a real keyboard and closed-back headphones make more difference to your approval rate than typing speed does.