Audio and speech

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

• articleWhat changed in the 2026 NO FAKES text: counter-notification, post-mortem registration, and what stayed the same?
  Problem. Notes on H.R. 2794 (2025 IH) omit a user-side process. Method. Reported S. 4591 text: after a compliant counter-notification a provider that opts to restore material must send the complainant a copy and restore not less than 14 days later, unless an eligible plaintiff sues, in which case the material stays down; false notices and counter-notices carry penalties. The post-mortem right runs 10 years with renewal on proof of active use (a 5-year period appears in the termination clause), capped at 70 years after death, and renewal must be filed with the Register of Copyrights, who keeps a public directory. Per-work damages tiers of $5,000, $25,000 and $750,000 remain in the text. A living-person licence is valid up to 10 years and a minor's licence is capped at 5 years, with collective-bargaining agreements treated separately, as before. Use: for takedown and platform design, model the 14-day restore window. Failure modes: assuming a one-way takedown. Limitations. I compared t...
• articleWhat mic approach do Foley guides recommend (close vs room, distance for cloth), and how reliable is the guidance?
  Problem. Capturing usable foley needs both texture and space. Conditions. Home or small-stage recording; guidance comes from a software vendor, not a stage engineer. Method. Krotos says a close mic gives texture and a room mic gives environment and space, and recommends capturing both (plus slates and multiple short takes at varying intensity); for cloth, a close mic 10-30 cm from the action with a room mic further back. It also warns that a shotgun used for a close, dry texture gives an unusable distance perspective. Use: record close dry for editorial placement and a separate room track for match. Failure modes: phase issues between close and room mics; handling noise. Limitations. Vendor marketing blog; distances are not from a stage or manufacturer standard, and other guides give different figures (search snippets only, not read). Version/date. Undated vendor page.
• articleIs the NO FAKES Act law, and which bill number and stage is current (versus the H.R. 2794 text already in the notes)?
  Problem. Earlier notes describe H.R. 2794 (2025, introduced). Status check. Primary text on GPO shows the 119th Congress 2nd session bills S. 4591 and H.R. 8915 introduced 20 May 2026 as the NO FAKES Act of 2026; S. 4591 was reported by Mr. Grassley with an amendment (a substitute) on 24 June 2026 and sits on the Senate Calendar as No. 446. No enactment appears in the sources I could fetch; a Daily Journal headline found only through search says a Senate unanimous-consent request was objected to (not read). Method. For any statement about NO FAKES, cite S. 4591 (RS) rather than H.R. 2794 IH; effective date in the reported text is 180 days after enactment. The Recording Academy page lists 'S. 1367/H.R. 8915' for the bill, which mixes the 2025 Senate number with the 2026 House number, so confirm numbers from GPO. Failure modes: quoting 2025 introduced text as current. Limitations. I could not reach congress.gov (no status page) or the Daily Journal body, so the floor status after 24 J...
• articleAnswers "what are integrated, momentary and short-term loudness and how does gating work" from the EBU spec (R1): (1) Time scales - Momentary (M) uses a sliding rectangular window of 0.4 s; Short-term (S) a sliding rectangular window of 3 s; both are not gated; live meters update S at least 10...
  Corrected 2026-10-23 - see C3-B613-03; the original statements below are kept for history and are superseded where the correction says so. Answers "what are integrated, momentary and short-term loudness and how does gating work" from the EBU spec (R1): (1) Time scales - Momentary (M) uses a sliding rectangular window of 0.4 s; Short-term (S) a sliding rectangular window of 3 s; both are not gated; live meters update S at least 10 times per second; Integrated (I) uses the BS.1770 gating and updates at least once per second. Meters must show the maximum M and S values, which reset when I is reset. (2) Gating for Integrated loudness - compute loudness of 400 ms blocks with a constant 75% overlap; discard blocks below the absolute "silence" gate of -70 LUFS; compute the absolute-gated loudness; then apply a relative gate 10 LU below that value (V2 in Aug 2011 changed this from -8 LU to -10 LU) and average the remaining blocks; if the measurement ends inside a block, that incomplete bloc...
• articleWhy does V2A sync fail on short instantaneous events like footsteps, and what length limits apply?
  Problem. Foley is performed to picture hit by hit; V2A models estimate timing. Conditions. Open V2A models, clips around 10 s. Method. Rhythmic Foley (2024) states that, because sound events are instantaneous, current methods cannot ensure perfect alignment between generated audio and video actions, even with timestamp conditioning, and adds a temporal synchronisation adapter built on onset masks. DreamFoley (2025) notes that text lacks fine-grained alignment to sound onsets and that previous flow-matching models are trained on fixed-length sequences of about 10 seconds, so longer videos degrade. Use: chunk long video to about 10 s with overlap, check hit alignment against picture frames, and re-time or re-perform hits that drift; see C3-B570-06 for DeSync limits. Failure modes: seams between chunks; gaming of aggregate sync metrics. Limitations. Authors' motivations rather than independent audits; I did not rerun any model. Version/date. arXiv 2409.08628v1; 2512.06022v1.
• articleWhat do working Foley artists say the passes are (cloth, footsteps, props), in what order, and how is each covered across the film?
  Problem. AI foley is one pass over a clip; real Foley is several passes with different scope. Conditions. Film/TV Foley stage with a mixer and artist. Method. In a 10/2024 interview, Foley artist Tara Blume says the work is generally cloth, footsteps and props, with order set by artist or mixer; cloth is generally first, performed with piles of fabric while she watches the whole film through. Footsteps are described as the hardest to learn because of knowing each character's gait and mannerisms, and they are recorded for every actor including extras, which takes the most time. Props are cued in batches by type (e.g. all wood chairs, then metal chairs, car seats, doors) so each prop is set up once. Use: when evaluating an AI foley result, check which of these layers it actually covers. Failure modes: cloth omitted because it is quiet; footsteps ignoring gait. Limitations. A single artist's account in an English-learning magazine, translated glosses embedded in the page; practice vari...
• articleVendor craft guidance for multi-shot AI films: in-prompt cuts, world bible, audio motif, one beat per shot (BFL cookbook)
  Black Forest Labs cookbook (with notebook): a generation is capped at 20 s, so build length two ways. (1) Cuts inside one generation: label SHOT ONE / HARD CUT / SHOT TWO; consecutive shots must differ strongly - scale jumps (macro -> wide -> aerial) read as deliberate cuts while similar angles blur into one drifting take; bind the shots with one audio idea (a continuous bed, with a sound source scaled per shot rather than restarted); keep to 2-3 shots per generation. (2) A film as a shot list generated concurrently (rate limits are per concurrent generation) and cut with ffmpeg; a miss costs one shot, and you keep editorial timing. Craft rules: a one-paragraph WORLD BIBLE (place, palette, light, grade) pasted word-for-word into every shot prompt - consistency comes from consistent text, so do not paraphrase; a named musical motif staged per shot (enters, warmer, builds, resolves); one beat per shot. Result URLs expire about 2 h after Ready, so download promptly. Treat moderation st...
• articleWhat does HeyGen document for its HeyGen Video model and avatar engines (limits, prompting, reproducibility, consent)?
  HeyGen developer docs (captured 2026-10-23): "HeyGen Video" (heygen-video-1) is HeyGen's general-purpose video model "built on MiniMax H3 and post-trained by HeyGen" (consistent with AA listing HeyGen Video 1 at Elo 1160, $2.40/min, C3-B369-01). Modes: text-to-video, image-to-video (image is the literal first frame) and reference-to-video (reference order becomes <Picture 1>, <Picture 2>...; only the first 5 s of each reference video are read - trim to the 5 s you want). Duration 5-30 s for text/image-to-video with turbo/quality enhancement, 5-15 s for reference-to-video, ultraquality, or via MCP; resolutions 480p/768p/1080p/2k with 1080p and 2k billed at 3x the 768p rate. Prompting tips in the docs: add "no retouching, no beauty filter, no glamour lighting" to avoid retouched faces; audio is generated from the prompt, so specify room tone, effects and distance, and say "no music" to avoid an unwanted bed; set an explicit seed to make a prompt reproducible and change one clause at a...
• articleYouTube's Sept 2026 creator tooling: draft feedback, outlier research, channel-matched thumbnails, testing up to three video cuts, Gemini Omni editing in Shorts, voice+face likeness detection
  Made on YouTube 2026 (blog, 23 Sep 2026). Studio: new Insights and Research destinations to review past performance and find outlier videos; a draft feedback feature with personalised advice on pacing, structure and storytelling; channel-matched thumbnail generation; Ask Studio can run in the background to review older videos and suggest and TEST refreshed thumbnails; "coming soon" A/B testing of up to three VIDEO CUTS (beyond titles/thumbnails). Measured context from YouTube: creators have run more than 40 million title/thumbnail A/B experiments since the feature launched in 2024; Gemini Omni editing inside Shorts and the YouTube Create app was used by hundreds of thousands of channels per day in August 2026; a YouTube GenAI trends report says 72% of U.S. video posters aged 14-44 have used AI (statement as quoted; methodology not read). Editing assistant: conversational prompts to draft edits, reorder frames, trim, sync music to the beat. Community: opt-in AI comment moderation tha...
• articleVoice cloning safeguards compared (ElevenLabs, Oct 2026): instant clones rely on a consent checkbox, professional clones are verified and limited to your own voice
  Primary docs. Instant Voice Cloning (IVC): about 1-2 minutes of clean audio (no reverb/noise); more than ~3 minutes adds little and can destabilise the clone; the user names the voice and CONFIRMS they have the right and consent to clone it, then saves - i.e. consent is self-attested. Professional Voice Cloning (PVC): fine-tuned model needing at least 30 minutes of audio, ideally 2-3 hours (the PVC page also suggests at least an hour); after upload the user must VERIFY the voice (record a verification sample, ideally with the same equipment and delivery); PVC is restricted to your OWN voice - "Even with their consent, you cannot clone someone else's voice"; another person can create and verify their own PVC and share it privately with you. PVC slots by plan: none on Free/Starter, 1 on Creator/Pro (and legacy Scale), 3 on Scale (and legacy Business). Policies: Terms of Service and Prohibited Use Policy apply; infringement claims by voice owners are reviewed. Comparison: TikTok AI Sel...
• articleWhat does audiocraft's LICENSE_weights say about hosting, redistribution, modification, training and output use of the AudioCraft model weights?
  Scope: a text observation of the generic weights licence file; per-checkpoint model-card review has not been done. audiocraft LICENSE_weights at 896ec7c4 is generic CC BY-NC 4.0 legal code that names no model, checkpoint or weights (curator text search: 0 matches); hosting, training and outputs are not mentioned.
• articleSeedance 2.5 limits from BytePlus's own API reference: 30 images / 10 videos / 10 audio, 4-30 s, 1080p (10-bit HEVC), draft tasks, and no direct real-face references
  Primary ModelArk docs (model id example dreamina-seedance-2-5-260628) settle creator-reported numbers. Omni reference-to-video accepts 0-30 reference images, 0-10 reference videos and 0-10 reference audio clips plus optional text (audio-only input allowed); reference videos 2-30 s each (4-30 s for edit tasks) with a 30 s combined cap; audio clips 2-30 s with a 30 s combined cap (Seedance 2.0: max 3 videos/3 audio, 15 s caps; first/last-frame and multi-image modes: 2.5 takes 1-30 images vs 1-9 for 2.0). So the "50 references" claim is the sum of maxima (30+10+10), and "12 references" fits 2.0 (9 images + 3 videos/audio). Output: Seedance 2.5 default 720p, also 480p and 1080p (1080p output uses 10-bit colour and H.265/HEVC); Seedance 2.0 supports up to 4K; 2.0 fast/mini 720p max. A separate LAS doc says 1080p is unsupported for 2.5 in reference-image scenarios (R2) - check per endpoint. Duration: 2.5 default -1 (auto) or 4-30 s (2.0: 4-15 s); edit tasks require duration -1, ratio adap...
• articleAsset-library production tools (Artlist Studio, Apr 2026): characters with generated angles and voices, multi-angle locations, shot/angle/lighting presets and structured prompts
  Creator-reported: statements attributed to the creator(s) come from discovery's reading of auto-caption transcript excerpts; the curator checked the video metadata (title, date, length) and any cited documentation pages, not the spoken claims. Nothing was tested; promotional or sponsored framing applies where stated. Mikkel Lassalle (Artlist affiliate, 2026-04-21). Artlist Studio project model: create a character from one reference image, then "generate angles" for multi-view consistency and pick a voice (custom voices announced as coming); a library of pre-made characters; locations generated from a prompt plus multiple angles for location consistency; in prompts you @-tag characters and locations. Prompt helpers: angle presets, shot presets (close-up, extreme close-up, full, wide, extreme wide), lighting presets, and a "structured prompting" mode with fields for characters, location, action, composition, style and mood. Generate a style frame, "set start frame", then in a directin...
• articleStyle frames first, prompt like a cinematographer; and the self-initiated brand concept film as a portfolio format
  Creator-reported: statements attributed to the creator(s) come from discovery's reading of auto-caption transcript excerpts; the curator checked the video metadata (title, date, length) and any cited documentation pages, not the spoken claims. Nothing was tested; promotional or sponsored framing applies where stated. Dreamina video (promotional, 2026-07-14): lock a style frame before animating because it drives all later animations; generate it with GPT Image 2 from a reference photo of the creator; write prompts "like a cinematographer" - camera angle, lighting, atmosphere, lens, colour grade, mood, environment, subject detail; use multiple reference images and prompt tweaks to converge on a style; then animate in Seedance 2.0 with omni-reference (image, video, audio). Lamborghini concept film (2026-08-14): a self-initiated, no-brief commercial for a real car brand mixing traditional VFX and AI; the transcript is mostly the voiceover and music (no technical breakdown audible). Righ...
• articleWhat can AI agents do inside an editing workflow today (trimming, synthetic B-cam angles, motion graphics, LUT grading, sound), and where do they fail?
  Creator-reported: statements attributed to the creator(s) come from discovery's reading of auto-caption transcript excerpts; the curator checked the video metadata (title, date, length) and any cited documentation pages, not the spoken claims. Nothing was tested; promotional or sponsored framing applies where stated. Dan Kieft (2026-09-29, Higgsfield-sponsored) uses a desktop chat agent ("GPT Astra", as he names it) with MCPs for Higgsfield, DaVinci Resolve (paid version) and Blender, plus computer-use control of Premiere Pro/After Effects: (1) trim dead space from transcription, insert editor-note adjustment layers and mark candidate B-cam/C-cam spots - works but slow (about 14 minutes for some trims via screen control; the Resolve MCP is much faster) and needs monitoring; local desktop app needed for large camera files (cloud upload limit 512 MB); (2) synthetic B/C-cam angles: give a photo of the room and the A-cam clip, generate side angles with Seedance 2.5 (plausible shadows an...
• articleWhich paid provider models can run inside ComfyUI as built-in "Partner" (API) nodes, and how does that change workflow design?
  Creator-reported: statements attributed to the creator(s) come from discovery's reading of auto-caption transcript excerpts; the curator checked the video metadata (title, date, length) and any cited documentation pages, not the spoken claims. Nothing was tested; promotional or sponsored framing applies where stated. The ComfyUI docs index (captured 2026-10-23) lists 298 built-in Partner node pages: Video 126, Image 101, 3D 45, Audio 16, Text 10. By name prefix the largest groups are Tripo (24), Recraft (20), Bria (19), Kling (19), ByteDance (18, i.e. Seedance/Seedream), Luma (15), Vidu (15), Flux/BFL (12), Wan API (12, incl. Wan 3 reference/image-to-video), Gemini (10), MiniMax (9), PixVerse (9), Meshy (8), ElevenLabs (8), Rodin (7), HeyGen (7), Grok (7), OpenAI (7), Runway (7), Tencent (6), Ideogram (6), plus Magnific, Fish, Topaz, Krea, Reve, LTX, Sync, Beeble, Veo, Claude and Higgsfield entries. Implication: a single ComfyUI graph can mix local open-weight models with hosted clo...
• articleHow do AI music and sound-design tools differ in control: one-shot song generators vs timeline-based scoring (Suno, Lyria on Artlist, ACE Studio)?
  Creator-reported: statements attributed to the creator(s) come from discovery's reading of auto-caption transcript excerpts; the curator checked the video metadata (title, date, length) and any cited documentation pages, not the spoken claims. Nothing was tested; promotional or sponsored framing applies where stated. Three sponsored demos by the same creator show a spectrum. Suno (2026-06-08): text-to-song with generated lyrics, instrumental toggle, exclude-styles, vocal gender, "weirdness" and "style influence" sliders, a sound-effects generator; generate 4-6 variants to choose direction; he states paid plans grant commercial usage rights (not verified here). Lyria 3 Pro on Artlist (2026-03-30): up to 3-minute tracks, custom/auto lyrics or instrumental, genre, mood, theme (commercial, documentary, social, film) and tempo controls, prompt enhancer, and an "Artlist sound" mastering option; Google lists Lyria 3.5 as its current flagship music model in the Gemini API (C3-B343). ACE Stu...
• articleIs Fish Audio S2 "open-source", and what can it do for AI voice-over (emotion tags, cloning)?
  Creator-reported: statements attributed to the creator(s) come from discovery's reading of auto-caption transcript excerpts; the curator checked the video metadata (title, date, length) and any cited documentation pages, not the spoken claims. Nothing was tested; promotional or sponsored framing applies where stated. Mikkel Lassalle (2026-03-10, Fish Audio link) calls Fish S2 "probably the strongest AI open-source model" and demonstrates bracketed emotion/breath tags (whispering, soft or loud voice, inhale, exhale), cloning "from as little as 15 seconds of audio", multilingual output, streaming and an API; workflow: draft a trailer script, polish with ChatGPT, generate, iterate quickly, cut picture to the narration. Licence check on Hugging Face (2026-10-23): fishaudio/s2-pro and the newer fishaudio/drama-3 (2026-10-17) are under a "fish-audio-research-license", and s1-mini is CC-BY-NC-SA-4.0 (gated); the fish-speech code repo shows NOASSERTION. So the weights are openly downloadabl...
• articleHow can you build a whole AI short shot-by-shot locally with consistent characters and voices (LTX 2.3 + FLUX.2 Klein), and what licences apply?
  Creator-reported: statements attributed to the creator(s) come from discovery's reading of auto-caption transcript excerpts; the curator checked the video metadata (title, date, length) and any cited documentation pages, not the spoken claims. Nothing was tested; promotional or sponsored framing applies where stated. Mickmumpitz (2026-05-04): per shot, FLUX.2 Klein generates a start frame from up to 4 references (character face close-up + slightly turned full-body image auto-merged into a sheet; environments/assets too), then LTX 2.3 generates video with audio; an "LTX ID" LoRA plus a voice sample keeps the character voice consistent; each shot group can reference previous shots and color-match the first shot; a shot duplicator and assembler build the edit, and swapping the reference images and voice sample re-casts the whole film without touching prompts (keep character description vague in prompts for this). Paid version adds a 360-degree environment and crop group for consistent ...
• articleWhat do openly readable sources say about letting sound lead, follow or bridge the picture when moving between scenes?
  Partially verified: the curator re-read none of the cited sources; S1, S4 rest on the producer's pinned reading. Two openly readable sources address it, from different angles. Pudovkin (1929 address and an essay on asynchronism) argues that sound should be edited like picture, heard without its source, and not tied to exactly synchronised speech; his examples are mainly inside scenes. Bordwell (2008) classifies across-scene links by sound and image, with sound-to-sound and sound-to-image links common and image-to-sound rare. No primary editor account of sound bridges (J-cuts, L-cuts) with public text was found; only secondary tutorials turned up in search and were not used.
• articleFor a first dubbing handoff, what exactly goes into the Netflix premix dialogue WAV and the dialog editorial session, what stays out, and how are M&E and dialogue kept apart?
  Two dialogue-only deliverables are defined: (1) a mono premix dialogue WAV for QC, an offline bounce of the whole picture length with group walla attenuated 10 dB and a -6 dB true-peak limiter on the dialog tracks or bounce bus; (2) a dialog editorial session of raw, untreated, synced dialogue, one track per character, zipped store-only, uploaded only after the premix passes QC and arbitration and mostly needed for centralized mixing. The M&E is not a studio deliverable in these articles: Netflix provides it, the studio listens to it while recording (do not re-record reactions already in it), and the mixer leaves it unaltered. The premix article never says whether M&E is summed into the premix bounce; "dialogue units" suggests not, which is an inference. Stems named: the OV dialogue stem (reference only) and the 5.1 dialogue pre-dub for extended versions.
• articleWhat do the two Netflix dub premix/mix articles require for timing and sync (picture reference, frame rate, lip-sync, stretching, QC order), and is any tolerance or timecode field given?
  Sync is specified qualitatively: premix dialogue must be in sync with the provided picture and the bounce must span the whole picture or guide-track length; the editorial session must keep the OV printmaster frame rate, sample rate, bit rate and file format; stretching is a last resort with a listening check; the mixer should tighten lip sync and calibrate the monitoring chain. No numeric tolerance, timecode start, reel or sync-pop is stated in either article (bounded absence check). Timing therefore lives in region positions of the session and in the length of the WAV; script event timing is a separate TTAL item covered by C2-B10-G2-04, and priority moments by C2-B10-G2-05. DAPT keeps begin/end as its own data area and ptulsconv uses clip timecodes (C3-B212-UO-AM02-01/-02).
• articleIn Netflix's two dubbing premix/mix articles, how is performance and delivery intent (effort, breath, emotion, direction) specified for the dubbing studio, and is there any direction-note or emotion field?
  Neither article defines a direction-note, emotion or performance-note field, and neither mentions the dubbing director. Performance intent is specified only by reference to the original version (OV): efforts, breaths and screams must match it, dialogue level, projection and tonality must match it, and reactions are recorded only where the original has them. The two numeric items are a level example (-14 dB regular, -6 dB loud, unit unstated) and a warning not to make an over-upfront dub. Where Netflix states directing intent and lip-sync priorities (Key Moments, PLDL/Show Guide) is in other pages already covered by C2-B10-G2-05; the DAPT/ptulsconv comparison (C3-B212-UO-AM02-01/-02/-03) shows formats that at least offer a free-text note, Reason or EFF slot, which these two Netflix articles do not.
• articleI am building a dubbing or ADR handoff: where do I put emotion/effort notes, timecodes and M&E handling, and does any public format tell me?
  In the two public handoff documents inspected in this batch, none defines an emotion or breath/vocal-effort field. W3C DAPT requirements (Group Note, 08 May 2025) say emotion and intonation data are a possible extension whose format is left undefined (stated for audio description), and ask for a free annotation string per dubbing script event; ptulsconv has a Reason tag and an EFF line-category flag. Timing and mix are separate: begin/end and Mixing Instructions in DAPT; clip timecodes, Ver and Reel in ptulsconv. A beginner should therefore keep three separate columns or fields (timing, direction note, mix/stem note) and name the convention in the file, since no standard supplies one.
• articleIn a documented practitioner ADR cue-sheet tool (ptulsconv), which fields describe why/how a line is re-recorded, which are timing, and how are they split across director, supervisor and talent outputs?
  ptulsconv 4.0.2 documentation (open-source Pro Tools ADR report generator) describes line tags: cue number, character, actor, prompt line, Reason (R), priority, time budget in minutes, and boolean flags including EFF (Effort), TV, OPT, ADLIB. Timing comes from clip timecodes plus Ver and Reel tags. It has Reason and EFF flag but no emotion/intent field; any performance note must be a custom {note=...} field. It generates distinct Director, Supervisor and Talent outputs. No M&E field is documented; a Dept tag is shown with Foley as an example.
• articleDoes the W3C DAPT dubbing/audio-description script format (a public standard) have a performance-intent field, and how does it keep that apart from timing and from mix/M&E handling?
  At W3C DAPT Candidate Recommendation Draft 26 June 2026, timing, free-text script-event annotation, character/text, and mixing instructions are four separate data-model areas. There is no dedicated emotion, effort or breath field: the closest slots are the free-text Script Event Description (ttm:desc), its small provisional descType registry (pronunciationNote, scene, plotSignificance) and x- user-defined types, plus Rate/Pitch for synthesized speech. No M&E or stem concept appears in the document; mixing is gain/pan. So an agent can use DAPT to keep three kinds of data apart, but must define its own performance-intent convention.
• articleWhat does the REAPER ReaScript API documentation say about BWF timecode start offsets and project drop-frame frame rate?
  The curator re-read all cited web sources (S13); nothing was executed. REAPER's scripting API can insert media at its BWF timecode start offset (InsertMedia flag 4096) and read the project frame rate and drop-frame state; the API text has no iXML, pull-up or sample-rate-conversion documentation (the user guide was too large to read).
• articleWhich audio track positions do EBU R 123 and R 48 assign to mixes, international sound (M&E), audio description and multichannel content?
  Partially verified: the curator re-read S8, S9; S4 was not re-read by the curator and rests on the producer's pinned reading. EBU R 123 (2009) tabulates audio track allocations for 2-16-track file exchange (e.g. 4b stereo mix on 1-2 with international sound or effects on 3-4; 4c mix plus audio description or foreign language; 8a/8b stereo plus discrete 5.1), defining M&E as Music and Effects and allowing other allocations; R 48-2005 gives older tape allocations such as case 2b (mix on 1-2, international sound on 3-4).
• articleWhat does WCAG 2.2 require for captions, sign language and keyboard operation of media, and what does W3C WAI say about presenting sign language?
  Partially verified: the curator re-read S33, S35; S34, S36 were not re-read by the curator and rest on the producer's pinned reading. WCAG 2.2 requires captions for prerecorded audio in synchronized media at Level A (except clearly labelled media alternatives for text) and sign language interpretation only at AAA; WAI notes most policies do not require sign language, judges sign-language avatars not adequate, and advises placing the signer (usually bottom right, above tickers) with a visible toggle. These are content criteria, not legal duties.
• articleHow do open implementations (libtimecode, libltc, FFmpeg) compute drop-frame timecode, and which frame rates do they support?
  libtimecode 85af0c9a header restricts drop to 2997/100; FFmpeg (83195411 = 4bef4614 bytes) applies drop to multiples of 30 with 4-label skip at nominal 60; libltc bf84b010 skips labels 0 and 1 at each minute whose units digit is not 0 when dfbit is set (static code reading, not run).
• articleAcross China's Measures, Korea's Article 31, California's AI Transparency Act and New York's synthetic-performer ad law, which media, actors and label forms does each text address?
  Scope: synthesis bound to the dated legal-text observations C2-B60-K4-06, -07 and -08; it is not additional evidence. Curator-checked synthesis across China, California and Korea texts: media scope, label form and addressed actors differ as offered; technical specifics are delegated outside the statute in China (national standard) and Korea (decree). The cross-text term mapping is synthesis, not any source's statement.
• articleWhat does Netflix's Sound Mix Specifications v1.6 require for printmaster, stems, dialogue loudness and naming?
  Scope: dated archive snapshot of the Netflix Sound Mix Specifications v1.6 (Netflix-branded deliveries); preferred over v1.4 for loudness. Netflix Sound Mix Specifications v1.6 (change log 2024-10-29; Internet Archive copy of 2026-08-09) applies to branded content only and keeps -27 LKFS +/-2 LU dialog-gated with peaks not above -2 dB true peak (BS.1770-1); the -24 LKFS program-gated fallback for under-15% dialogue now sits in a best-practices section stated to be not mandatory (BS.1770-3 or -4).
• articlePROPOSAL (1916 practice as the basis): what check could an agent run on its own short script draft using only what Sargent names in Chs. LXI-LXII?
  1916 practice: Sargent names checks but gives no pass mark: speech individual and natural, playing time longer than reading time, business marked apart from speech, a reconstruction that explains the emotions felt, the plot improved, rest then re-plot, a reader who asks questions, many plots before polishing. PROPOSAL (untested, synthesis): an agent turns these into a five-item checklist on a synthetic draft and compares drafts made with and without the loop. No source here tests the checklist.
• articleIf a talking picture were written in 1916, how did Sargent say the writing differs: dialogue rules, timing, layout and settings, and does dialogue change photoplay writing as he states it?
  1916 practice: Sargent treats talking-picture writing as stage-play writing, not a change to photoplay writing (dialogue defeats the end of the motion picture; the phonograph is unlikely to heighten the photoplay). His rules: dialogue natural, not stilted blank verse; characteristic and individual (banker vs bully, maid vs mistress); timed by playing (four minutes read may need five or six); brief plays with few settings (a hedged prediction); manuscript with speech and business separated. By ID only, C3-B237-S61-02 and -04 (Baker) share the aim of natural, individual speech; Sargent gives no test.
• articleWhat did Sargent predict in 1916 about talking pictures (Ch. LXI), and what should a photoplay writer expect or not expect from them?
  1916 predictions, not facts: Sargent expects talking pictures (his term for a phonograph, a projection machine and a timing device) never to replace the photoplay or the silent drama, though they may become an established amusement. He expects the camera to support recorded song or speech, with the chief use in seeing and hearing operatic or dramatic stars and in vaudeville specialties, and he doubts sound will be used to heighten the photoplay. He reports a 1913 revival that did not hold interest. His advice to writers is not to study the work much until its permanent success is assured. Later history was not examined.
• articleDid Netflix's M&E delivery guidance (fully filled M&E, stems, true-peak) change when the M&E guides and the 2.0 article moved to the Knowledge Hub?
  The scripted and nonfiction M&E guides read on 2026-10-22 match their Wayback copies (2026-08-09 and 2026-06-09) apart from markup, a stray word and a dropped Help Center pointer: both still ask for a fully filled long-play discrete M&E matching the original mix, with nonfiction depth of fill agreed with the Netflix Post Manager. The Near Field 2.0 article differs from its 2026-06-13 copy in one delivery-format detail: SMPTE ST 2067-21:2023 is now listed alongside the 2016 and 2020 editions for IMF Application #2E; its 2.0 M&E rule (fully filled, no dialog, -2 dBFS true-peak) is unchanged.
• articleDo the loudness, true-peak and stem requirements in Netflix's current Sound Mix Specifications page differ from the archived v1.6 text, and what did the last logged revision change?
  Comparing the Knowledge Hub page read on 2026-10-22 with the Wayback copy of v1.6 from 2026-08-09, the requirement text is the same: -27 LKFS +/- 2 LU dialog-gated (BS.1770-1, whole programme), -2 dB True Peak ceiling, -2.3 dBFS limiter recommended, DME stems that sum to the 5.1 mix, theatrical exempt from the loudness target. After removing whitespace, punctuation and markup, the two texts differ only in dropped section numbers and one heading rewording. The newest change-log entry on both is OC-1-6 (2024-10-29). One unlogged edit shows in the archive record: the Best Practices note was reworded between the 2025-01-23 and 2025-10-12 captures.
• articleNetflix Partner Help Center links for the nonfiction M&E guide, the Sound Mix Specifications and the dub-card article give 404: were they deleted, and where does each live now?
  At 2026-10-22 about 11:05 UTC the three Help Center article URLs answered HTTP 301 to pages on the Netflix Knowledge Hub (studiopartner.netflix.net/studio/...), while the Zendesk help-centre JSON API answered 404 for the same ids; the 404 is an API-route result, the pages were moved, not deleted. The successor slugs are branded-m-e-creation-non-fiction, branded-sound-mix-spec-and-best-practices and dubcard-conformance. A Wayback memento shows the redirect existed by 2026-10-12, after the last archived 200 copy of v1.6 with body (2026-08-09); the exact move date is not recoverable from these sources.
• articleWhere should cuts fall in a dialogue scene (line boundaries, reaction, overlap) and how are rolled edits and split edits used to create cutting points?
  Only one inspected source describes dialogue cutting points in procedure form: the Wikibooks Movie Making Manual 'Scene Editing' page (revid 4341818, community text). It says a rough cut that follows each speaker looks like a tennis match, and recommends locking the dialogue audio and rolling the picture edit forward to show the first speaker's reaction or backward to show the next speaker's anticipation (the L cut / J cut), with the gap between lines the only adjustable audio element. Leake et al. place one cut decision per dialogue line. The OER textbooks give no cutting-point rule; early manuals time a title to the utterance.
• articleWhat does Unity's own documentation say about the Audio Recorder's sample rate and bit depth, and which part of CA-B309's audio claim rests only on mirror source code?
  Official Recorder 5.1.7 docs state: WAV is the only Audio Recorder format, only Mono or Stereo is supported (from Project Settings > Audio > Default Speaker Mode), and the audio goes to the Recorder rather than the system output. They state no sample rate, bit depth or sample format anywhere in the five official pages read (0 hits for sample rate, Hz, bit depth, float, 32-bit). So CA-B309's "WAV header rate = AudioSettings.outputSampleRate at Stop(), 32-bit IEEE float" is absent from Unity's documentation and rests on AudioRecorder.cs from the mirror; the "no rate/bit-depth property" and "mono or stereo only" parts are corroborated. The mirror's RecorderAudio.md equals the official page word for word.
• articleHow should these Netflix partner help-centre pages be cited and dated, given that the HTML carries no date and the Zendesk API gives three timestamps?
  None of the four assigned pages prints a revision date. The Zendesk article API gives created_at, updated_at and edited_at: Live Action updated 2023-12-31T18:24:45Z (edited 2020-12-07T06:28:31Z), Animation updated 2023-11-30T20:02:40Z (edited 2020-06-25T00:10:03Z), 5.1 Mix Room updated 2024-04-06T11:27:25Z (edited 2020-12-07T06:30:17Z), Secondary Audio updated 2024-03-18T14:25:34Z (edited 2022-10-03T14:11:23Z). In every case updated_at is later than edited_at, and the API does not say which tracks a body change, so these are metadata, not proven revision dates. The practical citation is the HTML sha256 with the fetch time, the API values labelled as metadata, and, for a page that moved, an archive memento with its timestamp.
• articleWhere do the Netflix Live Action and Animation dialogue-recording pages agree and differ on processing, microphone distance and deliverable, and which rule applies if they conflict?
  The two pages share room, booth and chain-matching text but differ on three points. Processing: Live Action says avoid dynamic compression, limiting or EQ while recording (processing is for mixing); Animation says avoid compression and EQ but allows light peak limiting to avoid clipping and a gentle high-pass for rumble. Distance: Live Action gives about two feet for a super-cardioid boom and three feet for a shotgun; Animation gives a Neumann U87 at 30-50 cm. Deliverable: only Animation states a Pro Tools session with 24 fps timecode and 24 bit/48 kHz. Neither page says which controls when content type is unclear.
• articleI searched for the Netflix "Secondary Audio Delivery Specification": does that page contain the requirements, where do the file format, channel order, loudness and peak numbers actually live, and can I still read them?
  The Secondary Audio Delivery Specification page (API updated_at 2024-03-18) is a one-sentence pointer to "Localization, Accessibility, and Dubbing Branded Delivery Specifications"; readers voted it unhelpful (2 of 12 on the page). The live target now redirects (301) to knowledge.netflix.net, which redirects to a login, and the help-centre API returns 404 for it. An Internet Archive memento of the target (page version 5.1, captured 2024-08-10) is readable and holds the numbers: 48 kHz, 24 or 16 bit WAVE Extensible, 5.1 required and 2.0 optional, listed 8-channel orders, mono duplicated on channels 1 and 2, -27 LKFS +/- 2 LU dialogue-gated, -2 dBFS true peak, downmix trims. It is a past version of a spec whose current text was not readable.
• articleFor a Netflix dub, how are dialogue and M&E kept separate, what track or stem layout do the Live Action, Animation, 5.1 Mix Room and Secondary Audio pages state, and where do they say nothing?
  Dialogue is delivered separately from the Netflix-supplied M&E. The Animation page defines the dialogue deliverable: a Pro Tools session (AAF if another DAW), one character per track in dialogue-count order, walla/group tracks last, optional VO track, no plug-ins, no automation, pickups on new named tracks; the M&E delivery carries "Optionals" tracks the studio maps by muting regions. The Live Action page only says the dubbed voice should blend with, not sit on top of, the M&E. The 5.1 room page describes only the monitoring layout (5.0 plus subwoofer). Stem rules appear on the archived delivery-spec page the Secondary Audio page points to: for a dubbed version, a 5.1 dialogue stem that equals the 5.1 mix when combined with M&E; for the original, D, M and E stems; Atmos beds and objects separated by D/M/E (required for series, documentary, unscripted and comedy). No page gives M&E stem naming, format or who builds the combined mix check.
• articleWhat sync or timing tolerance, picture-reference rule and QC check do the Netflix Live Action, Animation, 5.1 Mix Room and Secondary Audio pages (and the delivery spec the last one points to) state for dubbed dialogue, and what is absent?
  No numeric sync tolerance for dubbed dialogue (frames or milliseconds) appears in the four assigned pages. The sync statements are qualitative or about equipment: the 5.1 room page requires picture displays and DAWs to be in frame-accurate sync (four named causes of drift); the Animation page asks for frame-accurate futz edits, clips "in sync", no time-stretch artifacts, 24 fps timecode (or as provided) and a programme start of 00:00:00:00 or the proxy. The archived delivery-spec page (version 5.1, 2024-08-10) adds that all audio must sync to final picture including 1 second of black either side, no extra leader, and measurable levels (-27 LKFS +/- 2 LU dialogue-gated; -2 dBFS true peak). Its only time tolerance, half a second, is for subtitle files, not audio. QC criteria are not listed on the four pages beyond rejection for tonality differences.
• articleHow does an open head-tracked binaural rendering specification (AOMedia IAMF) express head-locked versus world-locked rendering and reverberant filter profiles, and does it cite or encode the externalization, room-acoustics or latency evidence?
  The AOMedia IAMF specification at commit 55683fae00e541cd4f0849918f19af0e04bc629c lets an element be rendered head-locked or world-locked when head tracking is enabled, and signals a binaural filter profile of AMBIENT, DIRECT or REVERBERANT. In the text searched (whole index.bs, whitespace-collapsed), the words externalization, HRTF, BRIR and latency do not occur and the profiles carry no stated acoustic parameters, so the specification names the switches that the psychoacoustic evidence is about without citing it or giving values. The mapping to the evidence (world-locked rendering ~ source fixed in the world; REVERBERANT profile ~ reverberant versus anechoic) is a synthesis.
• articleDo individual HRTFs matter for localization, front-back confusion and externalization in head-tracked headphone playback, and how do room reverberation and training modify the answer?
  In one open-access head-tracked VR experiment (Jenny & Reuter 2020; 39 listeners, headphones, free field, noise from a visible moving drone), listeners rated their own measured HRTF as easier to localize, easier to externalize and more natural than the KEMAR or two other nonindividualized HRTFs, and reported front-back confusions in 27% of KEMAR responses versus 0% with their own. A 2020 review reports mixed results for externalization in reverberant conditions (no individualization effect in Begault 2001 and Leclere 2019; a gain in Werner 2016), and a 2021 Acta Acustica abstract reports that adding a dry room response accelerated learning of non-individual HRTFs. Which factor dominates in reverberant, head-tracked content is not settled by the sources read.
• articleBeyond the head-rotation study already catalogued (ips012), what do open-access sources report about head movement and head tracking for externalization and front-back confusion with generic or individual HRTFs, and what do they say about tracking latency?
  Catalogue ips012 already holds the head-rotation primary results (Brimijoin 2013; McAnally & Martin 2014) and the 22 to 33 ms update latency of that implementation. Added here: Hendrickx et al. 2017 (abstract only) report that head tracking with non-individualized HRTFs substantially improved externalization, especially for frontal and rear sources, and that it persisted after movement stopped; a 2020 review adds that self-initiated movement helps more than source movement and that benefit is larger for central sources; and Jenny & Reuter 2020 report self-reported front-back confusions of 27% with a generic HRTF versus 0% with the listener's own in head-tracked VR with free head movement. The sources read assert that latency matters but report no measured latency threshold.
• articleWhich cues set how far away a virtual sound seems, and what do open-access sources say about the role of reverberation and the direct-to-reverberant ratio when placing a sound at a distance?
  A 2016 open-access review (Kolarik et al.) reports that the main auditory distance cues are sound level, reverberation (notably the direct-to-reverberant ratio, DRR) and frequency, that distance is judged most accurately when level and DRR are both available, that DRR sensitivity is about 2 to 3 dB near 0 and +10 dB DRR and 6 to 8 dB at -10 and 20 dB DRR for wideband noise, and that near sources are overestimated and far sources underestimated. Reverberation helps distance but degrades azimuth localization. One primary study read here (Gil-Carvajal 2016) shows playback-room reverberation changing distance ratings for static headphone playback. No source read tests DRR with head-tracked binaural rendering.
• articleWhat happens to externalization, distance and direction when a binaural recording made in one room is played back in a different room, and does the cause lie in what the listener sees or hears?
  In one open-access experiment (Gil-Carvajal et al. 2016; 18 listeners, individual BRIRs from a 0.4 s room, static head, headphones), playing the binaural speech back in a much more reverberant room (2.8 s) left only about 18% of distance ratings correct, while direction and compactness ratings did not depend on the room; changing only what listeners could see had no effect, whereas changing what they could hear did. Reviews report a different study (Werner 2016) that found a visual effect, and caution that it lacked head motion; the sources therefore support the auditory-congruence finding for static headphone playback but not a settled answer for tracked playback.

More