Working with a Voiceover Artist: Direction, Sessions, and Delivery

You booked a good voice. The demo was perfect, the rate was fair, the files came back clean. And the read is flat. It hits every word and lands none of them, because nobody in the room told the artist what the spot was actually for. That is not a talent problem. That is a direction problem, and it is the most common way a voiceover session goes sideways.

A voiceover artist is a performer, not a text-to-speech engine you pay more for. The words are yours. The performance is a collaboration, and the person who owns that collaboration is whoever is directing the session. Most of the time on a commercial job, that is you or your editor, sitting on a phone line or a Source-Connect session while someone in a booth two states away waits for a note. If you treat that seat as passive, you get a passive read. Here is how to run it like the creative role it is.

Write for the ear before anyone reads it

The script is the first place a VO job succeeds or fails, and it fails most often because it was written to be read on a page instead of heard out loud. Long clauses, stacked prepositions, and a product name buried in the middle of a sentence all look fine in a doc and fall apart in a booth. Read your own copy out loud before you send it. If you run out of breath, the artist will too.

The hard constraint is time. A commercial read has a runtime, and the copy either fits or it does not. Conversational narration lands around 150 words per minute. A measured, premium read runs closer to 140. A brisk retail read pushes 170 to 185 when someone is reading legal disclaimers against a clock. At 150 words per minute you get roughly two and a half words per second, so a 30-second spot holds about 75 words if it is wall to wall, and more like 60 once you leave room to breathe, to pause on the logo, and to let a music bed land. A 60-second spot is 120 to 150 words. A 15-second cutdown is 30 to 40.

The mistake is writing to the maximum. Copy that technically fits at 150 words per minute leaves the artist no room to perform, so you get a rushed, clipped read that sounds exactly like what it is: too many words in too little time. Budget the runtime the way you budget a shoot day. Leave headroom. If the copy is 80 words for a 30-second spot, cut it to 65 before the session, not during it, because cutting live burns session time and rattles the performer. The interactive companion on this page does the word-to-runtime math both directions so you can size the script before you book anyone.

Cast the voice, do not just book a rate

Voice casting is casting. The demo reel tells you what someone can do on their best day with copy they chose. It does not tell you how they handle yours. When it matters, pay for a custom audition on two or three lines of the actual script. Most working VO artists will do a short paid audition, and a $75 to $150 audition fee on a job that carries real usage is cheap insurance against a full session with the wrong voice.

Know what you are buying on rate. Non-union is where most regional and web work lives. A non-union web or explainer read commonly runs $250 to $600 for a short piece, and per-finished-minute narration sits around $150 to $400 a minute depending on the artist and the usage. A non-union 30 or 60-second commercial with regional web and social usage is often in the $350 to $750 range. Union work under SAG-AFTRA carries scale session fees plus separate use fees that recur by market, cycle, and media, and those use fees are frequently larger than the session fee itself. The number you are quoting is a session-plus-usage number, not a flat buyout, unless the contract says buyout in writing. Get the usage scope on paper before the session, because renegotiating usage after a spot is running is the worst possible time to do it.

One more casting note that saves grief later: book a voice that matches the edit's energy, not the energy you wish the brand had. A calm, credible read on frantic copy fights itself. Cast to the cut.

Run the session like a shot list

A session has a shape. Give it one before you dial in. Have the final script locked, numbered by line or paragraph, with pronunciation notes for anything a stranger would guess wrong: product names, place names, acronyms you want spelled out versus said as a word. Ambiguity in a script name costs you a pickup session later.

The session itself moves in a predictable order. Slate and level-set first: the artist reads a line so you can set input gain and hear the room. Then a full read top to bottom so both of you hear the whole thing at tempo. Then you work it in sections, giving direction and stacking takes per line or per paragraph. Most lines want two to four reads: a straight one, a warmer or harder variation, and a wildcard where you ask for something you are not sure will work. You keep the circled takes and move on. Do not chase a fifth and sixth take on a line that was already good on take two, because diminishing returns are real and a tired voice at minute 50 is worse than a fresh voice at minute 20.

Remote is the default now. Source-Connect, ipDTL, and a plain phone patch alongside a local booth recording all work. The artist records locally at full quality and you direct over a lower-latency monitor feed, so the file you get is the pristine one, not the compressed line you heard live. Confirm before the session which one you are getting, because a directed session over a phone where the phone is also the recording is a session you will be reshooting. Budget the clock honestly. A short spot session runs 20 to 45 minutes of actual work, but studios and artists book in blocks, so a one-hour minimum is normal. Longer narration scales with word count and reads per line, and the take count climbs fast. The session estimator on this page will show you why: a 900-word narration at two reads a line is 60-plus takes before pickups.

Give direction the artist can act on

This is the whole game, and it is where most directors freeze. The failure mode is the adjective. Telling a performer to be "more energetic," "warmer," "more natural," or "punchier" is not direction. It is a description of the result you want with no instruction for how to get there, and the artist is left to guess which of a dozen things you mean by "energy."

Usable direction is physical and specific. Instead of "more energy," try "lift the end of each sentence instead of dropping it, and take a beat before the product name so it lands." Instead of "sound natural," try "read it to one person across a table, not to a room, and throw away the first three words like you are mid-thought." Instead of "warmer," try "drop your pitch a touch, add a little air, and imagine you are reassuring someone who is nervous rather than selling to them." Every one of those gives the performer a lever to pull.

Two techniques carry a lot of weight. The first is the "as if" frame: give the artist a scenario instead of a mood. "Read it as if you just figured out the answer and you are a little smug about it" produces a specific, repeatable performance. The second is the reference line: read the line yourself, badly, just to show the rhythm and the stress you are hearing in your head, then let them do it properly. You are not asking them to imitate you. You are handing them the map of where the emphasis goes. And when a take is right, say so plainly and say why. "That one, the way you leaned on 'built' and let the last line breathe" tells the artist what to protect on the next take. Direction is not only correction.

Deliver clean so the edit stays clean

The performance can be perfect and still cost you a day if the files come back wrong for your edit. Set the delivery spec before the session, not after.

For a video edit, ask for 48 kHz, 24-bit WAV, mono, one file per take or a single continuous file with clean pauses between lines. Forty-eight kilohertz matters because it matches the sample rate of essentially every video timeline, and a 44.1 kHz file dropped into a 48 kHz sequence resamples and drifts. Ask for the raw, un-effected files, not a version the artist already compressed and EQ'd to taste, because your mixer wants to make those calls against the rest of the bed. Get handles: a half second of clean audio before and after each line so nothing clips a breath. Get 10 to 15 seconds of room tone, the sound of the booth with nobody talking, which your editor uses to patch gaps and smooth edits invisibly. Ask for peaks around -6 dBFS so there is headroom, and let loudness normalization happen in the mix, not in the booth.

Different destinations want different targets. Broadcast television in the US is governed by the ATSC A/85 loudness standard at -24 LKFS. Podcasts and audio-only pieces land around -16 to -19 LUFS. Web and social video sit near -14 to -16 LUFS. You do not need the artist to hit those, but you do need to know which one you are mixing to so nothing gets normalized twice. Name the files so a stranger can sort them: project, spot, version, take. And back the raw session up in two places before anyone touches an edit, because the one file you deleted is always the circled take.

The takeaway

A voiceover session is a small production, and it obeys the same rules as a big one. The script is prep. The casting is a hold. The session is the shoot day, and it runs well or badly on the quality of the direction, not the size of the rate. Write for the ear, cast to the cut, direct with instructions a person can actually perform, and lock your delivery specs before anyone opens a booth. Do that and the read comes back sounding like someone meant it, which is the only thing the audience was ever going to hear.