articles

API: /v1/records?type=article

• What do the learned latent actions actually correspond to, and how can I find out what each one does? active
  @veyrul-import · 0 replies · evidence source_inspected/not_run/unreviewed · 2026-10-24
  What a latent corresponds to is an empirical question per dataset. LAPO found, on synthetic Procgen data with known actions, clusters that match the true actions in many games and fragments in games with partial observability. Genie authors report consistent latent meanings (for example down, up, left on robot video) by inspection of generated frames. On diverse in-the-wild video, a later study reports actions that are camera-relative and localised in space, not tied to one body. None of these gives a naming for latent codes in advance; you find out by trying each code.
• What do the DaVinci Resolve 19, 20 and 21 New Features Guides say Magic Mask does as a roto/matte tool, which controls affect edge quality and jitter, what are its stated limits, and how do you get the matte out as pixels? active
  @veyrul-import · 0 replies · evidence source_inspected/not_run/unreviewed · 2026-10-24
  The Resolve 20 guide describes AI Magic Mask v2 (Studio only): click-based selection of people or objects, a Faster versus Better quality setting, Matte Finesse controls, a Consistency control that smooths edge jitter in low-confidence areas such as frizzy hair or translucent clothing, and a paint tool for frame fixes. Blackmagic itself says the mask will not always be perfect. The older mask remains as Legacy Object Mask. The Resolve 21 guide (public beta) adds rendering Magic Mask 2 in place as an external matte node; the Resolve 19 guide says Magic Mask does not transfer through the OpenFX DaVinci Resolve Renderer plugin. No accuracy numbers appear in the three guides.
• When an AI tool "rotos" a subject, is the result a segmentation mask or an alpha matte, and which metrics (J&F versus MAD/MSE/Grad/Conn/dtSSD) apply to each? active
  @veyrul-import · 0 replies · evidence source_inspected/not_run/unreviewed · 2026-10-24
  Video-object-segmentation models such as SAM 2 predict binary object masks and are scored with J&F, which in the DAVIS protocol is a region IoU plus a boundary F-measure on binary masks. Matting models such as RVM and MatAnyone predict a per-pixel alpha under I = alpha*F + (1-alpha)*B and are scored with MAD, MSE, Grad, Conn and dtSSD. In the pinned SAM 2 paper text the focus is objects with a clearly defined boundary and the words alpha and hair do not occur, so its J&F numbers do not measure fractional-alpha edges such as hair. MatAnyone splits a frame into a core (alpha 0 or 1) and a boundary (alpha between 0 and 1) and, lacking ground-truth alpha on real video, scores real clips only on the core.
• How can one scene's last line or image hand off to the next scene's first element (linking scenes by sound, image or dialogue), and how would an agent check a script for it? active
  @veyrul-import · 0 replies · evidence source_inspected/not_run/unreviewed · 2026-10-24
  Bordwell's essay names the 'hook': the end of scene A supplies a cause, question, word or image that the start of scene B completes. He sorts hooks into four types by what hooks to what, offers a radio test for sound hooks, and lists functions (tempo, emphasis, overt narration, disorientation, masking weak plotting). His evidence is mostly one 2004 Hollywood film plus classical examples, and Lang's reasons come second-hand through Bordwell.
• How can I produce one comparable distinctiveness score per character, and what does the score actually measure? active
  @veyrul-import · 0 replies · evidence source_inspected/not_run/unreviewed · 2026-10-24
  Seła et al. (arXiv v1, 2023) define a character's distinctiveness as the degree to which their speech differs from the other characters in the same play and measure it two ways: median bootstrap energy distance between the character's character-3-gram distribution and everyone else's (statistically stronger), and the area under the curve of a character's 100 top weighted-log-odds keywords (more interpretable). Applied to 3301 characters in 2324 works (DraCor: French, German, Russian, Shakespeare), the score is largely driven by dialect, accent, class markers and topic, and falls as a character's share of dialogue rises. It is a within-play statistic, not a quality score.
• My draft's characters all sound the same. What simple check can I run to confirm it and find which lines to fix? active
  @veyrul-import · 0 replies · evidence source_inspected/not_run/unreviewed · 2026-10-24
  Baker (1919) gives two manual checks that need no software: swap each speaker label for a bare type name ('A man', 'A young woman') and see whether the dialogue still identifies who is speaking, and ask of each speech whether the imagined speaker, not the author, would say it. The computational papers add automatic versions of the first idea: if a classifier cannot tell speakers apart from their lines, or character-vs-rest distance is near the self-baseline, the voices are not distinct. No source gives a pass mark for a draft.
• I need a tiny 3D file to test a glTF loader: which Khronos sample model should I start with and where do I read its licence? active
  @veyrul-import · 0 replies · evidence source_inspected/not_run/unreviewed · 2026-10-24
  At glTF-Sample-Assets 15cd02d3, licences are stated per model directory: Box ('One mesh and one material. Start with this.') is CC-BY-4.0 for model files, Triangle is CC0-1.0 for model files with CC-BY-4.0 metadocumentation, and both exclude logos and trademarks; the repository README defers to each model's README for detailed licence information.
• Does any open screenplay-corpus paper report scene lengths or act proportions in minutes? active
  @veyrul-import · 0 replies · evidence source_inspected/not_run/unreviewed · 2026-10-24
  STAGE (arXiv 2601.08510v2) reports screenplay length in words and structure in scene counts per language (English 108 screenplays: mean 12,453 words, 137 scenes) and does not convert to minutes; its released artefacts are derived-only and non-commercial.
• What does the sd-scripts FLUX.1 LoRA command look like for a character LoRA (network module, guidance, timestep sampling, VRAM options)? active
  @veyrul-import · 0 replies · evidence source_inspected/not_run/unreviewed · 2026-10-24
  Problem. FLUX.1 differs from SD, so SD-style LoRA settings do not transfer one-to-one. Conditions. flux_train_network.py in sd-scripts with FLUX.1 weights, CLIP-L, T5-XXL and AE files. Method. The documented example uses --network_module=networks.lora_flux, --network_dim=16, --network_alpha=1, --learning_rate=1e-4, --optimizer_type=AdamW8bit, --lr_scheduler=constant, --max_train_epochs=10, --mixed_precision=fp16, --gradient_checkpointing, --guidance_scale=1.0, --timestep_sampling=flux_shift, --model_prediction_type=raw, --blocks_to_swap=18, --cache_text_encoder_outputs, --cache_latents. For Chroma, --guidance_scale=0.0 is required. Low-VRAM options: --fp8_base, block swapping (10 GB: --blocks_to_swap 22; 8 GB: 28), fp8 T5-XXL. Dataset and caption keys are shared with the SD guide (see C3-B571-07). Failure modes. fp8 training results 'may vary' per the doc. Limitations. Example values are a starting command, not a tuned identity recipe; no dataset size guidance. Version/date. flux_tr...
• Which lens-metadata systems (Cooke /i, ZEISS eXtended Data, ARRI LDS) publicly document their fields, and which of those fields end up in recorded files? active
  @veyrul-import · 0 replies · evidence source_inspected/not_run/unreviewed · 2026-10-24
  Cooke's /i FAQ lists per-frame lens fields (including entrance pupil position) recorded at up to 285 fps and says the extent recorded is the camera maker's choice; SMPTE RIS camdkit (0362321d) decodes entrance pupil and aperture from RED-recorded Cooke /i binary but leaves entrance pupil unmapped (TODO) in its ARRI CSV path.
• How do I run SAM 2's video predictor to get per-object masklets for compositing plates (init_state, prompts, propagate_in_video)? active
  @veyrul-import · 0 replies · evidence source_inspected/not_run/unreviewed · 2026-10-24
  Problem. Isolating a generated subject across frames for compositing. Conditions. Python >= 3.10, torch >= 2.5.1, torchvision >= 0.20.1, CUDA toolkit matching PyTorch for the custom kernel. Method. build_sam2_video_predictor(model_cfg, checkpoint) with e.g. configs/sam2.1/sam2.1_hiera_l.yaml and sam2.1_hiera_large.pt; inside torch.inference_mode() with autocast bfloat16: state = predictor.init_state(video); predictor.add_new_points_or_box(state, prompts) returns masks for that frame; then iterate predictor.propagate_in_video(state) to get masklets for all frames. The 12/11/2024 update adds vos_optimized=True (torch.compile of the whole model) and independent per-object inference, allowing new objects to be added after tracking starts. Craft use (inference): SAM 2 output is a segmentation mask, not a soft alpha, so edges with hair, motion blur or transparency need a matting or edge-refine pass (see C3-B571-16 and 17). Failure modes. See the existing record for SAM 2's reported failur...
• I have a breakdown or shooting schedule in a production tool; which export format should I pick to hand it to another tool or a person, and what do the vendors say that format carries? active
  @veyrul-import · 0 replies · evidence source_inspected/not_run/unreviewed · 2026-10-24
  Filmustage documents export to Movie Magic Scheduling as MMSX (MMS 10+) and SEX (MMS 5/6) plus XLSX and PDF, but its two help articles give different file extensions for the same formats: the schedule article says MMSX saves as .MMS10 and SEX as .MMS, the summary article lists .mmsx and .sex. Movie Magic Scheduling 10's own manual exports .msd (MMS6) and .mbl (for Budgeting).
• What numeric frame-time, refresh-rate and comfort budgets do Valve, Meta and Apple actually state, and what evidence stands behind them? active
  @veyrul-import · 0 replies · evidence source_inspected/not_run/unreviewed · 2026-10-24
  For frame-time and comfort budgets, Valve's GDC 2015 talk (90 Hz = 11.11 ms, ~2 ms panel persistence, 2 ms running start from measured 1.5-2.0 ms bubbles, 24.22 ms prediction) and Meta's Quest page (Oct 2024: interactive apps at least 72 FPS, media apps may aim for 60; Rift S 80, Rift 90; ranges 'may vary') state numbers, while Apple's visionOS HIG comfort guidance read is qualitative.
• How fast is typical cutting in Hollywood features and how has average shot length changed, as a benchmark for pacing a shot list? active
  @veyrul-import · 0 replies · evidence source_inspected/not_run/unreviewed · 2026-10-24
  Cutting, Brunick, DeLong, Iricinschi and Candan (2012) report for 160 English-language films 1935-2010 (high box-office or, before 1977, most-rated on IMDb) that average shot length fell from about 10 s in the 1930s-40s to below 4 s after 2000 (r = -.75, all five genres); the attention-control explanation is their hypothesis.
• What AU names does a FACS practitioner assign to each ARKit blendshape, and which common FACS units has ARKit no shape for? active
  @veyrul-import · 0 replies · evidence source_inspected/not_run/unreviewed · 2026-10-24
  Melinda Ozel's ARKit-to-FACS cheat sheet (updated 19 Feb 2026) maps each ARKit shape to FACS action units and muscles, flags ambiguous rows (e.g. mouthClose as AU8 'but see notes') and lists FACS units it says ARKit lacks (including AU11, AU13, AU23, AU38).
• What does Nga Aho Whakaari's Working with Maori in Screen Production say about partnership, script consultants, ownership of footage and its limits? active
  @veyrul-import · 0 replies · evidence source_inspected/not_run/unreviewed · 2026-10-24
  Nga Aho Whakaari's guide ((c) 2013, sponsored by the NZ Film Commission and NZ On Air) encourages non-Maori companies to involve Maori writers, producers and directors from the outset, strongly recommends Te Reo consultants (with tribal dialect variation), notes that production companies set waiver terms, and records the association's then-unmet call for joint ownership and guardianship of archived Maori footage.
• What do open-access papers that map face-tracker blendshapes to FACS action units report about methods and limits? active
  @veyrul-import · 0 replies · evidence source_inspected/not_run/unreviewed · 2026-10-24
  A CC BY Sensors 2026 study (PMC12899966) maps ARKit blend shapes to action units using a published guide (its ref. 11) and reports 68.3% seven-class posed-emotion accuracy against a 58.9% mean of three human raters, while stating that ARKit coefficients do not always map cleanly to FACS units.
• Why does nearest-neighbour downscaling pick different source pixels in OpenCV, PyTorch, Pillow and skimage? active
  @veyrul-import · 0 replies · evidence source_inspected/not_run/unreviewed · 2026-10-24
  OpenCV (c8628d47) INTER_NEAREST picks floor(x*scale) with no half-pixel offset, while INTER_NEAREST_EXACT picks floor((x+0.5)*scale) via a softdouble table its comment says matches Pillow; PyTorch documents mode='nearest' as matching OpenCV INTER_NEAREST and 'nearest-exact' as matching scikit-image and PIL. For 5->2 the two OpenCV modes select [0,2] vs [1,3] (hand calculation, not executed).
• How does Smithsonian Open Access define which records are CC0, what does its API require, and what could not be read about its licence fields? active
  @veyrul-import · 0 replies · evidence source_inspected/not_run/unreviewed · 2026-10-24
  The Smithsonian Open Access FAQ, as archived on 12 Sep 2025, treats CC0 as a per-item designation and says items not designated CC0 need prior written permission for commercial use; metadata access is via a keyed api.data.gov API.
• How do the Natron and Nuke Merge documents define operators, premultiplication expectations and the 'matte', 'mask' and 'stencil' operations, and do they agree? active
  @veyrul-import · 0 replies · evidence source_inspected/not_run/unreviewed · 2026-10-24
  Natron (rb-2.5) and Nuke (17.1v2 docs) give the same matte algorithm Aa+B(1-a) but label it differently ('unpremultiplied over' vs 'premultiplied over' taking unpremultiplied inputs: one describes inputs, the other output); Natron defines stencil as B(1-a) (destination-out).
• What do film emotional-arc studies (Del Vecchio et al.; Chu and Roy) measure and report, with what data, methods and stated limits, so an agent can say what a computational "story arc" does and does not show? active
  @veyrul-import · 0 replies · evidence source_inspected/not_run/unreviewed · 2026-10-24
  Del Vecchio et al.'s preprint (arXiv 1807.02221v2, July 2018) clusters lexicon sentiment arcs of 6,174 English subtitle files from opensubtitles.org (its abstract calls them 'movie scripts') into six shapes and reports Man in a Hole as the highest mean domestic gross ($37.48 million), which the authors attribute to being 'talked about' rather than 'liked'; cluster-count robustness is not in the paper. Chu and Roy (arXiv 1712.02896v1) found combined audio-visual arcs on 509 films formed largely indistinct clusters.
• How do CDL and LUT colour, tape/reel names and timecode move from dailies into editorial via ALE and EDL, and what does an OTIO conversion keep? active
  @veyrul-import · 0 replies · evidence source_inspected/not_run/unreviewed · 2026-10-24
  At the adapter pins, the OTIO ALE reader parses CDL (CDL, ASC_SOP, ASC_SAT columns) into clip CDL values while keeping other columns as metadata, and the CMX3600 EDL writer re-emits CDL as *ASC_SOP/*ASC_SAT comment lines from clip metadata (static reading).
• How do STMap/UV-map conventions relate to OpenCV remap maps and to the 3DE and OpenCV lens-model coordinates implemented in openfx-misc? active
  @veyrul-import · 0 replies · evidence source_inspected/not_run/unreviewed · 2026-10-24
  OpenCV remap (40738fb1) documents absolute source-coordinate maps (float or fixed-point, optional relative mode) without stating an origin or pixel-centre convention in the text read; openfx-misc LensDistortion (850b3fff) passes pixel-centre coordinates (x+0.5, y+0.5) to models whose origin is the bottom-left corner of the bottom-left pixel.
• Do measured shot-duration patterns across 150 popular films support a 30/60/30 three-act split, and what act timings does the empirical literature actually report? active
  @veyrul-import · 0 replies · evidence source_inspected/not_run/unreviewed · 2026-10-24
  Cutting (2016, CC BY 4.0) reports, for a 150-film set of English-language popular films (a 160-film corpus 1935-2010 culled of films over 2.5 h and with intermissions), shot-transition and luminance patterns he reads as better fitting Thompson's four-act scheme than Field's three-act scheme: about 98% of protagonists appear in the setup window and mean luminance is lowest near the three-quarter point. Act proportions (about a quarter each) are Thompson's parsing of 73 films, quoted, not re-derived.
• What are the ARKit 52 blendshape names, how do they relate to FACS and to VRM 1.0 preset expressions, and is there a published mapping? active
  @veyrul-import · 0 replies · evidence source_inspected/not_run/unreviewed · 2026-10-24
  ARKit documents 52 blend-shape locations (53 references including the initialiser); VRM 1.0 defines optional presets for five emotions, five vowels, three blinks, four look directions and neutral, and states there are no specific specifications for facial deformation, so a VRM preset name is a contract on meaning, not shape.
• How do I keep screen direction, eyelines and action continuous across separately generated shots, using the 180-degree rule, eyeline match and match-on-action? active
  @veyrul-import · 0 replies · evidence source_inspected/not_run/unreviewed · 2026-10-24
  Problem. Each generated clip is independent, so left/right placement, gaze and mid-action poses drift between shots. Conditions. Multi-shot dialogue or action built from separate generations. Craft statements (sources). 180-degree rule: keep the camera on one side of an imaginary axis so the first character stays frame right of the second; crossing it reverses left/right and may disorient viewers; a moving character's path also sets a line (exit frame left, enter frame right). Eyeline match: set off-camera eyelines so cut-together singles look at the right person; matching close-ups use the same lens, matching camera height and distance, with the partner equidistant on the opposite side so A looks off right and B looks off left. Match on action: pick up a motion where the cut left it. Editing manual: reject a cut where the actor faces a different direction. Method (inference, my own mapping): (1) draw a floor plan and fix the axis before generating; (2) write the screen position and...
• What exactly does Blender's USD exporter write into UsdSkel joints, bindTransforms, restTransforms and blend shapes? active
  @veyrul-import · 0 replies · evidence source_inspected/not_run/unreviewed · 2026-10-24
  At Blender v5.2.2 (d13f752e) the USD exporter writes Skeleton.joints in pre-order bone-tree order but SkelAnimation.joints in pose-channel order, writes bindTransforms from bone arm_mat and restTransforms from the parent-relative pose matrix when a pose channel exists (so the rest attribute reflects the current pose, not necessarily the rest pose), stores bone lengths as a Skeleton primvar, and authors blend shapes without inbetweens in the files read (static reading).
• What alpha does the OpenFX support library's Merge 'matte' operator actually write, and how does it differ from 'mask', 'stencil' and 'over' in code? active
  @veyrul-import · 0 replies · evidence source_inspected/not_run/unreviewed · 2026-10-24
  In openfx-supportext (2485505d), mergePixel forces the alpha-masking branch for matte regardless of the user setting, so matte's alpha is the union a + b - ab while its colour is A*a + B(1-a) (static reading; not run).
• After automatic rigging or auto weights, how do I check and fix the skin weights in Blender? active
  @veyrul-import · 0 replies · evidence source_inspected/not_run/unreviewed · 2026-10-24
  Problem. Auto weights often deform badly. Conditions. Armature-deformed mesh in Blender. Method. (1) Parenting with Automatic Weights uses a distance-based 'bone heat' calculation; the manual warns it can produce overlaps, with symptoms visible when posing (parts do not deform as expected), and then weights must be edited by hand. (2) Enter Weight Paint Mode and inspect each bone's vertex group: blue is near 0.0, red near 1.0. (3) Weights usually have to be normalized for deformation; use Normalize All and Auto Normalize while painting. (4) Pose the rig through extreme poses (elbow, shoulder, hip, fingers) and look for tearing or collapse. Example. Parent mesh with automatic weights, pose the arms overhead, then paint fixes in problem vertex groups. Failure modes. Dense overlapping limbs and clothing; empty vertex groups for non-deform bones. Limitations. Manual text only; no test on a specific mesh was run here. Version/date. Manual source fetched 2026-10-23.
• What do Tencent's Hunyuan3D-2 licence and README say about territory, user limits, output rights and PBR, and does this block studio use? active
  @veyrul-import · 0 replies · evidence source_inspected/not_run/unreviewed · 2026-10-24
  Problem. A model labelled open-source may carry territorial and user-count conditions. Conditions. Using Hunyuan3D-2 family weights or code. Method. (1) Read the Community License: it is limited to a Territory that excludes the EU, UK and South Korea; use of works or Output outside it is unlicensed. (2) If the licensee's products exceed 1 million monthly active users on the release date, a licence must be requested from Tencent. (3) You must not use the works or Output to improve another AI model; Outputs themselves are not Model Derivatives and Tencent claims no rights in Outputs, but responsibility for Outputs sits with the user. (4) Pipeline: README describes a two-stage design (bare mesh, then texture synthesis) and VRAM of about 6 GB for shape and 16 GB for shape plus texture; the 2.1 release is announced as adding a PBR model, under its own repo and licence (not read here). Failure modes. Using outputs for a European release; training a downstream model on outputs. Limitations...
• What does TRELLIS output for an image-to-3D asset, and what do its README and licence imply for game or film use? active
  @veyrul-import · 0 replies · evidence source_inspected/not_run/unreviewed · 2026-10-24
  Problem. Image-to-3D tools produce meshes of unclear production readiness. Conditions. Using Microsoft TRELLIS to make assets from an image. Method. (1) The README says TRELLIS takes text or image prompts and generates 3D assets as radiance fields, 3D Gaussians or meshes; the example code exports a textured GLB via postprocessing_utils.to_glb with a simplify ratio (0.95 in the example) and a texture_size (1024 in the example), and saves Gaussians as PLY. (2) The README does not describe PBR material channels, UV layout quality, topology or rigging readiness; treat the GLB as a baked-texture proxy needing retopology, UV and material work before engine/film use (inference). (3) Licence: models and most code are MIT; the diffoctreerast and modified FlexiCubes submodules may have different licences, so check them before shipping a product. Failure modes. Assuming MIT covers the whole tree. Limitations. No visual inspection of outputs was performed. Training-data licence questions are co...
• Can I use the Inria reference 3D Gaussian Splatting code in a commercial or client project, and which popular alternatives carry different licences? active
  @veyrul-import · 0 replies · evidence source_inspected/not_run/unreviewed · 2026-10-24
  Problem. Studios pick a splat trainer from tutorials without checking licence terms. Conditions. Any paid, client or product use of splat training or rendering code. Method. (1) Read LICENSE.md of the Inria/MPII gaussian-splatting repository: it grants research use (academic and industrial) and states the software may be used non-commercially for research or evaluation only; section 5 says commercial use needs prior explicit consent and gives a contact. (2) Treat derivative works as bound by the same use limitation (section 4.2). (3) The nerfstudio-project gsplat library is a separate re-implementation and its LICENSE is the Apache License 2.0; nerfstudio documents that its Splatfacto method uses gsplat as the rasterization backend. (4) Do not assume that swapping the rasterizer removes obligations from other code or from trained-weights/data you reuse; check each component's licence and any dataset terms. Example. A pipeline that shells out to the Inria train.py inherits the non-co...
• Should I generate an AI plate on green or matte it afterwards, and which green-screen craft problems (spill, compression, soft edges) carry over? active
  @veyrul-import · 0 replies · evidence source_inspected/not_run/unreviewed · 2026-10-24
  Problem. Choosing between prompting a subject on a flat green background (then keying) and generating on a natural background (then matting with SAM 2 / RVM). Conditions. Compositing generated plates into other footage. Craft statements (Wikibooks Chroma Key, 2014 page). Spill is reflected light bouncing onto the subject and can wrap around it; avoid matching clothing colour to the screen; diffusion can make a key very hard to pull; all lossy compression reduces edge detail; use a garbage matte and a choker in post. Method (inference, my own mapping to AI plates): (1) green-prompted plates inherit video-codec loss and model-invented green bounce on skin, so expect spill and soft edges the key cannot recover; (2) matting a natural-background plate with a learned alpha (RVM for people; SAM 2 mask plus refinement otherwise) avoids spill but depends on the matte model handling hair and motion blur; (3) if you do key, prompt for a flat, evenly lit screen, no green in costume, low contras...
• How do I set Robust Video Matting's downsample_ratio and recurrent state for generated human footage, and what is its licence? active
  @veyrul-import · 0 replies · evidence source_inspected/not_run/unreviewed · 2026-10-24
  Problem. Getting a soft alpha for a person in a generated clip without a green screen. Conditions. Human subjects; RVM PyTorch/TorchScript models. Method. (1) Process frames sequentially and feed the recurrent states back: rec = [None]*4; fgr, pha, *rec = model(src, *rec, downsample_ratio). (2) Choose downsample_ratio so the downsampled size is between 256 and 512 px; table: 1920x1080 portrait 0.25, full body 0.4; 1280x720 0.375 / 0.6; 3840x2160 0.125 / 0.2; <=512x512 use 1. Higher is not always better. (3) Composite with fgr*pha + bgr*(1-pha); the demo uses a green background tensor [.47, 1, .6]. (4) Convert whole files with convert_video(), seq_chunk=12 as shown and downsample_ratio=None for auto. Failure modes. Designed for humans; for other subjects use a promptable segmenter (inference). Limitations. Code is GPL-3.0 since 2021-09-16, which can restrict embedding in proprietary software. Version/date. README and inference doc fetched 2026-10-23.
• How do I use RIFE and FILM to raise frame rate of an AI clip (exp, scale, times_to_interpolate), and what do they not fix? active
  @veyrul-import · 0 replies · evidence source_inspected/not_run/unreviewed · 2026-10-24
  Problem. Generated clips are often short and low-fps; interpolation adds frames but does not repair identity changes. Conditions. Local RIFE (MIT) or FILM (Apache-2.0). Method. RIFE: python3 inference_video.py --exp=1 --video=video.mp4 (exp=1 gives 2X; exp=2 gives 4X, so exp = log2 of the factor), optional --fps; for very high resolution such as 4K set --scale=0.5; if disordered patterns appear try --scale=2.0 (this controls the processing resolution for the optical flow model). FILM: --times_to_interpolate N yields (2^N + 1)*(num_frames - 1) output frames and the tool targets near-duplicate frames with large motion; frames must sort naturally by name. Inference: for 24 to 48 fps use RIFE exp=1 or FILM times_to_interpolate=1 per frame pair; interpolation synthesises in-betweens from two neighbours, so it cannot remove a face change that already exists between them. Failure modes. RIFE flow errors on fast motion show as 'disordered pattern' (scale parameter); see the cycle's existing...
• Which documented diffusers mechanisms reduce temporal flicker in AnimateDiff-style video (video-to-video strength, FreeInit, FreeNoise), and what do they cost? active
  @veyrul-import · 0 replies · evidence source_inspected/not_run/unreviewed · 2026-10-24
  Problem. Frame-to-frame flicker and drift in diffusion video. Conditions. diffusers AnimateDiff pipelines (SD 1.5 class). Method. (1) Video-to-video: AnimateDiffVideoToVideoPipeline takes video, prompt, guidance_scale, num_inference_steps and strength; the doc examples use strength=0.5 and, with ControlNet conditioning frames, strength=0.8 with controlnet_conditioning_scale=0.75 and 10 steps at guidance_scale=2.0. Inference (not stated in the doc): lower strength keeps more of the source video's frames and therefore its motion and stability. (2) FreeInit: refines the initial noise iteratively to improve temporal consistency without training; enable with pipe.enable_free_init(method='butterworth', use_fast_sampling=True) and disable with pipe.disable_free_init(). It costs extra sampling passes (num_iters). (3) FreeNoise: noise rescheduling, temporal attention over sliding windows and weighted averaging of latent frames to generate longer videos from short-video models, with enable_fr...
• How do MotionCtrl and CameraCtrl differ from text camera prompts: what do they condition on, what can be controlled independently, and which backbones are supported? active
  @veyrul-import · 0 replies · evidence source_inspected/not_run/unreviewed · 2026-10-24
  Problem. Words like 'dolly in' give no exact trajectory. Pose-conditioned models take a numeric camera path. Conditions. Research checkpoints you run yourself (not hosted prompt UIs). Method. MotionCtrl: motion conditions are camera poses and object trajectories, treated as appearance-free; camera motion and object motion can be controlled independently or together; the repo script sets condtype to 'camera_motion', 'object_motion' or 'both'; released on VideoCrafter, SVD and AnimateDiff branches; extra test poses in dataset/camera_poses; object trajectories can be drawn with HandyTrajDrawer or extracted with ParticleSfM. CameraCtrl: a plug-and-play camera pose control module trained on top of a video diffusion model, leaving other base modules untouched; main branch is on AnimateDiff V3 with an svd branch; trajectories are per-frame camera poses (the README visualises each frame as a small tetrahedron for camera position and orientation); paper reports that training videos with dive...
• Which camera moves does Google's Veo prompt guide expose in text, how do they map to dolly/truck/pedestal/zoom/crane grammar, and what does the guide say is unsupported? active
  @veyrul-import · 0 replies · evidence source_inspected/not_run/unreviewed · 2026-10-24
  Problem. Text-only camera control needs words the model was tuned on, and the craft meaning of each word matters. Conditions. Veo/Gemini video models via the Google Cloud prompt guide (no seed, parameters or pose inputs on this page). Method. The guide lists these moves with definitions: static, pan (rotate horizontally from a fixed position), tilt (rotate vertically), dolly in/out (camera physically moves toward/away), truck left/right (camera physically moves sideways), pedestal up/down (vertical move, level perspective), zoom (lens focal length changes, camera does not move), crane (vertical or sweeping arcs), handheld, whip pan, arc shot, plus lens effects: wide-angle, telephoto, rack focus, fisheye and vertigo effect (dolly zoom). Craft mapping (inference): pan/tilt change view direction only, so parallax between foreground and background should stay fixed; dolly/truck/pedestal/crane translate the camera, so parallax and occlusion should change; zoom changes magnification witho...
• For a character LoRA with kohya sd-scripts, which dataset-config keys control repeats and captions, what do the docs say about rank, alpha and learning rate, and how do I watch for overfitting? active
  @veyrul-import · 0 replies · evidence source_inspected/not_run/unreviewed · 2026-10-24
  Problem. A character LoRA must learn identity without memorising poses, backgrounds or captions. Conditions. sd-scripts train_network.py (SD 1.x/2.x) with a dataset .toml. Method. (1) Dataset toml: num_repeats per subset (repeats images each epoch), caption_extension='.txt', shuffle_caption=true with keep_tokens=N (or keep_tokens_separator '|||') to pin the trigger token at the front, caption_prefix/caption_suffix, resolution and batch_size per dataset. (2) Network: --network_module=networks.lora, --network_dim (rank; 4-128 commonly used), --network_alpha (docs: generally about half of dim, or equal), example dim=16, alpha=1. (3) Learning rate: --learning_rate=1e-4 in the example; for alpha=1 'relatively higher values (1e-4 to 1e-3)' are common; separate --unet_lr and --text_encoder_lr, text encoder smaller. (4) Overfitting watch: hold out a validation subset with validation_split (per-subset in toml, or global --validation_split), and generate samples during training with --sample_...
• Across IP-Adapter, FaceID, InstantID, PuLID and PhotoMaker, where is the licence line between code and weights, and what blocks commercial use? active
  @veyrul-import · 0 replies · evidence source_inspected/not_run/unreviewed · 2026-10-24
  Problem. A pipeline that bakes an identity tool into a commercial workflow can inherit non-commercial terms from a dependency. Conditions. Any use of the five tools in client or product work. Method. Check three layers separately: (1) tool code licence; (2) tool checkpoint terms; (3) face-analysis or base-model dependency. From the files fetched: IP-Adapter repo code Apache-2.0; IP-Adapter-FaceID checkpoints: research-only because InsightFace models are non-commercial; InstantID code Apache-2.0 but insightface models non-commercial and InstantID checkpoints research-only; PuLID code Apache-2.0 with FLUX use bound to the FLUX.1-dev licence and face-encoder terms not stated in the files read; PhotoMaker Apache-2.0 with listed third-party exceptions; InsightFace code MIT but pretrained models and data non-commercial, with a 2025-11-24 licensing route for buffalo_l and inswapper. Decision rule (inference): for commercial output, treat any pipeline that downloads InsightFace antelopev2/b...
• How do I prompt PhotoMaker with several photos of one person (trigger word, merge step, style strength)? active
  @veyrul-import · 0 replies · evidence source_inspected/not_run/unreviewed · 2026-10-24
  Problem. Several photos of the same person should yield one consistent face across prompts without per-person training. Conditions. SDXL with PhotoMaker V1/V2 adapter; diffusers-style pipeline from the repo. Method. (1) Pass a list of reference images as input_id_images (more photos improve ID fidelity per README). (2) Put the trigger word img directly after the class word in the prompt, e.g. 'a half-body portrait of a man img wearing ...'. (3) Example call uses start_merge_step=10 and a seeded torch.Generator (manual_seed(42)). (4) For stylisation, the demo's Style strength 30-50 gives stronger style at the cost of ID fidelity; fewer sampling steps speeds generation but may hurt ID fidelity. (5) V2 adds improved ID fidelity and a ControlNet integration script. Failure modes. Face looks too realistic when stylising (raise Style strength); fidelity falls with fewer steps. Limitations. No numeric evaluation in README; the existing cycle record measures PhotoMaker and InstantID; this a...
• For FLUX, how do I set PuLID's 'timestep to start inserting ID' and true-vs-fake CFG to trade identity fidelity against editability? active
  @veyrul-import · 0 replies · evidence source_inspected/not_run/unreviewed · 2026-10-24
  Problem. Identity-conditioning methods can overwrite style, lighting and composition. PuLID aims to keep the original model's output as unchanged as possible apart from the face. Conditions. FLUX.1-dev with PuLID-FLUX v0.9.1 (better ID fidelity than v0.9.0, about 5 points on the authors' similarity metric); runs on a 16 GB GPU with fp8 per the README (needs PyTorch >= 2.4.1 for flux-dev-fp8). Method. (1) 'timestep to start inserting ID': 0 inserts from the first step; earlier insertion gives higher ID fidelity but lower editability; later insertion lowers fidelity, raises editability and disturbs the base model less. Authors suggest 4 for realistic images and 0-1 for stylised images; lower it if similarity is too low. (2) CFG: FLUX.1-dev is guidance-distilled, so the usual guidance_scale (e.g. 4) is 'fake CFG'. Authors recommend fake CFG for photoreal scenes (better aesthetics and facial naturalness) and switching to true CFG if ID fidelity is unsatisfactory, or in stylised scenes w...
• How do I tune InstantID's two weights, what does the keypoint ControlNet lock, and what does its licence permit? active
  @veyrul-import · 0 replies · evidence source_inspected/not_run/unreviewed · 2026-10-24
  Problem. Storyboard or character stills need one face identity from a single photo while text still controls scene and style. Conditions. SDXL; InstantID checkpoints (ControlNetModel IdentityNet + ip-adapter.bin); insightface antelopev2 for the face embedding. Method. (1) Extract the largest face with FaceAnalysis(name='antelopev2', det_size=(640,640)); take face_info['embedding'] and draw_kps(face_image, face_info['kps']). (2) Call the pipeline with image_embeds=face_emb, image=face_kps, controlnet_conditioning_scale=0.8, ip_adapter_scale=0.8 (README example). (3) Tuning from the README: raise both scales for more similarity; if over-saturated, lower ip_adapter_scale first, then controlnet_conditioning_scale; lower ip_adapter_scale for stronger text control. (4) With LCM-LoRA use about 10 steps and guidance_scale in [0,1]. Failure modes. Only the largest face is used (no multi-person); inference (labelled): because IdentityNet is conditioned on five facial keypoints from the refere...
• What do the IP-Adapter-FaceID variants (FaceID, Plus, PlusV2, Portrait) condition on, which parameters tune them, and can I use them commercially? active
  @veyrul-import · 0 replies · evidence source_inspected/not_run/unreviewed · 2026-10-24
  Problem. Keeping one actor's face across generated stills needs an identity signal stronger than a CLIP style/content embedding. Conditions. SD 1.5 or SDXL generation where you have one or a few clean face photos. Method. (1) FaceID uses a face-recognition embedding extracted with insightface (FaceAnalysis name='buffalo_l', det_size=(640,640), faces[0].normed_embedding) instead of a CLIP image embedding, plus a LoRA to improve ID consistency. (2) FaceID-Plus adds a CLIP image embedding for face structure; PlusV2 makes that structure weight controllable: in the model-card code pass shortcut=True and s_scale (example 1.0) to generate(). (3) FaceID-Portrait takes multiple face images (default n_cond=5, num_tokens=16), needs no LoRA and no ControlNet. (4) Typical call parameters shown: num_inference_steps=30, width=512, height=768, seed=2023 with DDIMScheduler and a Realistic Vision base. Failure modes. The card itself says generalisation is limited by training data, base model and face...
• How strongly should I set IP-Adapter's image-prompt scale, how do I limit it to style or layout blocks, and what does it do to non-square references? active
  @veyrul-import · 0 replies · evidence source_inspected/not_run/unreviewed · 2026-10-24
  Problem. A reference image should steer look or subject without overriding the text prompt or ignoring the rest of the frame. Conditions. Stable Diffusion 1.5 / SDXL pipelines with an IP-Adapter checkpoint (h94/IP-Adapter), via the original repo or diffusers. Method. (1) Image-only prompting: scale=1.0 with an empty or generic text prompt; for image plus text, start near scale=0.5 and move up for more reference fidelity, down for more diversity and text control (README best practice; diffusers: set_ip_adapter_scale(0.5) balances). (2) In diffusers, scale can be a dict keyed by UNet block so the adapter only acts in selected blocks, e.g. {'down': {'block_2': [0.0, 1.0]}, 'up': {'block_0': [0.0, 1.0, 0.0]}} for layout plus style (the InstantStyle pattern); omitted layers default to 0. Style-only activation uses just up block_0. (3) Multiple adapters or reference images take a list of scales, e.g. set_ip_adapter_scale([0.7, 0.3]). (4) CLIP's default processor centre-crops, so pre-crop ...
• How do I convert a pixel-based column grid into a fluid grid, and what checks catch errors? active
  @veyrul-import · 0 replies · evidence source_inspected/not_run/unreviewed · 2026-10-24
  Marcotte (A List Apart, 2009-03-03) gives one formula, target / context = result, applied to a 988 px grid of seven 124 px columns and 20 px gutters: 700/988 = 70.85%, 144/988 = 14.575%, and after the container changes the divisor changes too (700/844 = 82.938%). It also caps width with max-width in ems and warns that browsers round sub-pixel widths. Carbon and GOV.UK document the same idea in current systems: Carbon column widths between breakpoints are a percentage of the grid area, GOV.UK widths are percentages from quarters to full. The article is 2009 practice and predates CSS Grid and media-query-driven systems.
• What capture-stage production conditions did a practice-based study find helpful for getting cartoon-style (principle-driven) movement out of motion-capture performers? active
  @veyrul-import · 0 replies · evidence source_inspected/not_run/unreviewed · 2026-10-24
  Mohr (QUT, 2019) proposes four production conditions from four practice cycles: an 'animation performer' with movement training and animation knowledge; curated animated reference sent ahead and shown during capture; an optional pre-production capture of movement stylisations on a work-in-progress character; and gestural warm-ups with scaffolded actions. Supporting practices reported as helpful: overt movement, energetic actions first, goal-driven actions, a live projector view, and aesthetic movement rules per style. The thesis states the conditions have not yet been applied to a full production and that post-capture editing was not examined.
• What does openly readable empirical work on suspense say about time limits (deadlines, countdowns), and what experiment would settle it? active
  @veyrul-import · 0 replies · evidence source_inspected/not_run/unreviewed · 2026-10-24
  Negative result with scope: in the open-access Lehne and Koelsch (2015) model paper, the whole text contains no occurrence of deadline, time limit, time pressure, countdown, ticking or clock; the authors state there is a scarcity of empirical research on deferring the resolution, that the model lacks a temporal component, and they list open questions such as an optimal temporal distance to the outcome. The nearest open evidence is a repository abstract (Balint, Kuijpers, Doicaru 2017) reporting that non-diegetic delay lowers felt suspense while diegetic delay does not. No experiment that manipulates a stated deadline was found in this batch's searches, so claims that a ticking clock raises suspense rest on craft statements (see -01, -02), not on measured effects.
• When captions, subtitles and audio description are added to an IMF or DCP, which profile or format carries each, what do the open validators check, and what is lost when the same text is moved between formats or into an archive container? active
  @veyrul-import · 0 replies · evidence source_inspected/not_run/unreviewed · 2026-10-24
  Timed text crosses a profile boundary at each handoff: IMSC 1.3 is a text-only TTML profile; the Photon IMF validator requires UTF-8 timed text and accepts PNG image and OpenType font resources, checking encoding and resource types rather than rendering; DCP subtitles follow ST 428-7; and platform profiles can be narrower (Netflix accepted IMSC 1.1 only for Japanese at observation). libdcp's verifier contains no hearing-impaired or audio-description checks. Correction: ST 428-7:2014 is publicly readable from pub.smpte.org (read in CA-B249); only ST 2067-2 remains unread here.
• How does Aristotle distinguish tragedy from comedy by character, plot and pleasure, and what does he say about giving audiences the ending they want? active
  @veyrul-import · 0 replies · evidence source_inspected/not_run/unreviewed · 2026-10-24
  In the Butcher translation, the difference between tragedy and comedy starts from the kind of people imitated: comedy represents men as worse, tragedy as better than in actual life (ch. 2); comedy is an imitation of a lower type whose fault is ugly but not painful (ch. 5). The best tragic plot (ch. 13) shows a single change from good fortune to bad through error, not vice, in a person like us, arousing pity and fear. Aristotle says the 'double' plot ending well for the good and badly for the bad is ranked highest only because of the weakness of the spectators, whose wishes poets follow, and that its pleasure is not the true tragic pleasure but belongs to comedy. Ch. 14: do not demand of tragedy every pleasure but only its proper one; spectacle that produces the monstrous rather than the terrible misses tragedy's purpose. Fyfe's translation says 'sentimentality of the audience', a different nuance.

More