In the US, the only federal loudness law for commercials is the CALM Act, and it binds TV stations, cable operators and other pay TV distributors, not streaming apps or social platforms. FCC rules (47 CFR 73.682(e) for stations and 47 CFR 76.607 for cable and satellite) make the ATSC A/85 recommended practice mandatory for commercials, and A/85 sets -24 LKFS, measured over the full mix with ITU-R BS.1770, as the target for content delivered without loudness metadata, with true peaks below -2 dBTP. Streaming publishers and social platforms set their own targets, with one exception: since July 1, 2026, a California law bars streaming services serving California viewers from transmitting ads louder than the content they accompany.
For a producer that means three things. Order a broadcast mix at -24 LKFS integrated for any TV or cable placement. Get each streaming publisher’s target in writing, because the newest ATSC guidance allows streaming services to choose anywhere from -23 to -27 LKFS. And treat social loudness as a decision the plan makes, because the social platform spec pages we checked publish no target. This guide explains each rule, how a spot is measured and which versions belong in the deliverable matrix. It describes the rules as of November 2026 and is general information, not legal advice: whether a rule applies to a given placement is for the media owner and the brand’s counsel to decide.
Which loudness rules are actually law?
Two federal rules and one state law are binding; everything else is a delivery spec agreed in a contract. The table sorts outlets by what governs them.
| Outlet | What governs loudness | Who the rule binds | Target or test |
|---|---|---|---|
| Broadcast TV stations | 47 CFR 73.682(e), under the CALM Act | The station | ATSC A/85:2013 (-24 LKFS where there is no metadata or other arrangement) |
| Cable, satellite and other MVPDs | 47 CFR 76.607, under the CALM Act | The cable operator or other MVPD | ATSC A/85:2013 |
| Streaming services with ads, for viewers in California | California Business and Professions Code section 22776 (SB 576), from July 1, 2026 | The streaming service | Ads not louder than the content, consistent with the FCC’s CALM rules |
| Other streaming and connected TV | No federal rule; publisher delivery specs | The advertiser, by contract with the publisher | The publisher’s stated target |
| Social and online video platforms | No loudness law; platform specs | The advertiser, by the platform’s terms | None published on the spec pages we checked |
The CALM Act in brief
The Commercial Advertisement Loudness Mitigation Act (Public Law 111-311), approved December 15, 2010, told the FCC to adopt a regulation “limited to incorporating by reference and making mandatory” ATSC A/85 “and any successor thereto”, “only insofar as such recommended practice concerns the transmission of commercial advertisements by a television broadcast station, cable operator, or other multichannel video programming distributor”. The FCC’s rules have applied since December 13, 2012. 47 CFR 73.682(e) applies them to television broadcast stations, and 47 CFR 76.607 applies them to cable operators and other multichannel video programming distributors (MVPDs), a category that includes satellite providers.
The word that limits the law is “transmission”. The rules govern what a station or MVPD sends to viewers. They say nothing about what an app, a website or a social platform plays, which is why a spot can be compliant on cable and still be the loudest thing in a streaming break.
Who is on the hook
The rules bind the station or MVPD, not the advertiser or the production company. They give those transmitters ways to show compliance. A station or MVPD may rely on a network’s or programmer’s published certification for commercials embedded in that programming. Some must run unannounced 24-hour spot checks of non-certified programming: stations with more than $14,000,000 in 2011 receipts, and MVPDs with more than 400,000 subscribers at the end of 2011 (all non-certified programming at 10,000,000 subscribers or more, a random 50 percent below that). For commercials they insert themselves, they are treated as compliant if they install, maintain and use A/85 equipment and keep records.
In practice the obligation travels upstream through delivery specs. The NAB’s CALM Act best practices summary (August 2012) recommends that a station “provide a program and commercial delivery specification to their program and commercial suppliers specifying a target loudness that matches what the station adopts”. That is how advertisers meet the CALM Act: as a line in a network’s spec sheet, a rejected file or a gain change applied to their spot, not as a letter from the FCC.
Which edition of A/85 is the law
The edition the FCC incorporated is A/85:2013, dated March 12, 2013 (47 CFR 73.8000 for stations and 47 CFR 76.602 for MVPDs). ATSC approved a newer edition, A/85:2026-07, on July 8, 2026. 47 CFR 76.602 states that “to enforce any edition other than that specified in this section, the FCC must publish a document in the Federal Register”, and as of November 2026 the eCFR still names the 2013 edition. Networks may write the 2026 edition into their own specs before the FCC acts; the core delivery numbers (-24 LKFS, about 2 dB of tolerance, true peak below -2 dBTP) are the same in both.
What does -24 LKFS mean?
-24 LKFS is the average loudness that A/85 recommends for content delivered without loudness metadata, measured with the ITU-R BS.1770 algorithm. LKFS stands for loudness, K-weighted, relative to full scale: K-weighting is a filter that approximates how the ear responds to different frequencies, and full scale is the loudest level a digital file can hold. A spot at -24 LKFS averages 24 dB below full scale on that weighted measure.
The measurement standard is ITU-R BS.1770, “Algorithms to measure audio programme loudness and true-peak audio level”; the version in force, BS.1770-5, was approved in November 2023. Section 6 of A/85:2013 sets the delivery numbers: “For delivery or exchange of content without metadata (and where there is no prior arrangement by the parties regarding loudness), the Target Loudness value should be -24 LKFS.” It anticipates measurement variations of “up to approximately ±2 dB”, adds that content loudness “should not be targeted to the high or low side of this range”, and asks for true peaks below -2 dB TP “to provide headroom to avoid potential clipping due to downstream processing (such as audio coding used in delivery)”.
| Term | Meaning, per A/85 and BS.1770 | The number to know |
|---|---|---|
| Integrated loudness | The average loudness over the whole content, start to finish, in LKFS | -24 LKFS for delivery without metadata |
| Tolerance | Measurement variation A/85 accepts around the target | About ±2 dB, not to be aimed at deliberately |
| True peak | The maximum level of the waveform between samples, in dBTP | Below -2 dBTP |
| Anchor element | The element a program’s loudness is judged by, usually dialog | Used for programs, not for ads |
| Full-program mix | Every element and channel of the soundtrack, except the LFE | How ads are measured |
| Dialnorm | AC-3 metadata telling the decoder how loud the content is | Fixed at 24 when a station runs everything at -24 LKFS |
Dialnorm explains why -24 became the default. The NAB guide describes the simplest station setup as a fixed target with a single encoder setting: “A/85 recommends a target loudness of -24 LKFS for stations wishing to use a fixed dialnorm value”, and to operate at that level “the dialnorm parameter must be set to 24”. In 2012 it reported that “most major networks have chosen to operate with a fixed target loudness of -24 LKFS”. When the whole chain runs at one fixed level, a commercial delivered at that level plays at the same loudness as the program around it.
The 2026 edition keeps the -24 target and the 2 dB tolerance, adds a tolerance of about ±0.5 dB for true peak meters, and says true peak should be measured before encoding because values “may be different post-decode” (A/85:2026-07).
How is a commercial measured: full mix or dialog?
A commercial is measured on its full mix over its whole duration, not on its dialog. A/85:2013 section 5.2.4 says that “a measurement of the long-term integrated (or average) loudness of the content’s full program mix over its duration should be reported as the Dialog Level of the short form content”, and it defines short-form content as advertising, promotional and public service material, typically under about two to three minutes. Programs are measured differently: around the anchor element, usually dialog. The 2026 edition’s quick reference table makes the split explicit, with long-form loudness measured with dialogue gating per BS.1770-1 and short-form loudness measured on the full-program mix with BS.1770-3 or later.
Four details change the number a meter reports:
- Silence. The FCC notes that “since 2013, the Recommended Practice has required that any completely silent passages during an advertisement be excluded when calculating its average loudness, but the same does not apply to very quiet passages” (FCC NPRM, Federal Register, March 11, 2025).
- Slate. The AICP delivery spec states that the first 7 seconds of slate and silence “are not part of LKFS measurement” and specifies the measurement as “Full Mix, not just dialog”. Measure the spot, not the leader.
- Meter mode. A meter set to dialog-gated mode measures speech only. That is the right mode for a program and the wrong one for an ad under A/85, and the two readings of the same spot can differ by several dB.
- Downmix. A/85:2013 asks for surround mixes to be measured both in 5.1 and in the 2-channel downmix, because “loudness build-up” in the downmix “can result in downmix loudness exceeding that of the surround mix by up to 3dB”.
A/85 also explains why an ad can measure correctly and still sound loud. It warns that in wide dynamic range short-form content, “the louder elements of this type of material will increase the loudness measured with a long term integrated method, and consequently reduce the perceived Anchor Element loudness after normalization.” In plain terms: a quiet opening and a loud ending can average to -24, and viewers hear the loud ending.
Example (fictional). A 30-second spot has 20 seconds of soft dialog that measures -30 LKFS on its own and a 10-second music tag that measures -20 LKFS. Integrated loudness averages energy, not decibels, so convert each section to relative power (10 to the power of LKFS / 10), weight it by its seconds, average, and convert back:
- Dialog: 20 x 10^(-3.0) = 20 x 0.001 = 0.020
- Tag: 10 x 10^(-2.0) = 10 x 0.01 = 0.100
- Average: (0.020 + 0.100) / 30 = 0.004, and 10 x log10(0.004) = -24.0 LKFS
The spot passes a -24 LKFS check, yet its dialog sits 6 dB under the target and its tag 4 dB over. The mixer rebalances: dialog to -25, tag to -22.5. Dialog: 20 x 10^(-2.5) = 0.0632; tag: 10 x 10^(-2.25) = 0.0562; average (0.0632 + 0.0562) / 30 = 0.00398, which is -24.0 LKFS again. The integrated number is unchanged, but the gap between talk and tag shrinks from 10 dB to 2.5 dB. The arithmetic is simplified (steady levels in each section, K-weighting already applied in the section readings); under the rule the FCC describes, only completely silent passages drop out, so both sections count.
What do networks ask for in practice?
Networks and stations turn A/85 into a fixed delivery target, usually -24 LKFS on the full mix with true peaks below -2 dB, and correct files that miss it. The AICP’s North American file deliverable spec, which a group of US industry bodies endorsed for commercial delivery, sets exactly that: -24 LKFS full mix, true peak below -2 dB, PCM audio at 48 kHz and 24-bit. Network and station sheets can differ, so the sheet attached to the media order governs.
The NAB guide also describes what happens when a file misses. “If the average loudness of any short form content item is measured with a difference of more than +/- 2dB from the specified target loudness value of the station, the average loudness of this content should be adjusted to match the target value by the difference, in units of dB.” File-based tools make “a one-time loudness shift to all elements of the content”, while inline processors at the end of the chain “may not be audibly transparent”.
So a spot delivered 3 dB hot is either turned down 3 dB in one move or compressed in real time by a processor the brand never hears in review. Neither outcome makes a hot mix sound louder on air than a compliant one, and the second can make it sound worse. The practical rule is to deliver on target and keep the dynamics you approved.
How a studio mixes to -24 without losing impact
A broadcast mix at -24 LKFS is a monitoring and metering discipline, not a filter applied at the end:
- Set the meter mode first. Use a BS.1770 meter in full-mix, integrated mode, the same mode the network will use for short-form content.
- Mix at a calibrated monitoring level so the target is reached by balance rather than by a limiter at the end.
- Control true peak with a true-peak limiter set below -2 dBTP, and check it again after the final bounce, before encoding.
- Keep the sections close. As the example shows, a narrow gap between dialog, music and the tag reads more evenly against the program than a wide one at the same integrated number.
- Check the downmix when delivering 5.1, against the same target.
- Measure the delivered file, not the session. Exclude slate and black, measure the full duration, and record the meter and its BS.1770 version in the QC report.
Mistakes that send a spot back
Most loudness problems come from measuring the wrong thing or measuring at the wrong moment, not from a bad mix. These are the ones to check for before delivery:
| Mistake | Why it happens | Fix |
|---|---|---|
| Measuring with slate and black included | The meter runs over the whole file, and 7 seconds of silence or tone shift the average | Measure the content only, as the AICP spec describes |
| Measuring in dialog-gated mode | The meter was left in the mode used for programs | Switch to full-mix, integrated measurement for every ad |
| Measuring the session, not the file | Bounce, sample rate conversion or encoding changes levels and peaks | Measure the exported file that will be delivered |
| Checking stereo only on a 5.1 delivery | The downmix can read up to 3 dB louder than the surround mix | Measure both the 5.1 mix and its 2-channel downmix |
| Aiming for the edge of the tolerance | A mix at -22.1 “passes” one meter and fails another | Aim at the target; A/85 says not to target the high or low side |
| Trimming a compliant 30 into a 15 without measuring | The shorter edit keeps the loud tag and loses quiet passages | Measure every length as its own file |
| Raising a quiet version by gain alone | True peak rises with the gain and can pass full scale | Re-limit and re-measure any version made louder |
| Mixing on small speakers only | Low end and sibilance that the meter counts go unheard | Check on calibrated monitors and on a TV-sized system |
The last three rows matter most for versioning, because they appear only after the hero mix has been approved and the cutdowns and language versions are built from it. A cutdown schedule that leaves no time to re-measure each file is the usual reason a compliant campaign ships one non-compliant version.
Audio post is part of post-production, which is where the broadcast mix, the stems and any other loudness versions are made from the same session.
What applies to streaming and connected TV?
No federal loudness rule applies to streaming or connected TV as of November 2026; one state law does, and publisher specs fill the rest. That is the most common gap in delivery plans, because a CTV spot often comes from the same edit as a cable spot and is assumed to follow the same rule.
The FCC asked, and has not regulated streaming
In March 2025 the FCC published a notice of proposed rulemaking on the CALM Act (MB Docket No. 25-72, FCC 25-16). It noted “growing concern with the loudness of commercials on streaming services and other online platforms” and asked: “To what extent does the Commission have the authority to address these complaints, or would such action require additional statutory authority from Congress?” It also stated that the NPRM “does not propose any specific regulations on streaming providers” and that the Commission “will not proceed with any such regulation against any streaming provider without first seeking public comment in a subsequent notice of proposed rulemaking”. Comments were due April 10, 2025, and replies April 25, 2025. As of November 2026, the eCFR text of 47 CFR 73.682(e) and 76.607 still names only stations and MVPDs.
California’s streaming law (SB 576)
California enacted SB 576 as Chapter 336 of the Statutes of 2025, approved by the governor on October 6, 2025. It adds sections 22775 to 22777 to the Business and Professions Code (bill text). The operative rule, section 22776, reads: “On and after July 1, 2026, a video streaming service that serves consumers residing in the state shall not transmit the audio of commercial advertisements louder than the video content the advertisements accompany, consistent with the regulations adopted by the Federal Communications Commission pursuant to the Commercial Advertisement Loudness Mitigation (CALM) Act (Public Law 111-311) for television broadcast stations, cable operators, and other multichannel video programming distributors.”
Three parts of the text matter to advertisers:
- The definition is broad. A “video streaming service” is an entity that makes available directly to the consumer, over internet protocol, either video programming or “video content the entity makes available for users to view”. It excludes TV stations, cable operators and other MVPDs (already covered by the FCC rules) and any service that runs no commercial advertisements.
- The duty sits with the service. Like the federal rules, the statute regulates what the service transmits. It names no duty for advertisers or production companies, so its effect on a spot arrives through the service’s delivery spec.
- There is no private lawsuit. Section 22777 states that the chapter “does not create a private right of action”.
The governor’s announcement described the law as building on the 2010 CALM Act so that it “will also apply to streaming services”. Which apps fall inside the definition, and how the rule is enforced in practice, are questions for each service and for counsel.
What streaming publishers ask for
The newest ATSC edition gives streaming services a framework. Annex L of A/85:2026-07 recommends that a streaming service present all content “at only one specific and consistent Target Loudness” and that “selecting a Loudness value between -23 and -27 LKFS is recommended, unless there is a prior arrangement otherwise”. A footnote explains the range: some streaming operators use -23 LKFS (in line with the European EBU R128 s2 guideline), others -27 for more headroom, and others -24, the A/85 broadcast recommendation or “a default in the absence of any other agreed-upon value”. The annex also says ads should be measured on the full-program mix, that an inserted ad’s loudness should match “the delivery channel’s Loudness target value within +/-2 dB”, and that what is vital is “that the Loudness of Short-form Content not be greater than that of the Long-form Content it accompanies”.
Why programmatic insertion raises the stakes
Most streaming ads are inserted automatically, and that changes where loudness can be fixed. A/85:2026-07 defines programmatic advertising as the “automated downstream insertion of Short-form Content” and devotes a section of Annex L to loudness at the point of insertion: the inserted ad’s loudness should match the channel’s target within 2 dB either way. In a broadcast chain, a station operator can measure a file and correct it before air. In an automated break, the ad arrives from an ad server and is stitched into the stream with little or no human listening, so the file has to be right when it leaves post.
Codecs add a second reason. Annex L separates codecs that carry loudness metadata (AC-3, E-AC-3, AC-4 and others) from codecs that do not, such as AAC and MP3, which are the usual audio formats in the MP4 renditions served to streaming players. With metadata, a decoder can be told how loud the content is. Without it, the loudness is whatever is baked into the audio, which is why Annex L recommends a single consistent target for everything a service plays. For an advertiser, the practical reading is simple: the MP4 renditions inherit the loudness of the mix they were encoded from, so each rendition should be measured after encoding against the publisher’s target, not assumed to match the master.
Buying platforms tend to default to the broadcast number. The Trade Desk’s CTV best practices, for example, ask advertisers to “ensure that the loudness of audio averages -24 LKFS”. When a publisher’s own sheet names a different target, the sheet governs.
The arithmetic is simple and unforgiving. A -24 LKFS spot inserted into a service that runs at -27 is 3 dB above the service’s target, outside the 2 dB tolerance Annex L describes, and louder than the program it interrupts. The service will either turn it down or play it hot, and in California the second outcome is the one the statute forbids the service from transmitting. The fix belongs in the deliverables: a version at the publisher’s target.
Example (fictional). Tolvane, an invented kitchenware brand, books a 30-second and a 15-second spot on national cable and on three streaming publishers. The cable network’s sheet says -24 LKFS. The three streaming sheets (illustrative values, not any real publisher’s) say -24, -27 and -23 LKFS. The post team plans:
- Version A at -24 LKFS: cable, streaming publisher 1 (0 dB from target) and publisher 3 (1 dB under its -23 target, inside the 2 dB tolerance, which publisher 3 confirms in writing).
- Version B at -27 LKFS: publisher 2. It is made from version A by lowering the whole mix 3 dB, a gain change that also lowers true peak by 3 dB, from -2.5 to -5.5 dBTP, so no new limiting is needed.
Two loudness versions x 2 lengths = 4 mixes to measure and deliver, instead of one per publisher (4 outlets x 2 lengths = 8). Had publisher 3 refused a file 1 dB under its target, a third version at -23 would bring the count to 3 x 2 = 6. Every name and number in this box is an illustration.
Turning a mix down is a gain change. Turning one up is not always: raising a -24 mix to -21 also raises its true peaks by 3 dB, and a mix peaking at -2.5 dBTP would then peak at +0.5 dBTP, over full scale. Louder versions need limiting and a fresh QC pass, so they cost more time than quieter ones.
What about social and online video?
No US loudness law applies to social or online video ads, and the platform spec pages we checked set no loudness target as of November 2026. YouTube’s recommended upload settings list audio codecs (AAC-LC, Opus or Eclipsa Audio), channels and a 48 kHz sample rate, with no loudness figure. TikTok’s In-Feed ads specifications (updated June 2026) and LinkedIn’s video ads specifications list formats, sizes, durations and bitrates, with no loudness figure. Meta’s Facebook Feed video specifications recommend “stereo AAC audio compression at 128 kbps+” and list sound and captions as “Optional, but recommended”, again with no loudness target.
California’s statute could matter here, because its definition covers entities that make “video content” available “for users to view” and run commercial advertisements. Whether a given social or video-sharing platform is a “video streaming service” under section 22775 is for that platform and its counsel; the practical effect on an advertiser would again arrive through the platform’s own rules.
Without a published target, the plan has to set one. Two choices are defensible:
- Deliver the broadcast mix. A -24 LKFS file with true peaks below -2 dBTP is the A/85 default “in the absence of any other agreed-upon value”, it sits inside the streaming range Annex L describes, and it avoids a separate mix.
- Make a louder online mix. A brand may want more level for phone speakers. That is a creative choice, not a platform requirement, and it needs limiting and its own QC. If anyone cites a platform loudness requirement, ask for the platform’s page.
Whichever you choose, write the number into the plan so every file for that destination is measured against it. Placements that often play without sound need captions more than a loudness decision; whether captions are required and how to deliver them is covered in captions and audio description for video ads.
Which loudness versions should go in the deliverable matrix?
Order one version per distinct loudness target, not one per outlet, and make quieter versions from louder ones by gain alone. The table sets out the default for each outlet.
| Outlet | Governing document | Target | Measured how | Version to order |
|---|---|---|---|---|
| Broadcast TV and cable | FCC rules (A/85:2013) and the network's sheet | -24 LKFS, about ±2 dB; true peak below -2 dBTP | Full mix, integrated, BS.1770; slate excluded | Broadcast mix; 5.1 only if the network asks, with the downmix checked |
| Streaming or CTV publisher at -24 | Publisher sheet | -24 LKFS | Full mix, integrated | The broadcast mix |
| Streaming publisher at another target | Publisher sheet (A/85:2026 Annex L suggests -23 to -27) | As stated, matched within ±2 dB | Full mix, integrated | A version at that target, unless an existing one is accepted in writing |
| Streaming services serving California | Bus. and Prof. Code 22776 (binds the service) | Not louder than the content | Consistent with the FCC's CALM rules | The publisher's target version |
| Social and online video | Platform specs (no target published) | Set in the production plan | Full mix, integrated | The broadcast mix, or a plan-set online mix with its own QC |
| Language, described or edited versions | Same as the outlet | Same as the outlet | Each file measured on its own | A separate measurement for every mix |
The decision rules behind the table:
- List every outlet and its target. Get each streaming publisher’s number in writing, with the date, before the mix is booked.
- Start from the broadcast mix at -24 LKFS if any TV, cable or CTV placement is on the plan.
- Group targets. A publisher whose target is the same as an existing version, or which confirms in writing that an existing version is acceptable, shares that version.
- Derive quieter versions by gain. Lowering a mix lowers its true peak by the same amount.
- Plan louder versions as new work. Raising a mix needs limiting and a new QC pass.
- Measure every edit separately. A 15-second or 6-second cutdown has its own integrated loudness; trimming a compliant 30 does not make a compliant 15. How those shorter edits are structured is covered in cutting down a 30-second spot.
- Keep stems. Separate dialog, music and effects stems let a version be rebuilt for a new target or language without reopening the whole mix.
Multiply the result through the matrix: loudness versions x lengths x languages gives the number of mixes to measure. In the fictional Orlenna plan used across this site, every placement is online (paid social, online video, the product page and a trade show screen), so no loudness law applies; each of its 24 files is measured against the target the plan sets for its destination. A plan with a CTV flight, like the Tolvane example, adds versions only where publisher targets differ. Our article on CTV ad specs covers the file side of the same deliverables: masters, renditions and VAST tags.
What to ask for in the QC report
Ask for a loudness report for every delivered file, not one for the master. A useful report states:
- File name and length, matching the deliverable matrix.
- Target and tolerance from the governing sheet, with the sheet’s date.
- Integrated loudness in LKFS, measured on the full mix over the content only (slate and black excluded).
- Measurement method: full-program mix, and the BS.1770 version the meter uses.
- Maximum true peak in dBTP, measured before encoding.
- Channel layout, and for 5.1 files the downmix loudness as well.
- Dialnorm value, if the file or the destination uses AC-3 metadata.
- Meter or software and its version, and the date of measurement.
- Result: pass or fail against each destination that receives the file.
Reports like this do two jobs. They catch a hot or quiet file before a network or service corrects it, and they give the brand a record if a station, a streaming service or a viewer later questions a spot’s level. They do not certify legal compliance; the transmitter’s obligations remain its own, and counsel decides what the brand needs to keep.
The delivery specs checklist includes loudness and true peak among its checks for every destination, and the deliverables page shows where loudness sits in a handover.
What to do next
Before the mix is booked, list every outlet in the media plan, mark which are broadcast, cable, streaming or social, and request each streaming publisher’s loudness target in writing. Decide the online target in the plan rather than in the mix session. Then order one mix per distinct target, measure every length and language separately, and ask for the per-file report above. If you are planning a campaign that runs on TV, streaming and social at once, you can send a brief with the placement list, and the production plan will show each loudness version as a line in the deliverable matrix.
