Audio and speech

Dialogue, TTS, voice, cleanup, separation, translation and localization

• articleWhat is the EBU SQAM test material, what does it contain, and what use does the EBU allow?
  The EBU SQAM CD (handbook Tech 3253, September 2008) is a set of recordings for subjective sound-quality tests, grouped from alignment and artificial signals through single instruments, voice, speech, orchestra and pop (70 tracks). The handbook states no licence. The EBU's generic test-sequence terms forbid commercial use and redistribution, and the producer's reading of the SQAM download page says downloading means agreeing to R&D-only use.
• articleCan I build the EBU Tech 3341/3342 loudness test signals myself, or must I use the EBU download, and what are the download terms?
  EBU's loudness-meter compliance tests (Tech 3341, 23 cases; Tech 3342, 6 Loudness Range cases; both November 2023) mostly use synthetic 48 kHz tones; a few cases use authentic programme segments (e.g. a promo-like and a drama-like excerpt). The test sequences may be used only to assess audio equipment in internal R&D. Commercial use, copying, modifying, publishing and redistribution are all forbidden (EBU terms v1.0, July 2019), so they can't be bundled into a product or public test suite.
• articleWhat do Ofcom's access-services guidelines say about audio description style and subtitle presentation (speed, layout, synchronisation)?
  The 2015 Ofcom Code's Annex 4 guidance asks audio description to be in the present tense, to avoid filmic terms like camera angles and the phrase 'we see', and not to cover dialogue or key sounds. It asks pre-recorded subtitles to run normally no faster than 160-180 words per minute, usually in up to two lines, synchronised with speech. The current Code (last updated 15 April 2024) no longer contains this annex and refers to separate Ofcom best-practice guidelines, so cite the 2015 text as a dated version and check the current guidelines.
• articleWhat did the ITC's 2000 guidance say about how to write and record an audio description (tense, what to describe, sound levels)?
  The ITC's May 2000 audio description guidance is a describer's manual: seven steps from choosing and viewing the programme through drafting, reviewing, setting the programme sound level, recording and checking. It asks for present-tense description (continuous for ongoing action), not talking over sound effects, and lowering the programme level under description. It is historical guidance from the start of UK digital TV, not current Ofcom policy.
• articleWhat do Video.js v10.1.0 docs and v8.24.2 code do with caption, subtitle and description tracks, and what gaps do the maintainers state?
  Video.js changed approach between versions. Video.js 10 renders captions natively so FCC-required OS caption preferences apply (decision dated 2026-02-26), offers menus only for captions and subtitles, has no caption-appearance menu, and points audio description to a described audio track. Video.js 8.24.2 had a Descriptions button and showed description cues in its own display with aria-live assertive, giving captions and subtitles priority when both were on. Moving from v8 to v10 drops the built-in text-description path.
• articleHow does Able Player (v5.0.0) deliver text-based and separate-video audio description, and what are its documented gaps?
  Able Player 5.0.0 (MIT) offers two audio-description routes: a WebVTT descriptions track voiced through the browser's Web Speech API (or written to an ARIA live region), or a separate described-video file (data-desc-src). It can pause the video while a description is read, and puts descriptions in its transcript. Open issues report description problems on iOS and Safari.
• articleWhat do the HTML <track> kind values and TextTrack modes mean for description tracks, and is there evidence that browsers voice kind="descriptions" themselves?
  The HTML Standard defines a 'descriptions' text track as text intended to be synthesised as audio when the picture can't be seen, and makes voicing it a browser (user agent) job: a showing descriptions track is made available non-visually. Evidence that browsers actually voice such tracks is thin, so web players implement it themselves.
• articleHow does the CRTC's NER-based live-caption accuracy standard work (score formula, error types and deductions, sample size, evaluator certification, monitoring and remediation)?
  Canada's CRTC (Broadcasting Regulatory Policy 2019-308, August 2019) replaced the 95% verbatim test for English live captions with an NER accuracy score of at least 98. The score is (captioned words minus NER error deductions) / captioned words x 100, with deductions of 0.25 for benign errors and omitted detail, 0.5 for omitted main meaning and 1.0 for false information. Broadcasters monitor two live programs a month and report annually by 30 November. The Commission may ask for monthly calculations if compliance issues arise.
• articleWhich EN 301 549 V4.1.1 clauses does Annex ZB map to the EAA's AV, access-service and e-reader rows, what is outside its scope (e-books), and what is the presumption-of-conformity status at the pinned sources?
  EN 301 549 V4.1.1 Annex ZB maps European Accessibility Act requirements to its clauses. For access services in audiovisual media (Table ZB.5 row 5(b)(ii): subtitles, audio description, spoken subtitles and sign language transmitted fully, synchronised and user-controllable) it lists clauses 4.2.1, 4.2.2, 4.2.4, 4.2.5, 5.1.3.7, 5.1.5, 5.4, 5.5, 5.10, 7, 9.1.2, 10.1.2, 11.1.2 and all control-operation requirements. E-books are out of its scope. Conformity is presumed only for directives under which the standard is cited in the Official Journal; no EAA citation was found in the sources read.
• articleWhat does EN 301 549 clause 7 (ICT with video capabilities) require of a media player, and what changed between V3.2.1 (2021-03) and V4.1.1 (2026-09)?
  EN 301 549 V4.1.1 (September 2026) clause 7 requires video players with synchronised audio to offer a subtitle mode and to show subtitles within 100 ms of their timecode. They must keep subtitle data when transmitting, converting or recording, let users adapt subtitle presentation where the player controls it, and offer spoken subtitles and selectable, synchronised audio description. Where the player controls activation, it needs a single-operation toggle for subtitles, audio description or spoken subtitles at the same level as the volume control. V3.2.1 (2021) said 'captions' and required these controls at the same level of interaction as the primary media controls.
• articleFrom which date must a provider of an AI audio system mark its outputs under AI Act Article 50(2), what does Article 113 now say after the Digital Omnibus, and where is the official text?
  Under the EU AI Act as amended by the Digital Omnibus on AI (Regulation (EU) 2026/1744, OJ 24 July 2026), the Act applies from 2 August 2026, and a new Article 111(4) gives existing systems a grace period. Providers of systems generating synthetic audio, image, video or text that were on the market before 2 August 2026 must comply with the Article 50(2) machine-readable marking duty by 2 December 2026. Article 50 paragraphs 1-6 are unchanged; paragraph 7 (codes of practice) was replaced. How the dates apply to a given system is a legal question this record does not answer.
• articleWhich audio, video and image formats can c2pa-rs 0.91.4 and c2patool 0.28.2 embed and read manifests for, and which of C2PA 2.4 Appendix A's embeddings have no handler at that pin?
  The C2PA Rust SDK (v0.91.4) can embed manifests in WAV, AVI and WebP (RIFF), MP3 and FLAC (ID3-style), and MP4/M4A/MOV (BMFF). Fragmented MP4 (DASH) is supported only for file-based operations. It has no Ogg, font or HTML handler, although C2PA 2.4 Appendix A defines embeddings for Ogg Vorbis (added in 2.3), fonts, HTML, ZIP and text. A ZIP handler exists in code but isn't in the docs table. Ogg Vorbis/Opus support is an open experimental pull request (#2073), not merged as of 2026-10-22. c2patool supports whatever library version it was built with.
• articleHow can an agent read the C2PA Trust List and Conforming Products List as data to see which vendors, media types and spec versions are covered, and what do those lists not tell it?
  C2PA's Conforming Products List (270 records: 240 generators, 30 validators) records which media types each product can sign, not individual signing events. By curator recount, 81 generators actually list audio container types (most often WAV 63, MPEG 57, MP4 51), 151 list video and 208 list image; 33 carry an mlModel key, 32 of them empty.
• articleExactly which language rows were added, removed or changed between Doc 7's table (version B, 2018-2021) and the registry table today?
  The ISDCF register (330 rows) is far larger than Doc 7's 62 entries. 270 rows have no Doc 7 counterpart, mostly region and script variants plus new bare languages. Only KG was dropped, and 91 register codes are four letters long (e.g. ARSA, AZCY).
• articleI found an old ISDCF Doc 7 language table; can I still use its DCNC language codes, or has the registry changed them?
  Codes from the old Doc 7 table (version B) mostly still hold. In the current ISDCF register, 58 of its 61 coded languages keep both DCNC code and RFC 5646 tag. Three changed: Kyrgyz is KY (Doc 7 said KG), Flemish VLS now maps to nl-BE (was vls), and Malay MSA maps to ms (was zlm). Look codes up in the register, not the old PDF.
• articleHow is word error rate defined, what does the published critique say about it, and how much does the choice of text normaliser (for example Whisper's) change what is counted?
  WER counts substitutions, deletions and insertions against the reference word count, so it can exceed 1 (upper bound max(N1,N2)/N1). Morris et al. (2004) argue that measures of information communicated suit most applications better. WER also depends on text normalisation. Whisper's English and basic normalisers delete bracketed and parenthesised text (so tags like [music] or (laughs) vanish) and standardise numbers and spellings before scoring. WhisperX reports insertion error rate and 5-gram duplicates to catch hallucinated repetition. WER says nothing about timing, speaker labels or punctuation.
• articleWhat do peer-reviewed critiques say about how TTS MOS tests are reported and how test design changes the scores?
  A bare TTS score like 'MOS 4.2' is underspecified. Kirkland et al. (SSW 2023) found most 2021-22 Interspeech/SSW papers omit the scale labels, increments and listener instructions. In their own test, asking about 'quality' rather than 'naturalness' raised MOS (3.52 vs 3.21). The VoiceMOS 2022 organisers state MOS values can't be compared across listening tests, because each test has its own listeners and systems. ITU-T P.800 also warns that results on the older 0-4 scale must be raised by one to compare. Report the question, scale, listener pool and stimuli with any MOS.
• articleWhat room-acoustic and listening-level numbers do EBU Tech 3276 and ITU-R BS.1116-3 state for a reference listening room, and what are they scoped for?
  EBU Tech 3276 (2nd ed., 1998) and ITU-R BS.1116-3 (2015) set the same room targets. Average reverberation time is Tm = 0.25 x (V / 100 m3)^(1/3) seconds (BS.1116-3 measures it over 200 Hz-4 kHz). Early reflections in the first 15 ms should be at least 10 dB below the direct sound between 1 and 8 kHz. Background noise should preferably stay at NR 10 and never exceed NR 15, with -18 dBFS alignment. They differ on listening level: Tech 3276 uses 85 - 10 log(n) dB(A) for n channels, BS.1116-3 uses 78 +/- 0.25 dBA. Both are for assessment and reference listening rooms, not mix-room certification.
• articleWhat do the REW help pages say about what room measurements and EQ can and cannot fix, and what RT60 target they give for a small mixing room?
  REW's help says room EQ is valid only at the measured position. EQ can tame response peaks, especially low ones, but not deep narrow dips, which are position-sensitive and need headroom EQ doesn't have. A 1/3-octave RTA is too coarse to see those dips. In domestic-size rooms RT60 isn't meaningful below a few hundred Hz (modal behaviour, no diffuse field). Recommended RT60 is about 0.3 s under 50 m3 and 0.4-0.6 s up to 200 m3. C50 indicates speech clarity and C80 music clarity. REW's room simulator treats every source as omnidirectional.
• articleWhat loudness (LKFS) and true-peak limits do the ABC Australia, Free TV Australia and CBC/Radio-Canada documents state, and which BS.1770 revision do they cite?
  Australian and Canadian broadcasters use -24 LKFS and -2 dBTP, not the UK's -23 LUFS and -1 dBTP fail line. ABC (V4.7, 2015) sets -24 LKFS average dialogue level (per Free TV OP48) and -2 dBTP maximum. Free TV OP-59 Issue 4 (2018) allows +/-1 LU and cites BS.1770-3. CBC/Radio-Canada v8.0 (January 2026) requires -24 LKFS +/-2 LU for both programme and dialogue loudness, -2 dBTP, measured with a BS.1770-5 algorithm. CBC v7.0 (2022) also sets -10 dBTP for audio-only content such as podcasts.
• articleWhat programme loudness target, tolerance and true-peak limit do the UK DPP-family delivery documents (BBC, Channel 4, ITV, Sky) state, and where do the documents differ?
  UK DPP-family delivery specs all target -23.0 LUFS, recommend -3 dBTP and fail QC above -1 dBTP, but the tolerance depends on which document you read. Channel 4 v5.2 and ITV live v6.3 (2023) allow +/-0.5 LU non-live and +/-1.0 LU live, while the Sky DPP-derived PDF (UK 4.3 2014 / Broadcaster 4.8 2017) allows +/-1.0 and +/-2.0 LU. Loudness range (18 LU) and factual dialogue range (6 LU) are guidance only. The BBC's 2026 file spec gives only a -23 dB dialogue-level setting and refers to DPP010 for QC. NRK v03 allows +/-1.0 LU for live.
• articleWhich W3C timed-text specs (TTML2, IMSC 1.0.1/1.1/1.2/1.3, WebVTT) exist, at what status, and which profile does a major deliverable spec ask for?
  As of 2026-10-22 there are separate dated IMSC Recommendations: 1.0.1, 1.1, 1.2 (August 2020) and 1.3 (21 May 2026). IMSC 1.3 is text-only and drops the image profile. An EBU-TT-D file is generally also a valid IMSC 1.2 Text Profile file, with listed exceptions. WebVTT is still a Candidate Recommendation Draft (20 May 2026), not a Recommendation. Platform profiles are narrower: Netflix accepts its restricted IMSC 1.1 Text Profile only for Japanese timed text, so W3C conformance alone is not delivery conformance.
• articleWhat do XAG 103 and GAG require for redundant visual and audio cues?
  XAG 103 asks that any visual or audio content critical to gameplay or story also reach players through at least one other sense. Its examples: gunfire shown visually also comes through spatial audio or haptics, and important sounds get an on-screen indicator of type and direction. Text-only dialogue should name the speaker and show where they are. The Game Accessibility Guidelines rate 'no essential information by sound alone' as Hearing, Basic, testable by a first-time playthrough with sound muted, and rate replicating supplementary audio information (such as shot direction) visually as Hearing, Intermediate.
• articleWhat objective methods (STOI/ESTOI, SRMR, speech-loudness measures, SI4B monitoring) are stated for judging dialogue intelligibility in a finished mix, and what limits do the sources give?
  STOI/ESTOI (pystoi) are intrusive: they compare a clean reference with processed speech, at 10 kHz over 15 third-octave bands from 150 Hz, in 30-frame segments, after dropping frames more than 40 dB below the loudest. Becerra et al. (2024) call intrusive metrics unsuitable for finished film dialogue because no reference exists, and advise caution with the non-intrusive SRMR. Torcoli et al. (2024) propose loudness-based checks instead: speech loudness deviation (critical below -10 LU) and speech-to-background loudness difference (critical below 0 LU), as a first proposal. On their data, the 0 LU threshold correlated best with self-reported listening effort.
• articleWhat do the Commission's Article 50 guidelines (20 July 2026) and the code's deployer section say about synthetic voice, audio deep fakes, music and audio-only labelling?
  For audio-only deep fakes, Section 2 of the EU code (for deployers) asks for a short audible disclaimer at the start in plain language (or an earcon with awareness measures), reminders at intervals for long-form or live audio and after interruptions such as ads, and a visual label when a screen is available. The Commission guidelines' examples count voice cloning of real podcast presenters as a deep fake. They do not count fictional-character voices (with no deception about the narrator) or loudness/noise processing that leaves the words unchanged. Speech in a specific real person's voice needs marking, assistive voices (AAC, customised neural voices) do not, music in an existing artist's style is a creative-work deep fake, and playlists are out of scope. The examples are illustrations, not determinations.
• articleWhat do Amazon's public Prime Video Tech Docs say about audio delivery: loudness, stems, channel layouts, sample rates?
  Prime Video's public Tech Docs, as archived, ask for fully mixed program tracks only (M&E and MOS removed), 48 kHz audio in the stated L-R-C-LFE channel mapping, and descriptive audio typed as Narration and paired with same-locale dialog. They list validation failures such as wrong channel counts and audio-video duration mismatches over 2.0 s. Neither archived page states a loudness target or true-peak limit, so a loudness number for Prime Video cannot be cited from these public pages.
• articleHow are hearing-impaired (HI) and visually-impaired narrative (VI-N) audio tracks represented in a DCP, and how do open tools model and check them?
  In a DCP, hearing-impaired (HI) and visually-impaired narrative (VIN) audio are dedicated channels inside the main sound track, not separate files. SMPTE ST 428-12 defines HI (0Eh) as a dialog-intelligibility channel and VIN (0Fh) as narration describing picture events, and allows several of each for different languages. libdcp maps the MCA ids 'hi' and 'vin'/'vi-n' to its own channel slots, but its verifier has no HI/VIN-specific check. ClairMeta checks only an even channel count. No inspected tool verifies that a VIN channel actually carries narration or is labelled correctly.
• articleWhat does Netflix's 'Dolby Atmos Home Mix Archive Specs' say about archive files, naming and metadata for an Atmos home mix?
  The Atmos requirements page specifies the archive print master as a Dolby Atmos BWAV ADM file (any bed configuration), with DME separation and group labelling deferred to Netflix's BWAV ADM Creation Guidelines. It gives no Atmos-specific filename pattern or ADM metadata table, so naming must come from the asset-naming page and those guidelines.
• articleWhat do Unity Localization Smart Strings and Godot gettext/tr_n docs state about plurals, genders and variables in dialogue strings, and what failure cases do they mention?
  Unity Localization's Smart Strings (1.5.13) give dialogue lines pluralisation, gender, lists and conditionals with named placeholders; the documented trap is that sources are tried in order, so the same placeholder can resolve to different data, and number formatting follows the table's locale. Godot's tr_n plural support is for zero or positive integers only; its gettext extractor silently skips non-constant string arguments, and NO_TRANSLATE/TRANSLATORS comments steer extraction; pseudolocalisation catches length and hard-coded text but is not a CJK/RTL font test.
• articleWhich of Pro Tools, REAPER and Resolve document keeping the bext TimeReference when a file is rendered, trimmed or sample-rate converted, and what does REAPER's history actually promise?
  REAPER's version history documents preserving the source start offset on render (v7.16, adjusted for take offset), a batch-converter choice between keeping the BWF start or using item position (v5.985), clearing offset data on consolidate/glue with source metadata (v7.35), and adjusting the written position after trimming silence (v7.37). No entry or API text says how a stored sample-count start is rescaled when the sample rate changes (no pull-up/pulldown terms). Pro Tools and Resolve remain unresolved here.
• articleHow do I get a field-recorder WAV to land at its embedded timecode in REAPER, and what do Avid Media Composer and Final Cut Pro do differently with the file's timecode frame-rate flag?
  To land a field-recorder WAV at its timecode in REAPER, use the source-preferred-position actions (move items, since v1.01; Media Explorer insert, v6.17) or InsertMedia flag &4096; since v6.19 REAPER also reads time references in mp3/flac/ogg/opus/wavpack. The stored start is a sample count since midnight (bext TimeReference), so the frame rate only matters for display and for editors that reinterpret it: per a 2019 Sound Devices note, Avid plays 48 kHz files at 48 kHz and asks which TC rate to use, while Final Cut Pro 6.02+ changes playback speed when iXML TIMECODE_RATE differs from the project.
• articleWhat do LUFS, LKFS, LU, dBFS and dBTP mean, and how do they relate?
  LKFS (ITU-R BS.1770) and LUFS (EBU R 128) are the same absolute unit, K-weighted loudness relative to full scale, and track level dB for dB. LU is the relative form (1 LU = 1 dB), used for differences such as distance from target or loudness range. dBFS is level relative to digital full scale. dBTP is a true-peak level estimated by oversampling (at least 192 kHz) to catch peaks between samples; R 128 caps production true peak at -1 dBTP. Loudness describes average perceived level, true peak describes clipping risk.
• articleHow do Opus R128_TRACK_GAIN tags, Vorbis-comment ReplayGain tags and the output gain header differ in reference level, and which players document or implement applying each?
  RFC 7845 makes the Opus ID-header output gain the default gain. The optional R128_TRACK_GAIN/R128_ALBUM_GAIN tags are Q7.8 dB values relative to the EBU R128 reference, applied on top of the output gain, and must be updated or removed if the output gain changes; REPLAYGAIN_* tags should not appear in Opus comments. Players map R128 tags to ReplayGain's reference by adding 5 dB (mpv: ReplayGain 2 at -18 LUFS; MPD: ReplayGain 89 dB). mpv applies no replay gain unless asked, and album mode falls back to track gain. In the FFmpeg files read, only REPLAYGAIN_* tags are exported as side data and the Opus header gain read is commented out of the demuxer.
• articleWhich public HRTF/SOFA-convertible measurement sets state terms an agent can check (MIT KEMAR, Viking), and where is the SOFA catalogue?
  MIT's KEMAR HRTF data (1994) are free to use without restriction provided the authors are cited (710 positions, 44.1 kHz); the Viking HRTF dataset v2 on Zenodo (2020) is licensed CC BY 4.0 in its record metadata and distributed as SOFA files plus 3D scans. Terms for third-party SOFA conversions of KEMAR were not readable (SOFA hosts refused connections to producer).
• articleWhat do practitioner and delivery documents say about ADR session setup, cueing, mic matching and what a dialogue editor expects back?
  One ADR recordist/editor describes a fairly standard dialogue-editor delivery: 24-bit 48 kHz stereo tracks with boom on the left and lavalier on the right, selected takes at the top and everything else removed, plus fill tracks, cueing and playback reverb/EQ to sell the match. Practitioner testimony, not a specification.
• articleOverview: from the set recorder to the delivered package, which carrier holds which kind of information at each audio handoff, and where does the timing or metadata thin out?
  Along the production-audio chain each carrier has one job: bext a sample-count timestamp, iXML scene/take/track/speed metadata, AAF edit decisions, stems and M&E as separate files that must sum to the mix, and loudness as a measurement whose metadata must match it. Foley handoff had no primary document in the producer's sources (gap).
• articleWhich Ofcom Broadcasting Code rules apply to material that could mislead viewers (factual programmes, simulated news) and to dramatised reconstructions of real events?
  In the January 2019 Ofcom Broadcasting Code, Section Two requires that factual programmes and portrayals of factual matters not materially mislead (2.2) and that simulated news in drama or documentary cannot reasonably be mistaken for real news (2.10); rules naming reconstructions sit in the fairness and privacy sections (7.9, 7.10, 8.19), not in Section Two. Later amendments are not covered.
• articleWhat audio deliverables, loudness, true-peak and stem/M&E rules does Netflix's Post Production Branded Delivery Specification state (2026-06-06 observation)?
  Netflix branded delivery v5.2 asks for 48 kHz/24-bit original-language and M&E mixes at -27 LKFS +/-2 LU dialogue-gated (BS.1770-1), true peak no higher than -2 dBFS for 5.1/2.0 and recommended limiters at -2.3 or lower on Atmos beds and objects, plus 5.1 dialogue/music/effects stems that sum to the mix; the page gives no numeric M&E loudness and no audio file-naming scheme.
• articleWhat does Ogg Opus (RFC 7845) preserve about sample rate, start offset, loudness and channel order compared with WAV or FLAC?
  In Ogg Opus (RFC 7845) the stored input sample rate is informational only (decode at 48 kHz), playback position subtracts pre-skip, the header carries an output gain, mapping family 1 uses Vorbis channel order that differs from WAVE and FLAC (permute channels), and loudness normalisation uses R128_TRACK_GAIN/R128_ALBUM_GAIN tags rather than REPLAYGAIN tags.
• articleWhy can one spoken-word file meet ACX, Apple Podcasts and AES TD1008 numbers differently, and what must I record when I claim a file 'meets spec'?
  Spoken-word targets use incompatible bases: ACX (audiobooks) sets an RMS window of -23 to -18 dB, -3 dB peaks and a -60 dB noise floor; Apple Podcasts measures LKFS and true peak per BS.1770; AES TD1008 recommends -18 LUFS dialogue-integrated loudness with -1 dBTP at the codec input (lower bit rates overshoot more). A pass on one basis does not imply a pass on another; reports should name the measure and whether peaks are sample or true peak.
• articleWhich of the bext chunk and the iXML chunk should I trust for timecode, speed and take identity when a recorder file reaches the dialogue editor?
  Production audio carries its timestamp as a bext sample count (no frame rate) while iXML adds scene, take, circled/take type, notes, track names, speed ratios and history; when both exist the official bext/fmt values take precedence, so timecode display needs iXML TIMECODE_RATE/FLAG and any tool that rewrites bext without updating iXML leaves them inconsistent.
• articleWhat channel layouts, stem separations and sidecar file rules does Paramount state for programme audio (including M&E and LFE)?
  Paramount sidecar audio must be 1, 2 or 6 interleaved channels in RF64 WAV with channel labels matching the layout, start together and match the video duration within one frame, and every 5.1 needs a matching stereo; LFE may hold only content under 120 Hz (full-range LFE fails QC); Atmos beds and objects must at least separate dialogue, narration, music and effects.
• articleWhat programme-loudness and true-peak limits do Paramount and Apple state for sidecar, Atmos and immersive-music deliveries?
  Paramount requires near-field full mixes at -24 LKFS +/-2 dB (BS.1770-3, without LFE for 5.1) with true peak at or below -2 dBTP for US deliveries, or EBU R 128-2020 -23.0 LUFS +/-1 LU with -1 dBTP for Europe; theatrical mixes are exempt; the Atmos home mix targets -24.0 LKFS +/-2.0 LU with true peak at or below -0.1 dBTP, measured in the Dolby Atmos Renderer or an equivalent.
• articleHow does a documented DAW (Ardour) treat embedded BWF timestamps, sample-rate mismatches and film/video pull-up when importing?
  Ardour's import dialog shows a Broadcast WAVE file's embedded timestamp, can place the file at that timestamp (or zero if none), and marks a sample rate that differs from the session in red because the file must be resampled; the page does not mention iXML.
• articleHow is SMPTE LTC structured per libltc documentation and code, and where do 23.976, 29.97 and 59.94 behave differently?
  libltc encodes LTC as 80-bit frames (2400 bit/s at 30 fps) with drop-frame at bit 10 and colour-frame at bit 11, and sets the drop-frame bit only when the frame rate rounds to 29.97 (rint(fps*100)==2997, in src/ltc.c); so 23.976 material is labelled as non-drop 24-frame counts (code reading).
• articleHow do iXML SPEED tags express NTSC frame rates and pulled-up recordings (for example 48048 Hz labelled as 48000), and which field should a conform tool trust when they disagree?
  iXML SPEED gives rates as integer ratios (e.g. 30000/1001 rather than 29.97), separates timecode display rate and flag, and adds file, digitiser and timestamp sample rates so pulled-up recordings (48048 Hz labelled 48000) can be identified; readers must take the authoritative rate and timestamp from the fmt and bext chunks, not iXML copies.
• articleWhat does the BWF bext TimeReference field actually store, and what does FFmpeg's WAV code do with it when I copy, re-encode or resample a recorder file?
  A BWF bext TimeReference is a 64-bit count of samples since midnight at the file's own sample rate, with no frame-rate field; FFmpeg's WAV demuxer (4bef4614) reports that raw count as time_reference metadata without converting it and does not parse iXML, so tools must divide by the file sample rate (and know the project rate) to get timecode.
• articleWhich language-code system applies where: DCNC title-string codes versus CPL metadata tags, and what extra rules does ISDCF Doc 7 add?
  Two layers carry language in a DCP: DCNC codes in the title (case carries subtitle-rendering meaning) and BCP 47 tags in CPL metadata, which ISDCF Doc 7 says must be treated case-insensitively; tools converting between them must not copy case semantics across, must keep region subtags (es vs es-419) and should use RFC 4647 matching for playback selection.
• articleHow do I read the audio-language and subtitle-language parts of an ISDCF digital cinema package name, and where do the same values sit in the CPL?
  In ISDCF digital cinema naming the language field lists the audio language first and subtitle languages after it; upper case means the projector renders the subtitles and lower case means they are burned in; the same languages are mirrored in the CPL (MainSound and MainSubtitle Language, MainSubtitleLanguageList).
• articleWhat do WPT results for the CSS features that WebVTT and IMSC rendering rely on (ruby, writing modes, bidi, emphasis) show per engine?
  The CSS layout tests beneath subtitle styling tell a different story from the WebVTT cue reftests: css-writing-modes bidi tests pass 88 of 94 in all four engines, while css-ruby passes 114, 114, 111 and 64 of 179 and text-emphasis tests 77, 90, 125 and 20 of 145 (Chrome, Edge, Firefox, Safari). So engine differences in WebVTT bidi are not explained by generic CSS bidi support; whether ruby results explain the cue failures is not established.

More