The voice over demo: what a casting agency actually listens for

A voice over demo is rarely heard the way it was edited. You built it as a progression: the plainest first, the scale after that, the best kept for the end. It gets opened by somebody with a specific project in mind, a delivery date, and other applications waiting in the same folder. Almost every misunderstanding comes from that gap.

What happens in the first few seconds

In practice, the first extract is the demo. It decides whether the listening continues, and it does not need to be the most spectacular thing in the selection. It has to say straight away what your voice does.

What stops the listening, on the other hand, almost never has anything to do with talent. A room you can hear: reverb, the hollow sound of an empty office, continuous hiss, background noise the ear catches before the first syllable. A music intro or an opening fade, which delays the one piece of information anybody is waiting for. A spoken self-introduction at the head of the file, taking up the most valuable seconds to repeat what the covering message already says.

What makes someone listen on comes down to very little: a clean voice, and an extract that resembles what you do most often.

A short, dense demo beats a long, complete one

Trade practice puts a voice over demo at around 60 to 90 seconds. That length is not a courtesy convention, it follows from what the person listening is actually doing: sorting, not assessing a career.

Past a minute and a half, a demo stops adding information and starts diluting it. Every average extract pulls down the average of the whole, and it is that average that stays in the memory, never the one strong passage. A four-minute demo does not show four times as much. Mostly it shows that a choice was not made.

The edit counts as much as the selection. Cut each extract on its strong moment instead of letting it finish the sentence, keep the silences between sequences short, and avoid the long fades that feel like filling time.

“If I cut, I lose half of what I can do.” True, and that is exactly the call being asked of you. A demo reel exists to open a door. The rest of the catalogue comes afterward, when a specific request turns up: keep your other extracts ready, sorted by register, and mention them in one line in your message.

Show a real range rather than one performance

Range is measured in registers you can hold. Three or four clearly distinct worlds, a few seconds each, say more than ten variations around the same tone.

The trade recognizes fairly stable families of delivery: advertising, which calls for a familiar, energetic tone; documentary narration, which asks for settled diction; training and e-learning, which live on instructional clarity; the trailer, which accepts theatricality. A useful demo shows which ones you hold, and in what order of preference.

Then comes the hardest rule to apply to yourself: show what you can do, and leave out what you would like to do. An extract placed out of wishful thinking sounds like exactly that. The calculation is bad in both directions: if it works, it triggers a call on a register the session will not confirm, and it is that session that becomes the reference, long before the demo.

For an actor starting out, an extract recorded specially for the demo is still legitimate, on two conditions. That the situation is realistic. And that it is never presented as a broadcast credit: a believable scenario well played stands on its own, a false credit gets checked with one message.

From the booth

When you listen to a demo, the first thing to set aside is your own taste. A voice you would not have chosen for yourself can be exactly right on a project, and the reverse holds just as often. The second observation is more useful still: the extract recorded fifteen times almost always sounds less alive than the first take, because over-control eats the spontaneity before it corrects the fault. On a demo that shows more than anywhere else, with no context to redeem it.

Technical quality: the room first, processing as little as possible

The room comes before the gear. A sensitive mic in a bare space mostly reveals the faults of the space, and recorded reverb does not come out in the edit, you live with it. A hand clap in the middle of the room settles it, and the verdict lands in one second.

Level next. If the person listening has to touch the volume between two extracts, the edit is not finished. A demo holds an even level from the first extract to the last, music included, with no passage that makes you lean toward the speaker and no peak that makes you jump.

Processing last, and this is where a lot of applications penalize themselves. A demo compressed to the ceiling, de-noised into artifacts and set in an ambient reverb ends up saying more about the processing chain than about the voice. The reflex makes sense, it is trying to hide the weaknesses of the take. It also hides the timbre, which is the one thing the agency is trying to hear.

There is a production reason behind that requirement. A multi-voice casting brings together actors who each record at home, with microphones and rooms that have nothing in common, and the studio has to create the illusion that those voices came from the same place. That matching is built on clean, lightly processed material. A voice that arrives saturated with effects joins onto nothing.

The mistakes that keep coming back

Impersonating a famous voice. It proves you have a good ear, and it erases exactly what makes a voice castable. There is no ideal timbre in voice over: low, settled timbres open documentary and training, unusual timbres open character and video game work. An impression puts the listening into comparison, and that comparison is never won.

The character demo with no direction. A run of voices and accents strung together outside any situation says nothing about your ability to hold a role under direction. Dubbing asks for rhythmic precision and acting hooked to a picture, not a collection of timbres. If that is your ground, show the character in a situation: an intention, someone to speak to, an imposed pace.

The dutiful read. Clean articulation, steady tempo, nothing to correct, and nothing happening. It is not a reading problem, it is an intention problem. Two questions are enough to expose it before you record: who are you speaking to, and what do you want from that person once the sentence ends?

The badly named file. A file called “demo_final_v3” in a folder holding dozens of them becomes unfindable within weeks. Your name, the language, the type of demo: ten seconds of work that decide whether your voice can be found again six months later, at the moment a project matches it.

Sending with no language or accent declared. The costliest variant, and it deserves its own section.

Languages and accents: declare at the level you actually hold

FRVOICEOVER works in 6 native languages: French, English, Spanish, Italian, German, Arabic. A demo that arrives with no language and no accent stated cannot be filed, however good it is.

Three levels are worth separating, and they get confused often: the languages you speak, the ones you act in, and the ones you can take direction in during a session. Only the last two interest a casting.

Do not overstate anything here. A declared language that is not native is audible from the first sentence, in where the stress falls and in the rhythm long before the vocabulary. The real risk is not at the moment you send: it comes later, in session, in front of a client, in a language you do not hold.

Same logic for regional accents. An accent held over one demo sentence is not an accent held over a twenty-minute training module. Say what you hold over the distance, and say it plainly.

What has to come with the demo

An agency does not only listen for today’s project, it files. That filing needs information the demo does not contain:

• the languages and accents, at their real level;
• the registers you hold, and the ones you would rather not do;
• your working conditions: home studio or not, ability to be directed remotely, usual turnaround;
• one sentence that says what your voice does.

That last line is the most useful thing in the whole application, and the rarest. A vocal identity fits in one sentence as soon as it is clear.

Then the housekeeping: a link that works, that does not expire, and a file that opens without creating an account. A submission that forces someone to write back asking for the language, the accent or a valid link becomes hard to use. At FRVOICEOVER the listening is done by voice direction, with sound engineering in the same team: a small team, in other words, where that follow-up email is precisely the one that does not happen in a busy week.

Before you send

Go through the list, honestly:

• a room you cannot hear;
• 60 to 90 seconds, strongest extract first;
• three or four registers you genuinely hold;
• an even level and discreet processing;
• a file named with your name and the language;
• languages and accents declared at the level you hold them;
• one sentence that says what your voice does.

If it is all there, your submission is usable. If a line is missing, take it back before sending: a demo reworked three weeks later beats a submission nobody can file. And if the list tells you the moment has not come, it has done you a service, because a demo reel gets prepared and never improvised.

The FRVOICEOVER casting brings together more than 100 voices, in every style, every working language of the studio and every age, and it stays open to applications. We promise neither an individual reply nor a place in the catalogue. What we do say fits in one sentence: a demo that meets these conditions goes on the join the catalogue page, and it arrives there in a state where it can be used.