
In Summary Before you send a video to a client, watch the exported file, not the timeline, and check five things: picture, sound, text and captions, legal and brand, and the file itself. Measure what has a number (loudness, true peak, video levels, caption line length) against the client's delivery spec. If the client has no spec, use the published standard for where the video will run. Below is a printable checklist with every threshold taken from the standards body's or platform's own document, and the menu paths to measure them in Premiere Pro and Final Cut Pro.
The short answer
Quality control (QC) is the last pass before delivery. It asks whether the file is correct, not whether the edit is good. That second question belongs to the rough cut review. Shootsta's 2026 QA guide puts the difference simply: "QA asks whether the video is correct. Review asks whether it is right."
Run the pass in this order:
- Get the spec. Ask the client for their delivery spec before you export. Their numbers win over any standard below.
- Export the deliverable. QC the file you will send, not your timeline. Exports can change color, sound and captions.
- Measure first. Loudness, true peak and video levels have numbers. Check them before you watch.
- Watch once, start to finish, with good headphones. Note every problem with a timecode.
- Read every word on screen and in the caption file. Names, titles, numbers, dates, URLs.
- Check the file. Name, format, length, versions.
- Fix, re-export, and check the new file again. A fix can break something else.
Before you start: whose numbers apply?
Most corporate and agency work has no broadcast spec. That does not mean anything goes. ATSC A/85, the US recommended practice for TV loudness, says its −24 LKFS target is for content delivered "where there is no prior arrangement by the parties regarding Target Loudness". In other words, an agreed spec comes first.
So write the target down before you start the pass:
| Where it runs | What to measure against |
|---|---|
| A client or partner with its own delivery spec | Their spec. Ask for it in writing. |
| European broadcast | EBU R128: −23.0 LUFS, true peak no higher than −1 dBTP |
| US broadcast | ATSC A/85: −24 LKFS (dialogue for long-form content), true peak no higher than −2 dBTP |
| A streaming service | A/85 recommends one target between −23 and −27 LKFS for streaming services, unless agreed otherwise. Ask which one. |
| Web, social, intranet, events | None of the sources below publishes a single number. Agree one target with the client and use it on every video in the series. |
Platform codecs, resolutions and frame rates are covered in our export settings by platform guide, with each value from the platform's own page. This checklist does not repeat them.
The checklist
Print this section, or copy it into your project doc. Each line is pass or fail.
1. Picture
- No black frames, flash frames or single stray frames at cuts.
- No missing media, offline clips or render placeholders.
- Every shot matches its neighbours in color and exposure, unless the change is intended.
- No unintended color or gamma shift between your viewer and the exported file. Play the file in the player the client will use. (Resolve users: see fixing color shift after export.)
- Titles, logos and captions sit inside the frame on the delivery aspect ratio.
- Video levels within range, if the spec asks for it. EBU R103 gives the ranges in code values:
| Bit depth | Nominal video range | Preferred min / max | Total signal range |
|---|---|---|---|
| 8-bit | 16–235 | 5–246 | 1–254 |
| 10-bit | 64–940 | 20–984 | 4–1019 |
For pre-produced and graded material, R103 says "the nominal limits contained in this document should be followed closely". Values outside the preferred range count as a gamut error. It also recommends that measuring equipment flag an out-of-gamut occurrence only once the error covers more than 1% of the image. R103 describes the nominal range as 0 to 100% of the video range between black level and peak.
2. Sound
- Dialogue is clear all the way through. Listen on headphones, then on laptop speakers.
- No clicks, pops, hum, dropouts or clipped peaks.
- No sudden jumps in level at cuts.
- Music and effects sit under speech, never over it.
- Both channels carry the right audio. A voice in only one ear is easy to miss on laptop speakers.
- Integrated loudness measured over the whole programme, against the target you wrote down. EBU R128 says to measure "in its entirety, without emphasis on specific foreground elements such as speech, music or sound effects". ATSC A/85 measures dialogue for long-form content instead, so use the method of the standard you deliver to.
- Tolerance. R128 allows ±1.0 LU only where the target "is not achievable practically (for example, live programmes)". A/85 expects about ±2 dB of measurement variation and says loudness "should not be targeted to the high or low side of this tolerance".
- True peak at or below the limit: −1 dBTP for R128 in production, −2 dBTP for A/85.
3. Text and captions
- Every name and job title checked against a written source, not memory.
- Numbers, dates, prices, URLs and phone numbers checked.
- No placeholder text ("Name Surname", "Lorem ipsum", "TK").
- Captions present if the video is published on the web. WCAG success criterion 1.2.2, Captions (Prerecorded), is a Level A requirement. WCAG says captions include dialogue, identify who is speaking, and include meaningful sound effects.
- Captions match the words spoken and stay in sync.
- Line length and lines per caption within the spec. Two published examples:
- Netflix: 2 lines maximum, 42 characters per line for English, and each subtitle on screen for between 5/6 of a second and 7 seconds.
- CEA-608 broadcast captions: 32 characters per line. iTT allows two lines per caption (four in Japanese).
- The caption file is in the format the destination accepts. YouTube's basic formats include .srt, .sbv/.sub, .mpsub, .lrc and .cap, and it suggests SubRip (.srt) or SubViewer (.sbv) if you are new to caption files. Those two must be "plain UTF-8" with no styling. For positioning, YouTube supports it in WebVTT (.vtt) and in iTT, a subset of TTML. It prefers .scc for CEA-608 captions.
- If you burn captions in, they do not cover a face or a lower third.
4. Legal and brand
- The current logo, colors, fonts and end card from the client's brand kit.
- Everyone identifiable on camera has signed a release.
- Music and stock licences cover this use, this territory and this length of time.
- Nothing confidential in frame: a screen, a document, a whiteboard, a badge, a number plate.
- No third-party logo the client has not cleared.
- Every claim and quote approved by whoever owns it (legal, product, the person quoted).
- Required disclaimers and credits present and readable for long enough.
5. File and spec
- Codec, container, resolution, frame rate and aspect ratio match the spec.
- Running time matches the spec, including any slate, black or countdown it asks for.
- Audio channel layout matches the spec (stereo, or split dialogue, music and effects).
- File named the way the client asked, with the version.
- One file per destination: the 16:9 master, the vertical cut, the clean version without text.
- Caption file, transcript and thumbnail included, if the client asked for them.
- The file opens and plays to the end on a second machine.
Where to measure in your editor
Premiere Pro. Open Window > Audio Track Mixer. In the first effect slot of the Mix track, open the effect menu and choose Special > Loudness Meter. Right-click the meter (Ctrl-click on macOS) and choose Edit to set a preset, then play the sequence from start to finish. Adobe's page explains that Integrated "averages the programme from start to finish". The meter measures; it does not change your level. For picture, the Lumetri Scopes Waveform set to Luma shows IRE from −20 to 120. See our Premiere audio normalizing answer and our Premiere SRT export answer.
Final Cut Pro. Final Cut checks captions as you edit, in its three caption formats: CEA-608, iTT and SRT. A caption that breaks its format's rules turns red in the caption lane, and the Validation section of the Captions inspector says why. Overlapping captions can be fixed with Edit > Captions > Resolve Overlaps.
DaVinci Resolve. See our Resolve audio normalizing answer.
Whatever you measure in, measure the exported file when you can, not only the timeline.
Where Eddie fits
Eddie does not run a technical QC on a file. It does not report codec, gamut or caption-file errors. It does help with several lines on the list, as long as you know where each result shows up.
- Loudness, in the cloud MP4. Under Export ▸ All export options… ▸ Video file, Final loudness sets the level of the whole mix: Web & social (−14 LUFS), which is the default, Broadcast EBU R128 (−23 LUFS), Broadcast US ATSC A/85 (−24 LKFS) or Off. For a broadcast delivery, change it from the default before you render. It applies to cloud MP4 renders only. The player does not preview it, and it does not change a Premiere Pro, Final Cut Pro or DaVinci Resolve export. Auto-level speech (−16 LUFS) brings each speaker to the same level before the mix and does reach those editor exports, as clip gain. Both apply to the next export; a file already exported does not change. See audio cleanup.
- Reading levels. Eddie can read the sound of one source over a window of up to ten minutes: short-term loudness in LUFS, the level of every 100 ms in dBFS, the noise floor, and the silences. It reads a source, not your finished mix, so measure the delivered file in your editor or meter.
- Captions and names. Eddie adds a caption track from the transcript's word timings. Filler words are left out of the caption text by default, which matters if the client wants verbatim captions. When a name is misheard, correcting the transcript fixes every matching occurrence and rebuilds the caption tracks built from it. An edit with captions has free Captions (.srt) and Captions (.vtt) downloads; SRT keeps the words and timing only. See captions in your editing app and Eddie's captions and titles.
- Hiding something in frame. Eddie can blur, pixelate or block a rectangle on a segment to hide a face, a plate, a screen or a document. The rectangle is static and follows nothing. It shows in playback and is baked into rendered video; editor project files cannot carry it.
- A second watch. Eddie can render an edit and return notes on story, pacing, the hook, audio issues and a cut list with timestamps, as text or as timestamped comments on the timeline. Treat it as a second pair of eyes, not a spec check. See Feedback v2.
- Client sign-off. Each edit has a review link: a web page where the client watches that cut and leaves notes at the timecode without an Eddie account. If the client reviews in Frame.io, see the Frame.io workflow.
What it will not do
- A checklist does not replace the spec. If the client has one, every number here gives way to it.
- Standards do not cover the web. None of the standards bodies or platforms cited here publishes one loudness number for web video. Pick one with the client and keep it.
- Eddie's final loudness does not reach editor files. It is set in the cloud MP4 only. If you finish in Premiere, Final Cut or Resolve, set and measure loudness there.
- Eddie does not check the delivered file against a spec. Measure codec, levels and caption files with your editor or a QC tool.
FAQ
What should I check before sending a video to a client?
Picture, sound, text and captions, legal and brand, and the file itself. Check the exported file, not the timeline. Measure loudness and true peak against the client's spec, read every name and number, and play the file to the end on a second machine.
What loudness should a corporate video be?
There is no single published standard for web video. Use the client's spec if there is one. For broadcast, EBU R128 sets −23.0 LUFS and ATSC A/85 sets −24 LKFS. For streaming, A/85 recommends one target between −23 and −27 LKFS. Otherwise agree a number with the client and use it across the series.
What are broadcast-safe video levels?
EBU R103 sets the nominal range at 16–235 in 8-bit and 64–940 in 10-bit, with a preferred tolerance of 5–246 and 20–984. Graded material should stay close to the nominal range.
How many characters per line should captions have?
It depends on the spec. Netflix allows 42 characters per line for English and two lines at most. CEA-608 broadcast captions allow 32 characters per line.
Which caption file should I deliver?
The one the destination accepts. SRT is the simplest: YouTube takes it as plain UTF-8 text without styling. For broadcast, ask for the format; YouTube prefers .scc for CEA-608 captions.
Who should do the final QC pass?
Someone other than the editor if you can. Shootsta's guide warns that an editor checking their own work "misses the same things twice". In a small team, one person can do it as a separate pass after editing is finished.
Sources
- EBU, R 128: Loudness normalisation and permitted maximum level of audio signals (2023; −23.0 LUFS target, ±1.0 LU only where the target is not achievable practically, −1 dBTP true peak, measure the whole programme)
- ATSC, A/85:2026-07 Loudness and True Peak Quick Reference (−24 LKFS without prior arrangement, about ±2 dB measurement tolerance, −2 dBTP, −23 to −27 LKFS for streaming), from the A/85 document page
- EBU, R 103: Video signal tolerance in digital television systems (version 3.0, 2020; nominal, preferred and total ranges in code values; 1% out-of-gamut threshold)
- W3C, Understanding Success Criterion 1.2.2: Captions (Prerecorded) (Level A; what captions include)
- Netflix Partner Help, Timed Text Style Guide: General Requirements (2 lines maximum; 5/6 second to 7 seconds per subtitle) and English (USA) Timed Text Style Guide (42 characters per line)
- Apple, About Final Cut Pro caption validation (CEA-608, iTT and SRT rules; 32 characters per CEA-608 line; Resolve Overlaps)
- YouTube Help, Supported subtitle and closed caption files (basic formats .srt, .sbv/.sub, .mpsub, .lrc and .cap; .srt and .sbv in plain UTF-8; positioning in WebVTT and iTT; .scc preferred for CEA-608)
- Adobe, Apply Loudness Meter to a track, Measure the loudness of your track and Available Lumetri Scopes
- Shootsta, Video Editing QA: The Complete 2026 Guide (QA versus review; who runs the pass)
Delivering a cut this week? Start a project for free and set the final loudness before you render the MP4.


