Animation and performance

2D/3D animation, rigs, retargeting, motion capture, facial animation and lip sync

• articleWhat do the sources say about the goro-goro (gara-gara) clown scene in wayang kulit: where it sits, what it contains, whether it is mandatory, and what does that teach about comic interludes in long-form pacing?
  In Surakarta-style accounts gara-gara is a pathet sanga scene with the punakawan clown-servants (Petruk, Gareng, Bagong; Semar appears in other accounts) filled with songs and jokes, after which the clowns accompany the knight. A pattern source cited by two articles says it is not a standardised scene in Surakarta (unlike Yogyakarta) and is presented if the play is short or a great knight is meditating in the forest; Krystiadi's list marks it 'if needed'. Yogyakarta gara-gara is described as seemingly detached from the story. So the widely repeated 'clown scene at midnight' is a pattern in some styles, optional in Surakarta pakem, and sources disagree on whether it is an integral story scene.
• articleHow is an all-night Javanese wayang kulit purwa performance divided (the three pathet) and what scene types are described in each part?
  Sources agree that an all-night Surakarta-style performance is divided into three pathet (nem, sanga, manyura) played in that order, each built from a palace court scene (jejer), further scenes (adegan) and typically a battle (perang), and that the dalang (puppeteer-narrator-singer) controls the sequence with sung suluk, narration and music cues. Times differ: one article gives 9 pm-12, 12-3, 3-5; an outreach handout gives 9 pm to about 6 am. A common outline: pathet nem sets up characters and the problem, pathet sanga holds the hero's withdrawal and, optionally, clown scenes, pathet manyura brings battles and resolution. Scene-level sequences in published lists differ by source and style.
• articleWhat do actors say in BAFTA's public A Life in Pictures transcripts about performing for the camera, shot size and lens, and which of these accounts conflict?
  Actors' own accounts in BAFTA lectures describe camera distance and coverage as things they experience rather than calibrate: one marvelled that a colleague grew more relaxed as the camera moved in; another found two simultaneous cameras freeing because he never knew his framing; an actor-director says audiences project onto a face held in a long close take and judges takes by what viewers feel; another avoids learning lenses so as not to become self-aware; another describes playing straight into the lens for one director and forgetting multiple cameras on long takes. None states a rule for scaling performance to shot size.
• articleWhat empirical evidence exists on Meisner-style partner-focused training, and what does it measure?
  The one empirical study found on Meisner-style training (Sun and Okada, 2023, CC BY) analysed what actors said during exercises: over a course, novice students shifted from describing their partner's behaviour to speculating about the partner's inner state, and professional actors showed more partner connection and balanced communication. Sixteen students joined and twelve completed; the authors flag the small sample, underpowered tests, uncontrolled age and gender, and limited transferability. The study measures communication in training, not on-camera performance, so it does not show that the technique produces better acting.
• articleWhat practical steps and contract wording does Equity (UK) publish for performers asked to be digitally scanned or voice-captured, and what legal framing does it give?
  Equity (UK) advises performers to: ask five specific questions before engagement (face/body scanning, voice capture, synthesised performance, data mining and related uses, each with a description of intended use); limit any capture to one named production; negotiate an AI-training carve-out; refuse in writing at contract stage if unwilling; not sign scanning permissions on set; have scans deleted once the production is complete; and keep up to date and share the advice. Equity says no UK law is written explicitly for AI and points to copyright, performers' rights and data protection. Its vision statement sets eight cloning principles, from consent for past, current and future performances to engagement under collective agreements. These are union advice and proposed clauses; the pages report no outcomes.
• articleIf I only have a text prompt, what do MAV3D and 4D-fy actually show, how were they evaluated, which one can I run, and how good is the motion?
  Corrected 2026-10-23 - see C3-B588-07; the original statements below are kept for history and are superseded where the correction says so. Partially verified: the curator re-read none of the cited sources; S1, S3, S4, S19, S8, S16, S9, S5 rest on the producer's pinned reading. Two papers cover text-to-4D with score distillation. MAV3D (Meta, January 2023) optimises a six-plane dynamic NeRF against a text-to-video model in three stages (about 6.5 h on 8 A100) and shows an R-Precision of 83.7 against adapted internal baselines of 6.2-10.6; no code is available. 4D-fy (arXiv v2, May 2024) blends 3D-aware, image and video diffusion priors, takes about 23 h on one 80 GB A100, and ships a repository with run instructions. In 4D-fy's own user study (26 evaluators, 28 prompts) users prefer 4D-fy for appearance (92%) and structure (89%) but only 41% prefer it for motion, and the motion difference is the one comparison that is not statistically significant. 4D-fy says motion is limited to sim...
• articleHow many frames, views and what resolution does each text/image/video-to-4D system produce, and what GPU time and memory do the authors report?
  Corrected 2026-10-23 - see C3-B588-07; the original statements below are kept for history and are superseded where the correction says so. Partially verified: the curator re-read none of the cited sources; S1, S11, S3, S4, S13, S14, S15, S16, S17, S5, S6, S8 rest on the producer's pinned reading. Reported per-asset cost spans about four to five orders of magnitude, all author-measured on different hardware. Seconds: L4GM (0.3 s per 16-frame pass plus about 2 s for views, RTX 4080 Super). Minutes: DreamGaussian4D (6.5-10 min, 80 GB A100), Diffusion4D (8 min, paper), SV4D (15-20 min fitting) and SV4D 2.0 (1 min generation + 21 min fitting, up to 30 GB, H100). Hours: Consistent4D (about 2-2.5 h, V100), MAV3D (about 6.5 h on 8 A100) and 4D-fy (about 23 h on one 80 GB A100). Clips are short: 12-32 frames per pass depending on the system (12, 14, 16, 21, 24, 32), time axes are continuous for the NeRF/neural-field systems, and no passage read states output duration in seconds for text-to-4...
• articleIf I have one video (or one image) of an object and no rig, which video-to-4D generation systems can produce an animated 3D object, what do they output, and what do they cost to run?
  Corrected 2026-10-23 - see C3-B588-09; the original statements below are kept for history and are superseded where the correction says so. Four released video-to-4D lines were read at pinned versions. Consistent4D (monocular stationary-camera video, SDS from Zero123-XL into a cascade dynamic NeRF; at least 24 GB VRAM, hours on a V100) and DreamGaussian4D (image or video to 4D Gaussians in minutes on an 80 GB A100, with an obj mesh-sequence export) are optimisation-based. SV4D and SV4D 2.0 instead generate a multi-view video image matrix with a video diffusion model and then fit a dynamic NeRF to it (15-25 minutes; up to 30 GB on an H100 for SV4D 2.0). All are object-centric; SV4D 2.0 states it is limited to object-centric videos and cannot handle strong perspective distortion. Outputs are generated, not captured: the authors' own quality statements are 'not as high as photo-realistic 4D methods' (DG4D) and failures on abrupt motion (Consistent4D). Weights and rights are in candidate...
• articleI have ARKit or MediaPipe blendshape values and need FACS action units: what published mappings exist, how were they made, and which should I trust first?
  Three mappings exist and are made differently: a practitioner ARKit table (Ozel), an expert-consensus MediaPipe-52 table (CNR clinical pre-proof, no validation against real facial behaviour data) and a learned AU-to-blendshape model (ICAT-EGVE 2024 poster). None was validated against manual FACS coding of the same footage, so any table is a hypothesis to check on your own clips; keep one-to-many cases (eyeSquint AU7 with AU44; cheekPuff AU33/AU34; jawOpen AU26/AU27) as multi-label.
• articleHow do Blender's Mesh Cache modifier time modes (Frame, Time, Factor) map to PC2 and MDD frame positions, and do PC2 and MDD agree in seconds mode?
  In Scene play mode Frame time is the scene frame scaled and offset, used directly as cache index by both readers. In Time mode the modifier passes ctime/fps (seconds) and the PC2 reader divides by fps again before applying start/sampling, while MDD compares seconds to stored times; by reading, PC2 Time mode is therefore inconsistent with MDD (code inference, not run). Curator addition: Scene-mode Factor also uses ctime/fps, so Factor time = frame_scale*(ctime/fps) - frame_start, so the clamped factor reaches 1.0 after (1 + frame_start)/frame_scale seconds of scene time, not at the last cache frame (code inference, not run; amendment CA-AM-0036).
• articleWhat is an X-sheet (exposure sheet) and how do Dragonframe, OpenToonz and Toon Boom Harmony each lay out frames, layers/exposures, holds and exposure edits?
  OpenToonz 1.7.1 docs and Harmony 25 Premium menu reference (bytes match producer pins) describe the Xsheet as a frame-row by layer-column timing grid with editable exposure; Harmony defaults to overwrite and offers insert mode.
• articleTwo new permissively licensed Tencent releases for creators: Hy-Nomas (Apache-2.0 interactive world video with camera control) and AuK (MIT speech generation/editing with zero-shot cloning)
  Hy-Nomas (card created 21 Oct 2026, Apache-2.0 - notably NOT the territory-restricted Tencent community licence used for HY-World 2.0, C3-B399-01): a segment-autoregressive video "world" model - each new segment is generated from text instructions and optional camera trajectories while a persistent KV cache carries context, enabling multi-turn, free-form interaction (introduce subjects, actions, style or environment changes) and camera navigation; 8-step distilled DiT with a PixelUMM decoder; default output 720x1280 at 24 fps (480x832 option); standard inference uses 8-way sequence parallelism (eight GPUs), with an INT8/FP8 single-GPU configuration that keeps a bounded history; ships Hy-Nomas-Bench (280 test cases) and Docker/offline JSON request format with timed segments. AuK (open-sourced 9 Sep 2026, MIT; 1.5B parameters; AuK-Flash 4-step variant): one natural-language instruction interface for zero-shot and instruction TTS, content/acoustic and paralinguistic speech editing, spe...
• articleCascadeur as an "assistive AI" 3D animation tool: local inference, documented training-data provenance, no training on user animations, video-to-pose mocap alpha, and a new MCP server hook
  Nekki's Cascadeur publishes an unusually specific AI-use disclosure (R1): ML tools are AutoPosing, Inbetweening, AutoInterpolation and Video Mocap; AutoPhysics, Ragdoll, Animation Unbaking and Fulcrum cleaning are explicitly non-ML. Pose/inbetween models were trained on Nekki's own game animation (hand-key + mocap for Shadow Fight and SPINE), supplementary mocap and purchased third-party packs "certified for AI use"; the video mocap uses open-source libraries plus post-processing. User animations are not collected for training, and all tools run locally (mocap downloads a ~700 MB mocap.pt package once). Video Mocap (Alpha, R2): import an .mp4 reference, select N timeline frames, get N keyframed poses - single actor only, poses can miss nuance because a 2D view hides depth; intended for animation drafts, then refine with AutoPosing/physics. 2026.2 (R3): Additive Layers (alpha), Easing on intervals, Inbetweening/Motion Generation retrained on larger datasets and now passing style betw...
• articleYouTube automatic dubbing (Help Center, Oct 2026): eligibility exclusions (over 120 min, little speech, unsupported/undetected language, too-fast speech, copyrighted content), language pairs, expressive speech, manual-review setting and experimental lip sync
  Per YouTube Help (R1): automatic dubbing generates translated audio tracks, labelled "auto-dubbed" in the description; viewers get their preferred audio language or can switch tracks. It is on by default for eligible creators; new uploads are dubbed automatically and published according to the channel's publication setting, and YouTube "may also generate dubs for previously published videos". A video is ineligible if it exceeds 120 minutes, has no/little speech or only music, is in an unsupported or undetectable source language, has speech too fast to dub without an unlistenable speed-up, or contains copyrighted content. Accuracy depends on setting the correct original video/channel language; dubs cannot be edited - fixing the original language triggers regeneration. Language support is asymmetric: English originals can be dubbed into ~20 languages (Arabic, Bengali, Dutch, French, German, Hebrew, Hindi, Indonesian, Italian, Japanese, Korean, Malayalam, Polish, Portuguese, Punjabi, R...
• articlebvhio (addb5461, MIT) accepts exactly one ROOT hierarchy followed by MOTION; the informal Wisconsin description notes BVH lacks a full basis-pose definition. No formal BVH specification was found by the...
  bvhio (addb5461, MIT) accepts exactly one ROOT hierarchy followed by MOTION; the informal Wisconsin description notes BVH lacks a full basis-pose definition. No formal BVH specification was found by the producer.
• articleWhat it measures (R1): an Elo-like composite on OpenArt's own Arena, anchored at MiniMax H3 = 1,000; weights 30% general capabilities, 17.5% talking, 17.5% singing, 17.5% multi-character speaker assignment, 17.5% multilingual. Outputs are generated by OpenArt from procured prompts - "Model...
  What it measures (R1): an Elo-like composite on OpenArt's own Arena, anchored at MiniMax H3 = 1,000; weights 30% general capabilities, 17.5% talking, 17.5% singing, 17.5% multi-character speaker assignment, 17.5% multilingual. Outputs are generated by OpenArt from procured prompts - "Model providers cannot submit, select, or regenerate those outputs" - but OpenArt operates the arena and sells access to the ranked models, so the post itself says it "should not be described as an independent third party". Scores are relative within this board only; no confidence intervals are printed, so the 6-point Seedance 2.0 vs 2.0 Mini gap is not a demonstrated difference. Wan 3.0 is second on the same snapshot's Video Overall but last here - a general video ranking is a poor proxy for dialogue work. Scope gap: only five native audio+video generators are ranked; dedicated lip-replacement (sync, VEED, HeyGen, LatentSync) and image+audio avatar models (OmniHuman, InfiniTalk, Fabric) are absent, so ...
• articleretarget_bvh (8501d53f) writes rotation keys for mapped bones and location keys only for the parentless (hip) bone, so bone translations other than the root are not retargeted; the add-on manifest declares GPL-2.0-or-later and Blender 4.2.0 minimum (the wiki says 3.0+, a version-statement...
  retarget_bvh (8501d53f) writes rotation keys for mapped bones and location keys only for the parentless (hip) bone, so bone translations other than the root are not retargeted; the add-on manifest declares GPL-2.0-or-later and Blender 4.2.0 minimum (the wiki says 3.0+, a version-statement conflict per producer).
• articleAssimp's FBX importer (c3f68d29) declares supported FBX versions 7100-7400 ('FBX 2011, FBX 2012 and FBX 2013'); newer files fall outside that window and are handled only as the code's newer-version path allows (producer: warning unless...
  Assimp's FBX importer (c3f68d29) declares supported FBX versions 7100-7400 ('FBX 2011, FBX 2012 and FBX 2013'); newer files fall outside that window and are handled only as the code's newer-version path allows (producer: warning unless strictMode).
• articleWhen exporting a simulated garment from Marvelous Designer, what do the Alembic and cache-animation (MDD, PC2, Maya cache) exports include, and what do the help pages leave out?
  Marvelous Designer hands cloth on as per-vertex motion: Alembic (HDF5 or Ogawa) or garment-only caches (Maya .mc/.mcx, Point Cache 2, MDD in two variants). Export options cover weld vs unweld seams (which changes vertex counts and object splitting), thickness, merged vertices, unified 0-1 UVs, optional avatar shape, scale, axis conversion and frame range; strain or stress as vertex colour is documented only for Alembic. Pattern, physical-property and sewing data travel only as an optional XML with the cache export. The pages do not say how fabric materials or simulation settings reach a renderer.
• articleOverview: for capture to render, what does each file or API in the chain actually carry (markers, blendshape weights, skeleton animation, caches, render products), and what is not carried?
  Synthesis (producer, supported by verified statements): each capture-to-render carrier defines its own fields but not the mapping to the next: C3D labelled markers, ARKit/MediaPipe 52 named coefficients, UsdSkel joints with world-space bind transforms, Alembic time-sampled geometry whose topology may change between samples, UsdRender settings with no universal RenderVar names, and EXR named channels whose layer naming implies no stacking order. None carries a rig, a character or an intent, so marker-to-joint, blendshape-name and RenderVar-to-channel mappings must be authored and recorded.
• articleHow do Dragonframe's audio workspace and Harmony's sound tools break dialogue into per-frame mouth shapes, and what must exist before automatic detection works?
  Dragonframe track reading is manual: you type the dialogue and phonetic cues, Dragonframe splits the words into frame cells, and you drag them into place against the waveform. Face sets come from Photoshop layers, and the reading appears in the X-sheet and can be exported and re-imported. Toon Boom Harmony instead detects phonemes automatically and writes one of eight standard mouth-chart letters per frame, but only if the mouth layer already has drawings named with those letters. The letters are labels for mouth shapes, not the sounds themselves.
• articleWhere do the declared 1.0 meta fields (and the 0.x to 1.0 migration output) and the VRM Public License text read differently?
  VRM 1.0's meta.md defines commercialUsage values personalNonProfit (default), personalProfit and corporation, but the English VRM Public License 1.0 text uses only 'Non-commercial' and 'personal use' wording (no 'corporation' or 'personal profit'); UniVRM (efd20f62) migrates VRM 0.x commercial 'Allow' to personalProfit, never to corporation, writes the Japanese licence URL on export, and defines an English URL constant that returned 404.
• articleWhat audio does ElevenLabs say Instant Voice Cloning needs (length, quality, format), and why can more audio make the clone worse?
  Problem: make a quick voice clone from a short sample. Conditions: ElevenLabs Instant Voice Cloning (IVC), docs fetched 23 Oct 2026. Method: record about 1 to 2 minutes of clear audio with no reverb, artifacts or background noise; at least 1 minute, and the page says more than 2-3 minutes yields little improvement and can be detrimental to stability. The number of samples is irrelevant; total runtime counts. Quality means how the audio was captured, not the codec, though MP3 at 128 kbps or higher is advised (the same page elsewhere strongly recommends MP3 192 kbps or higher). Why it matters: the AI tries to mimic everything it hears, including speed, inflection, accent, breathing, noise and mouth clicks, so keep tone, performance and level consistent. Example (inference): record a single take reading in the delivery style you want in the final narration, in a quiet treated room, and trim breaths or clicks you do not want copied. Failure modes: clone inherits room reverb or noise; wi...
• articleWhat does ElevenLabs Professional Voice Cloning require for training audio (hours, level, distance, noise, single speaker, verification)?
  Problem: prepare a dataset to fine-tune a high-fidelity clone of a narrator. Conditions: ElevenLabs Professional Voice Cloning (PVC) docs fetched 23 Oct 2026; paid plans with PVC slots. Method (docs): use one speaking voice only; style in the samples is replicated, so record in the delivery style you will use (an audiobook voice should be trained on reading); use the language you will mainly need. Amount: 'bare minimum' 30 minutes, 2-3 hours for the best result; total runtime matters, not sample count. Level: roughly -23 dB to -18 dB RMS with a true peak of -3 dB. Mic: any microphone works, but XLR into an audio interface is recommended; distance about two fists from the mic; record in an acoustically treated room (a thick duvet can make a temporary space); no background music or noise; no singing. Processing options in the app can remove background noise or split speakers, but the guidance is to feed clean audio. Verification: verify with the same or similar equipment and a similar...
• articleWhat does ComfyUI-AdvancedLivePortrait offer, what are its dependency pins, and what is unclear about its licence?
  Problem. Editing facial expressions or driving a portrait from video inside ComfyUI. Conditions. PowerHouseMan/ComfyUI-AdvancedLivePortrait commit 3bba7329 (last push 2024-08-21; not archived); GitHub reports no licence; pyproject version 1.0.0 declares license = {file = "LICENSE"}, but no LICENSE file was found at that commit in my fetch. Method. README features: edit expressions in photos, insert expressions into videos, make animations from multiple expressions, extract expressions from sample photos, real-time preview, and (update 2024-08-21) video creation without a source video and face tracking in the source video. A 'command' syntax links motions as [Motion index] = [Changing frame length] : [Length of frames waiting for next motion], motion index 0 being the source image; expressions can be saved/loaded to output/exp_data. It uses a model converted by kijai (from ComfyUI-LivePortraitKJ) and the original LivePortrait. Requirements include numpy>=1.26.4, opencv-python-headles...
• articleWhich tuning knobs, restrictions and debugging tools do the Unreal Motion Matching docs state, and what do they leave unsaid about data limits?
  Unreal's motion-matching docs give tuning controls (Block Transition notify states, Normalization Sets, Mirror Data Tables, schema channels) and state that montages cannot be used with the node, but set no minimum or maximum data amount, duration or memory budget.
• articleWhat is motion matching, and what animation data and setup does an engine need before it works?
  In Unreal 5.8, motion matching is the Pose Search plugin's Motion Matching node: a Schema defines query channels and links to Database assets of animations, replacing hand-built transitions, and locomotion clips must carry root motion with the Root Motion property enabled. The Unity Kinematica (preview) and Learned Motion Matching memory/training claims remain producer-verified.
• articleHow do I choose a low-pass cutoff for noisy motion data, and is there one correct number?
  There is no single correct cutoff for noisy motion data: Winter-style residual analysis gives a starting point but is known to oversmooth, and the author of the commonly copied 0.05 cm threshold warned it was never meant to be used out of context; compute a cutoff as a first guess and check velocities/accelerations on your own data.
• articleWhat rights does the NO FAKES Act of 2025 (H.R. 2794 as introduced) create for voice and likeness, and how do licences, minors and union contracts work? (bill, not law)
  Corrected 2026-10-23 - see C3-B614-14; the original statements below are kept for history and are superseded where the correction says so. Problem: understand what a proposed US federal digital-replica right would require of producers and performers. Status: an introduced bill (119th Congress, 9 April 2025); not law as of this text, and the introduced text may have changed. Method (source text): it defines a digital replica as a newly created, computer-generated, highly realistic representation readily identifiable as a person's voice or visual likeness, either in work they did not perform or in a version whose performance was materially altered; it makes the authorisation right a licensable property right that is not assignable during life; post-mortem it lasts 10 years for the right holder plus 5-year renewals on proof of active use, ending no later than 70 years after death. Licence rules: a living adult's licence is valid only up to 10 years, must be written and signed, and must...
• articleVoice changer (speech-to-speech) versus TTS for performance-driven dubbing: what does ElevenLabs say is preserved from the source performance, and what do I have to supply?
  Problem. TTS invents timing and emotion from text; a dubbing or ADR editor who wants a specific performance (whisper, laugh, cry, cadence) needs a method that keeps the actor's delivery and swaps only the voice. Conditions. You have a recorded or live source performance and a target voice (cloned or designed) in the same account. Method. Per ElevenLabs docs, Voice changer takes source audio (file or microphone), a target voice_id and a model (eleven_multilingual_sts_v2 covers 29 languages; eleven_english_sts_v2 is English-only, and the docs say the multilingual model often does better even for English). The docs state the output keeps tone, delivery, cadence, whispers, sighs, laughs and cries; the target voice supplies timbre only. It is positioned as a complement to TTS (fix pronunciation, add performance) and to voice actors (extend range). Failure modes. The docs warn input gain matters: quiet input hinders recognition, loud input clips. There is a remove_background_noise option,...
• articleWhat does Descript document about Regenerate and Smooth jump cuts (inputs, supported footage, batch behaviour, review controls), and what would that mean for AI talking-head clips?
  Problem. Text-based edits leave jump cuts; some tools repair them with AI. Conditions. Descript help pages (23 Oct 2026); English only; uses AI credits on current plans. Method (documented). Regenerate uses AI to smooth or rewrite parts of a video: heal jump cuts, tighten pacing, clean audio, or change what a speaker said (the last needs an authorised custom AI speaker). It currently supports English audio and video only. In video projects it regenerates Audio & video by default, with an Audio only option; sequences, B-roll, multi-speaker recordings and shots where the speaker's face isn't clearly visible cannot have video regenerated; Descript asks consent the first time because the process modifies a speaker's visual performance. Smooth jump cuts scans the whole composition and regenerates audio and video at each jump cut: Find first (lists every cut for review, recommended when the composition has intentional cuts such as scene changes or camera-angle cuts) or Smooth all (5 cuts ...
• 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...
• 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...
• articleWhen should animation be cut to the music (music-first) and when should music follow the picture, and what did early cartoon practice record?
  Problem. Choosing whether score or picture leads changes the production order. Conditions. Cartoon or short-form animation with tightly synchronised sound. Method. A University of Florida radio transcript says Disney felt image and story should come first with music scored to accompany them, while Stalling believed music should come first, with animators using its mood and beat; the claim is that Disney's studio ended up with two units, story-led Mickey cartoons and music-led Silly Symphonies. Archives West records that Stalling patented a 'tick' method of recording and depicting motion pictures in 1924. Cartoon Research notes that Steamboat Willie's precise synchronisation of effects and music to the action made it a sensation. Use (inference): choose music-first for rhythm-driven pieces (dances, montage; the generated track sets the timeline), picture-first when story beats are fixed and the track is composed or edited to hits (see C3-B603-07 on Mickey Mousing). Failure modes: mus...
• 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...
• articleHow HeyGen gates real-person digital twins: consent video levels, identity matching, and the enterprise "skip consent with indemnity" option
  Closes the HeyGen consent follow-up. Every video-based "digital twin" requires proof the depicted person agreed; photo-based and prompt-generated avatars (no identifiable real person) have no consent step (consent_status null). Three access levels (developer docs): (1) webcam recording on HeyGen's hosted consent page - all customers; the API (POST /v3/avatars/{group_id}/consent) returns a link valid 24 h for the subject, who reads a statement (with a spoken code in the interactive flow); (2) uploading a pre-recorded consent video - enterprise only; (3) skipping consent entirely - enterprise only with a signed indemnity agreement. Requirements: the same person as the training footage (identity-matched), one clearly visible face, the provided statement read (validated semantically, any language), under 35 s per developer docs (help centre says under 30 s - version discrepancy), MP4/MOV/WebM, good light and audio, no recording of a screen; if creating a twin of someone else, that perso...
• articleGetting clean AI lip sync (sync. labs, Oct 2026): model choice and per-second prices, input checklist, long-video chunking and compositing colour tips
  Primary vendor docs (extends the earlier lip-sync answer to curator request CA-DR-0001-07). Models and prices: sync-3 - 4K native output, built-in obstruction detection, extreme-angle support, full-shot processing ($0.107-$0.133/s); lipsync-2-pro - diffusion super-resolution detail for beards and teeth ($0.067-$0.083/s); lipsync-2 - best balance, preserves speaking style ($0.04-$0.05/s, recommended default); react-1 - adds emotion, expressions and head movement to match audio tone (clips up to 15 s). Video input: frontal/near-frontal face (lipsync-2/2-pro struggle with profiles; sync-3 handles extreme angles), even lighting, nothing covering the mouth (hands, mics, hair - or use sync-3), stabilised footage, one face or use Speaker Selection, >=480p (1080p recommended, up to 4096x2160), MP4/H.264. For compositing back onto source footage: export SDR BT.709 with explicit colour tags and H.264 yuv444p (not 4:2:0/4:2:2) to keep colours consistent. Audio: clean speech without music/noise...
• articleSeedance 2.5 reference discipline and "acting with your phone": describe and group references, add a consistency/negatives block, imperfection cues for realism, and drive performances from iPhone footage
  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 (Aug-Sep 2026, condensed reads). Reference discipline (R1): with many reference slots, describe each reference in the prompt and GROUP references that show the same person/object from different angles (separate angle images rather than one multi-angle sheet); state the host's subject rule (he cites "one to eight distinct subjects" across subject references); add a "keep consistent" block (format clarity e.g. handheld phone, colour and grain, lighting, shadow, mood) and explicit exclusions ("no brand logos, extra people, text on screen"). Realism (R2): make the still cinematic first (e.g. 35 mm, f/1.8, shallow depth of field, fine film grain, muted desaturated ...
• articleCoverage note with small facts: 11 more sponsored or self-promotional Mikkel Lassalle uploads (Feb-Mar 2026)
  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. Concrete items: Artlist platform tour (2026-03-12, R6) - claims 50M creators; video tools with negative prompts, multiple references, multi-output, four cinematic styles and clip chaining into a long scene; custom voice cloning from about 10 seconds of audio (consent rules apply). Artlist "Big Ad" challenge with a $60,000 prize (R12) - contests as a sponsor marketing tactic. Enter.pro (R2): an "AI development agent" test building a deployable app with payments in about an hour; it recommended Gemini 3 Pro or Claude Sonnet 4.5 for the app's text generation; he configured a freemium plan (5 free shot plans/month, unlimited on Pro). Framia Pro (R1): agent platform with cre...
• articleByteDance OmniHuman v2 vs v1.5 on fal (Oct 2026): single image + audio to talking/performing video; v2 takes 4-300 s audio with an optional multilingual prompt at $0.105/s, v1.5 caps audio at 30 s (1080p) / 60 s (720p), adds a speaker mask and turbo mode at $0.16/s
  Hosted on fal (R4): three OmniHuman endpoints active - original, v1.5 (updated 21 Apr 2026) and v2 (updated 15 Oct 2026). All take one image of a human figure plus an audio file and return a video whose length equals the audio; billing is per second of audio/video ("duration ... as used for billing"). v2 schema (R1): required image_url + audio_url; audio wav/mp3, max 15 MB, 4-300 s; image jpeg/png/webp/bmp/tiff/gif/heic/heif, max 30 MB, aspect ratio 0.4-2.5, 300-6000 px per side; optional prompt for scene, movement and camera in Chinese, English, Japanese, Korean, Mexican Spanish or Indonesian; no resolution, mask or turbo parameters exposed. v1.5 schema (R2): audio must be under 30 s for 1080p or under 60 s for 720p (the schema says 720p is "faster and higher in quality"), turbo_mode trades slight quality for speed, and mask_url restricts speaking to the person in the white mask area - the only multi-person control in this family on fal. Rendered price lines (R3, 23 Oct 2026): v2 $...
• articleProducing a 14-clip AI short in Higgsfield Cinema Studio: story blueprint, saved elements, structured JSON prompts, continuation references and voice unification
  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. Youri van Hofwegen (Higgsfield link, 2026-06-11). Steps: lock a story first (an LLM drafts beats/arcs) as the blueprint for all assets; build characters - one from a multi-angle photo collage with the "auto" model (two-panel full body + face), one fictional via an "AI cast" form (genre, budget, era, archetype, gender, ethnicity, age, body type, outfit; he says genre/budget shape production look) - and four locations with a "cinematic locations" model; save each as a named element (character/location) to pull into video prompts. Video: 14 clips, each prompted in structured JSON fields (he reports a significant quality jump vs paragraphs; written with a helper tool), per-...
• articleWhat planning framework and location-consistency method does a solo AI animation workflow use (SCENE, 3D world from images, shot-by-shot 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. Youri van Hofwegen (2026-05-05, OpenArt affiliate) plans before generating with "SCENE": Story (inciting incident; decide the ending first), Characters (who and rough look), Emotion (decide the palette/mood arc up front, e.g. tension -> danger -> relief), Narrative beats (three clips = setup, escalation, resolution; one job per clip because the model does not know the next clip), Every-clip rules (a fixed style line in every prompt as a "production bible"). Location consistency: OpenArt's "3D World" builds a navigable world from 1-4 multi-angle images of a location (generate angle variants first, "Ultra" mode), then "take shot" anywhere with focal length 23-300 mm; char...
• articleHow do you build a consistent multi-character cast with LoRAs (dataset, captions, training steps) for image and video models, per a July 2026 local workflow?
  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 (2026-07-17) gives five LoRA rules: (1) a made-up trigger word absorbs everything not captioned; (2) vary what should generalize (include wide shots, lighting, poses, clothing), or the character only exists as trained (e.g. close-ups); (3) decide what stays fixed versus editable; (4) captions decide editability: captioned clothes stay changeable but must then be re-prompted; (5) captions are reverse prompts, so match the target model's prompt style (long natural language for Krea 2; structured JSON with bounding boxes for Ideogram 4). Pipeline: an edit-model "character creator" generates a varied synthetic dataset from one or two images (add close-ups of det...
• articleHow can you do markerless AI performance capture (turn an actor into a creature) with masked video-to-video, and what goes wrong?
  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 (2026-10-07) uses MiniMax H3's reference variant in ComfyUI: feed the original performance video, lock audio to the source, and in-paint only inside a mask (auto-mask from text such as "head, helmet, phone, cable" or a manual Resolve/After Effects matte) so the camera stays locked and the result composites over the plate. Identity is fixed with a character sheet; expressive emotion sheets gave much better facial performance than neutral sheets ("keep the character's identity, look, and art style from picture number one"). Findings: mask size drives head size; full-body masks enable whole-body transformation; partial masks spanning head and lower body produce...
• articleWhat practical rules help AI-rendered animation keep continuity and correct actions (guide frames, prompts, start/end frame fixes)?
  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. From the Bare Minimum short (V2, LTX 2.3): describe shot events in order, then expressions; specify handedness explicitly (a wrong-hand reach was fixed only by prompting "reaching... with her left hand"); unseen props (a newspaper) become gibberish when interpolated from a distant guide frame - paint a corrected frame in Photoshop and add it as another guide frame; when start and end frames generated separately drift in minor details, LTX warps between them - composite the start and end frames together manually so shared details match; add a second guide frame early when one is not enough; dialogue audio should be clean dialogue, at least as long as the clip, or the mod...
• articleHow do you "AI-render" a 3D layout locally (3D passes -> video model), and how did the best local approach change from Feb to Aug 2026?
  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 Mickmumpitz videos trace one pipeline across 2026. Feb (V1): Inner-Reflections' VACE + SkyReels V3 R2V merge (Apache-2.0 card) reads control passes (VACE) and reference images (SkyReels); export an outline pass (Blender Workbench flat black + white outlines, or Freestyle as a render pass) and a normalized/inverted Z depth pass with a contrast curve; outlines keep composition but let the model invent detail, depth follows geometry more tightly, blend both if needed; keep control-net strength low for start-frame image generation (Z-Image Turbo + control nets); free workflow ~120 frames at 24 fps; paid version chains 81-frame chunks with 11 overlapping start frames f...
• articleHow can you puppeteer AI video with physical props or rough animation (Time-to-Move), and how reliable is it?
  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 (2025-12-17): Time-to-Move (TTM, Apache-2.0 repo) is training-free and works with diffusion video models (he uses Wan 2.2; paper also names CogVideoX and SVD). Make a control video by moving toys/cutouts on a stick or by rough After Effects/Blender animation; TTM's "dual-clock denoising" uses lower noise where the character moves (follow the motion) and higher noise elsewhere (clean background). His ComfyUI prep: SAM 3 masks the character (add negative clicks on the stick), Qwen Image Edit removes the stick in start/end frames (also fix hovering by masking larger and prompting "stand on the table... do not change the look"), then TTM with start+end frames fo...
• articleHow does 4D Anyone create a 4D Gaussian-splat capture of a person from one phone video, and what are its limits in a ComfyUI port?
  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 (2026-09-11) ports the 4D Anyone paper (ant-research, Apache-2.0 repo) to ComfyUI: from one (ideally tripod) vertical phone video, a 3D skeleton locks pose per frame and a fine-tuned video model generates up to ~48 consistent synthetic views (it keeps already-generated angles in context and reshuffles which views generate together: far-apart first for shape, neighbours later for smooth transitions), then gsplat trains per-frame Gaussian splats. He swapped the original skeleton code (not licensed for commercial production) for Meta SAM 3D Body (SAM License) and bundled a gsplat setup script. Settings: camera preset (high-quality vs quick 16-view), Turbo on (l...
• articleCan I get an animated file I can open in Blender, Unreal or Unity (mesh sequence, PLY sequence, glTF, Alembic) out of these text/image/video-to-4D systems, and what did the authors actually release?
  Partially verified: the curator re-read none of the cited sources; S1, S3, S2, S12, S11, S6, S7, S5, S17, S18, S9, S4, S14 rest on the producer's pinned reading. Among the eight systems, only DreamGaussian4D documents a command-level path to a DCC tool: its README exports a per-frame OBJ mesh sequence and loads it in Blender with the Stop-motion-OBJ add-on. MAV3D describes a mesh-sequence conversion in the paper but has no available code. L4GM ships weights and code whose inference script writes only mp4 renders; a PLY writer exists in its repository but is not called, so a per-frame PLY sequence is an untested route. SV4D / SV4D 2.0 releases sample view-by-frame image grids, and no fitting or export code is visible by path name. No source read mentions Alembic, glTF, FBX or USD output.
• articleWhat do text-to-4D, image-to-4D and video-to-4D generation mean in practice, and for MAV3D, 4D-fy, DreamGaussian4D, Consistent4D, SV4D/SV4D 2.0, L4GM and Diffusion4D, what goes in, what representation comes out, and is it something I can run?
  Partially verified: the curator re-read none of the cited sources; S1, S3, S11, S13, S15, S5, S8, S4, S6, S12, S17, S9 rest on the producer's pinned reading. "4D" here means a 3D object or scene that changes over time and can be rendered from new viewpoints; every system read produces a generated asset, not a captured one. Three routes appear in the primary sources: (a) per-asset optimisation against diffusion priors, by score distillation (MAV3D, 4D-fy, Consistent4D) or, for DreamGaussian4D, by fitting Gaussian deformation to a generated driving video with image-diffusion guidance, (b) a diffusion model first generates a grid of views by frames, then a 4D representation is fitted to it (SV4D, SV4D 2.0, Diffusion4D), and (c) a feed-forward network that reconstructs per-frame 3D Gaussians from a video (L4GM). Output representations are dynamic NeRFs or Gaussians; only the DreamGaussian4D abstract mentions animated meshes (per-frame, candidate 02) and no abstract describes a rigged as...

More