Edit a Table-Read Recording Into a Clear TV Pitch Sample
Edit a Table-Read Recording Into a Clear TV Pitch Sample
A table-read recording in a pitch has one job: let someone who wasn't in the room hear two actors make a scene work. Levels, trims, fades — all of it serves that. An edit that makes the scene sound smoother while making it mean something else has cost you the only thing the recording was there to prove.
So the order is short. Choose the exchange before you clean it. Leave the original file alone and build a session beside it. Fix what is technically wrong — a quiet microphone, a chair, a slate — without moving the words. Then compare the result with the recording and export a file you have listened to from start to finish, labeled as what it actually is.
That last step carries more weight than it sounds. A tidy twenty-five-second excerpt is easy to mistake for evidence that the scene works on screen. It isn't. It's evidence that the scene works on the page and in two actors' mouths, which is worth demonstrating and is not the same claim.
Choose the exchange before you clean the recording
Name the question the sample has to answer before you touch a fader. Is it about a relationship, a tonal turn, a rhythm of interruption? Whatever it is, the passage has to contain the answer, and it usually has to contain the setup too. A reply that lands because of what preceded it will not land in a pitch if you cut the setup out to save six seconds. The listener has no context except what you give them.
So select the fullest version of the exchange that still fits: what prompts it, the turn, and the response that shows the turn registered. Then stop. You are not sampling the scene, you are sampling the exchange.
Two things belong here rather than later, because they are not editing decisions. First, permission. Someone authorized the recording; that is not the same as authorizing an excerpt, and it is not the same as each performer agreeing to be in a pitch. Confirm the excerpt, the intended use and the people included, in writing, before you spend a day on it. Documentation describes a tool; it doesn't grant you rights to anyone's voice.
Second, source preservation. Note the file, the take, who is speaking and the start timecode of the passage you want, before you open anything. It takes four minutes and it is the difference between "let's try the other reading" and "we can't find where that came from."
Keep the source, and build the session beside it
Here's a constructed example to work through. Two actors, three tracks — one microphone per actor plus a room mic — a scene from a fictional pilot, and the passage the pitch needs. Joel's track sits roughly 10 dB below Nadia's. A stage manager slates the top: "Scene 14, Spare Key, take 3, March 4." The exchange runs a little over twenty seconds:
NADIA: Spare key? I'm back after midnight Thursday.
JOEL: Yeah.
[2 seconds]
JOEL: I lost it. Last week, maybe Monday.
NADIA: You lost the spare key.
JOEL: I'll get another one cut.
NADIA: You've been making me tea all week.
The turn is Nadia's last line. She doesn't react to the lost key; she reacts to the week of tea, which the admission has just explained. Everything before it is there to make that line land.
I did not record, edit or listen to any audio while writing this, and neither did anyone else on my behalf. The session above is invented, not captured. Run the same sequence on your own authorized material.
Put the recording into a new multitrack session as a copy, and keep the original out of the working folder. Adobe's Audition documentation draws a line worth respecting here: multitrack clip edits are a different operation from waveform edits, and the help page on arranging and editing multitrack clips describes clips as arranged and trimmed without altering the underlying file. Open the waveform and cut, and you're working on the recording itself. Do that once, on the wrong afternoon, and the take you wanted is gone from the file everyone else is still using.
Name the tracks after the characters or the actors. "Audio 1" and "Audio 2" will be a puzzle in a week, and the puzzle will be about someone else's voice. Put the take number in the clip name, not only in the file name.
The same help page distinguishes copies that reference the same source audio from copies that duplicate the media, and it lists snapping among the arranging behaviors. Both distinctions matter in practice. A reference copy follows the original, so edits you expect to affect one instance may affect several; an independent copy doesn't, which is what you want when you're going to process one line differently. Snapping will butt a take neatly against its neighbor, and it will fight you when you want a deliberate overlap. Turn it off on purpose rather than deciding the software is being difficult.
Leave handles: keep several seconds of audio outside your intended in and out points in the session. The excerpt will get shorter after review, or longer, and you want to make that change in the session rather than by reopening the recording.
Finally, find the slate. It is loudest on the room mic, but a stage manager speaking in the room usually bleeds onto every track in the file, which is why removing it is a whole-session trim at the head, not a tidy cut on one channel. If you leave it in, the first thing a pitch listener hears is "take 3."
Level the voices without moving the words
A 10 dB gap between the two speakers is not an aesthetic problem, it's an intelligibility problem, and this is the repair that carries the least risk. Raise the quiet clip as a whole. If the actor leaned in and out across the passage and one static change leaves the ends too loud and the middle too soft, write a level change over time rather than adding more cuts — every extra boundary you create is another join to check later. Names and locations for these controls vary by version, so match the behavior, not the menu.
If routing is available in your session, a bus that both actor tracks feed is a cheap convenience: one gain move then applies to the dialogue as a pair, and the room mic stays on its own path. Adobe's page on multitrack routing and EQ controls notes that track signals can be routed through buses and outputs, and that polarity inversion is available as an operation. Available is the operative word. Inverting polarity is not a general repair for a weak microphone, a dull-sounding actor or an uneven pair of tracks. It's a specific tool with specific uses, and its presence in a list is not a diagnosis.
Two noise decisions in the example. The slate comes off the head of all three tracks — that's a trim outside the exchange and changes nothing inside it. The chair scrape is different. It lands inside the gap, on the room mic, while neither actor is speaking. Lowering the room track gently across that second and smoothing the edges reduces the distraction without touching Joel's timing. A hard, deep dip will punch a hole in the room tone that may be more noticeable than the scrape was, and deleting the scrape entirely would shorten the gap — which brings us to the part of this job that actually has teeth.
The two seconds you'll be tempted to delete
The gap between Joel's "Yeah" and his admission runs a little over two seconds. It is the scene. It's where a man decides whether to say the thing, and Nadia's final line is the payoff for it.
Now build the other version, hear it, and then don't ship it. Close the gap. "Yeah. I lost it." The line arrives rehearsed, decided in advance, and Joel stops being someone confessing and becomes someone reporting. You have saved two seconds that no pitch reader was counting, and you've replaced the turn with a summary of the turn.
In a multitrack session, the operation that does this is usually the ripple kind: the selection is removed and what follows moves up to fill the space. Whether that happens on the selected track alone or across the session depends on what you've selected, and the Audition help page separates Delete from Ripple Delete for exactly this reason — check what your selection triggers before you use it. Either outcome is bad in a specific way. A single track that closes its gap slides out of step with the other speaker. A session-wide ripple moves the whole excerpt off the recording's timeline, and now the source comparison in the next step is guesswork instead of a check.
So the rule is narrow and useful: fix noise, don't fix timing. A performed hesitation, a breath, a beat before an answer — those are information about the character, and your cut is not neutral about them. Removing one changes whether a reply reads as reluctant, surprised, eager or immediate. If a pause is genuinely a technical accident — a dropped page, somebody's phone — you can reduce or remove it, but say what you removed. If you can't tell whether a hesitation was performed or accidental, leave it and note the doubt. Nobody has ever lost a pitch for a two-second pause in a good scene.
Check every join against the source
Go through the excerpt against the recording, boundary by boundary, and ask what your edit did to the time and to the meaning.
Start with the ones you know about. The slate is gone from the head of all three tracks. The transition into the exchange is a short fade, not a hard edge, so it doesn't click. The room track has a shallow dip where the chair was. Those are all describable in one sentence each, which is the test — if you can't describe an edit plainly, you may not have decided what it does yet.
Then the ones you'd rather not think about. Suppose a colleague preferred the admission line as delivered in take 2, and the temptation is to drop that reading into the take 3 gap. You can do that. What you then have is a composite: two performances presented as one continuous exchange. If you ship it, declare it — in the session log and in whatever text accompanies the file — because the moment a listener assumes they heard one unbroken performance and later learns otherwise, every other excerpt you send them gets audited. In this example, the team built the swap, heard how much the voice changed between takes, and stayed with take 3 throughout.
Keep a short log with the excerpt. It doesn't need to be elaborate:
- Source file, take number, date of the read
- In and out points of the excerpt, in the source's time
- Speakers, and the track each is on
- Every removal, every level change that isn't a plain clip gain, every substitution
- Who approved the excerpt, and when
Mark the session too — excerpt in, excerpt out, slate removed, substitution — so the reasons travel with the project instead of living in your memory of Tuesday.
Export a labeled excerpt, and say what it is
Listen to the exported file, on the device the listener will probably use, all the way through. Not the session — the file. Exports surprise people: an envelope that looked smooth, a fade that turned out to be on the wrong side of a cut, a channel that collapsed when it hit a phone speaker.
Choose the format for the destination rather than by habit. If the excerpt is going into someone else's edit, send it uncompressed; if it's going into a deck or an email, a compressed file is fine. Ask about loudness if the pitch has other audio it will sit beside, and match that, rather than adopting a single target for every pitch you ever send. Then check the result against the other audio in the deck by ear.
Label it as a table read, everywhere it appears: in the file name, in the caption, in the sentence you write above it. "Table read excerpt, Scene 14, unrehearsed, March 4. Twenty-five seconds. The pause before the admission is performed." That last clause is not decoration. The gap is dramatic information, so the text alternative for anyone who can't hear the file has to carry it too — a transcript with the pause marked, in the same notation you'd use for a deaf or hard-of-hearing reader, preserves the same turn that the audio preserves. A transcript that runs the lines together loses it.
And when the excerpt is approved, freeze it and version the file. If you revise it, change the number, so nobody pitches a version that hasn't been checked.
A clean reading excerpt is still a reading. It tells a listener what the scene does in two actors' mouths on a Thursday afternoon, which is often exactly what a pitch reader needs to decide whether to keep reading. It doesn't tell them what the camera will find, how the cast will play it, or what the final mix will sound like — and the label should say so. Protecting the pause and naming what the file is are the same instinct: the sample is only persuasive while it stays honest about what it is.
Frequently asked questions
What should be chosen before cleaning a table-read recording?
Choose the exchange the sample must answer. Include the setup, the turn, and the response that shows the turn registered; cutting the setup to save a few seconds removes the context a pitch listener needs.
How should the source recording and working session be handled?
Keep the original out of the working folder and build a new multitrack session beside it. Leave handles beyond the intended in and out points, name tracks after actors or characters, and treat waveform edits as different from multitrack clip edits.
Which timing changes are acceptable in a table-read excerpt?
Fix noise, not timing. A performed hesitation or beat before an answer is character information; reducing a genuinely accidental pause requires saying what was removed. If you cannot tell, leave it and note the doubt.
What has to be declared if an alternate take is dropped into the excerpt?
The result is a composite of two performances. Declare every substitution in the session log and in the text accompanying the file, because a listener who assumes one unbroken performance may audit later excerpts.
How should the exported excerpt be labeled?
Label it as a table read wherever it appears, state its source and limits, and carry the pause in a transcript for anyone who cannot hear it. Listen to the exported file start to finish, choose format for its destination, and freeze and version it when approved.