✦ Task type · Creator work
Short voiceovers, and the rights question you cannot skip
The recording is the easy half. The half that has changed is what you are permitted to do with someone's voice afterwards, and briefs written three years ago no longer cover it.
Short voiceover is one of the few task types where the deliverable is genuinely simple and the contract around it is not.
What to put in the public brief
- Language, and the accent or market if it matters
- Approximate script length in seconds
- The audio floor: quiet room, no background music, phone is acceptable
- Where the audio will be used, in plain terms
- Explicitly whether synthetic voice training is or is not included
Keep the script private. It is usually the campaign.
What to put in the private instructions
Give the script and one line of context
Who is being spoken to, and what they should feel. Direction like "warm but not salesy" means little; "you are telling a friend something useful you just found out" is actionable.
State the audio conditions plainly
"A carpeted room with the window shut, phone about a hand's width away." Non-professionals do not know what a good room sounds like, and this is cheaper than rejecting takes.
Ask for two takes, not ten
Different pacing on the second. Ten takes produces fatigue, and the tenth is reliably worse than the second.
Repeat the rights terms where they will read them
Consent given in the brief is worth more than consent inferred from a payment.
Proof worth requiring
Ask for
The raw audio file, unprocessed, at the highest quality their phone records, plus both takes.
Do not ask for
Noise reduction or EQ applied by the worker. Amateur processing is destructive and irreversible; you can always clean a clean file, but you cannot recover an over-gated one.
When this task type is the wrong tool
If you need broadcast quality, a consistent voice across a long campaign, or an on-camera performance, this is not it — hire a professional with a treated room.
If the voice must match footage you already have, a UGC product demo recorded with sound is usually better than a voiceover laid over silent video, because the room tone will not match and audiences hear it without knowing why.
Common questions
What audio quality can I expect?
Phone-recorded, in a quiet room, with no processing. That is genuinely fine for social video and unusable for broadcast. If you need a treated room and a condenser microphone, say so in the brief and expect the price and the available pool to change accordingly.
Can I use the recording to train a voice model?
Not unless the brief says so and the worker agreed to it specifically. A general usage licence does not cover synthesising new speech in someone's voice, several jurisdictions now legislate on this directly, and burying it in broad wording is the version of this that ends in a claim. Ask separately, pay separately, or do not do it.
What reward is fair?
Ten to thirty dollars for a short read of up to a minute, with one revision. Raise it for a specific accent or a language with few registered speakers, because scarcity sets this price far more than duration does.
How do I avoid a robotic read?
Give context, not just the script. A line read cold sounds read; the same line delivered by someone who knows who is being spoken to and why sounds spoken. One sentence of context is worth more than any performance direction.
Thinking of taking this task?
The same job, from the other side. Rewards are set by the poster, so treat these as the range this task type usually sits in rather than a promise.
- Usually pays
- $10 to $30 for a short read
- You need
- A quiet room and a phone
- Proof expected
- The raw unprocessed audio, usually two takes
Check whether the brief includes voice-model training. That is a separate permission and should be paid separately; you can decline it and still take the task.
Create a worker accountBrief a voiceover
Set the reward, the languages or accents you need and the number of takes. Workers reserve a spot; your script stays private until they do.
Post this task