- Remove dead detect_anomalies() (superseded by adaptive variant)
- Remove unused r_ref param from infer_color_from_neighbors()
- Remove write-only c_ema state from anchor hook
- Use math.exp() instead of torch.tensor()+torch.exp() in bilateral loop
- Use in-place .add_() for accumulation in bilateral filter
- Pre-normalize padded tensor once in compute_local_relationships()
- Fix video VAE fallback producing duplicate vectors in calibration
ComfyUI loads custom_nodes by directory name (ComfyUI-LCS) but Python
cannot import module names with hyphens. Register ComfyUI_LCS as a
sys.modules alias so SubspaceLab and other plugins can use
`from ComfyUI_LCS.core.xxx import ...` regardless of load order.
The anchor node was implemented but missing from __init__.py registrations.
Also removes probe node references (exposure, spatial variance, chroma AC)
which have been moved to ComfyUI-SubspaceLab.
Two bugs caused sharpness edits to shift color even when calibrated
with lcs_data:
1. Calibration order was wrong: DC removal before LCS color removal
operated in the wrong space (LCS basis has large DC components),
producing incorrect color projection and incomplete removal.
Fix: remove color first (in raw space), then DC.
2. Intervention DC removal broke orthogonality: subtracting
mean(pc1_dir) introduced a color component because the all-ones
vector projects onto LCS basis. Fix: re-orthogonalize against
LCS basis after DC removal.
Also: store lcs_basis in SharpnessData/safetensors so intervention
can re-orthogonalize; bump cache name to _grating to force
re-calibration with the corrected pipeline.
Grating stimuli achieve higher linearity (R²=0.94 vs 0.88) because each
contains a single spatial frequency — a purer probe of the sharpness axis.
Both methods discover the same 1D subspace (|cos|=0.986, 9.7° apart).
- Replace _gaussian_kernel_1d, _apply_gaussian_blur with _generate_grating_batch
- Generate stimuli lazily per batch to avoid 1.3GB all-at-once allocation
- Accept deprecated blur_levels parameter with warning
- Update cache filename to sharpness_{fp}_freq{suffix}.safetensors
- Update node descriptions to reflect new calibration method
When img2img denoise is low, all sigmas have low amplification so
self_anchor gets zero observe steps. This caused three issues:
1. Auto-intensity fallback hardcoded drift_signal=0.2, mapping to
maximum intensity 0.6 — excessive for img2img. Now measures actual
drift from the seeded baseline, or uses conservative 0.05 default.
2. EMA seed step returned early, wasting the first correction step
and leaving correction_index unincremented. Now falls through to
anomaly detection which naturally produces zero correction when
comparing EMA against itself.
3. The wasted seed step caused an envelope off-by-one (envelope[0]
= sin(0) = 0 on the real first correction). Resolved by fix 2.
Add feature entry, quick start guide, node table row, detailed
parameter/mode descriptions, and "How It Works" section for the
Color Anchor node. Also update the file structure listing with
new core and node files. Both EN and ZH versions in plain language.
Hoist bilateral filter and reference resize before the auto-intensity
block so the result is computed once and reused for both drift
measurement and correction. Replace drift_samples list with running
sum/count to avoid unbounded accumulation.
Default mode is now "auto", which infers the concrete mode from
connected inputs (reference+vae → reference, mask → smooth,
nothing → self_anchor) and derives intensity from runtime drift
signals instead of requiring manual tuning.
LTXV can produce latents with odd H or W (e.g. 11x20) which can't be
divided into 2x2 patches. Patchify now pads odd dims to even using
replicate padding before rearranging, and unpatchify crops back to the
original size. This keeps patch_size=2 consistent between calibration
and inference for all models.
Replaces hardcoded SCALE_FACTOR/SHIFT_FACTOR with model.latent_format.process_in/out
so LCS works with any model (FLUX, LTXV, SD, etc). Adds LTXAV packed tensor
unpack/repack support for audio+video models. All hooks (color, tone, sharpness,
observer) now use the model-aware conversion from args["model"].
LTXAV uses a flattened 1D latent layout (e.g. [1, 1, 466048]) where
spatial H/W < 2, making 2x2 patchification impossible. patchify() now
returns None for such formats, and all hooks skip intervention cleanly.
Video VAEs output [B, C, T, H, W]. Patchify now merges T into batch,
processes all frames, and unpatchify restores the 5D shape.
Updated all callers to pass extra_shape through.
Video VAEs like Wan don't support batch encoding — input B images
produces only 1 latent. Detect this by comparing output batch size
to input, and fall back to encoding images one by one.
Fixes both color calibration and sharpness calibration.
- Precompute edit_vec = (strength * sign) * dc_removed_pc1_dir once in
closure instead of recomputing pc1_dir mean removal every denoising step
- Remove SharpnessData.pc1_std (stored but never used since dropping
pc1_std multiplier)
- Blur single-channel gray noise, expand to 3-channel per-batch instead
of allocating full 201 MB 3-channel copy upfront
- Remove unused LCSData import
256 vs 512 difference (97.5% vs 97.3%) is within random variation —
different image_size consumes different random numbers from the same
seed, producing different images rather than a controlled comparison.
Keep 512 consistent with LCS color calibration. Remove per-blur-level
mean logging that was used during development.
256px gives the best PC1 (97.5%) and fastest calibration.
Larger images average more patches, reducing per-sample variance
but not improving PCA purity. Added image_size input to
LCSSharpnessCalibrate for experimentation.
RGB noise has independent per-channel variation that blur affects
differently, introducing inter-channel variance captured as PC2.
Grayscale noise (same value across RGB) ensures blur affects all
channels identically, raising PC1 from 95.8% to 96.7%.
Also added per-vector zero-mean to remove VAE brightness drift
and diagnostic logging for per-blur-level latent means.
pc1_std=12.78 represents the spread across blur levels 0-16 during
calibration, producing edits ~34x larger than the patch mean. Since PCA
basis vectors are already unit norm, strength directly controls the
L2 magnitude of the edit. Widen UI range to -5..5 to compensate.
Root cause of color shift and mosaic artifacts:
1. delta = strength * pc1_std was too large (pc1_std=12.78)
2. Removing residual when using LCS destroyed 62/64 dimensions → mosaic
3. Complex projection/residual/color-removal logic was unnecessary
The correct approach is simple: patches + delta * pc1_dir.
Adding along one direction preserves all other dimensions by construction.
LCS color removal only matters at calibration time (makes pc1_dir orthogonal
to color). At intervention time, just add along the orthogonal direction.
Removed lcs_data input from LCSSharpnessIntervene (only needed at calibration).
Removed all debug logging.
When calibration used LCS color removal, the intervention must also
remove the color component before projecting onto the sharpness basis,
then add it back after reconstruction. This ensures sharpness edits
don't affect color.
Workflow:
1. LCSSharpnessCalibrate(vae, lcs_data) → sharpness_data
2. LCSSharpnessIntervene(model, sharpness_data, strength, lcs_data=lcs_data)
Both nodes now have optional lcs_data input. When connected, color is
preserved during sharpness intervention.
When LCS color data is connected, the sharpness PC1 will be orthogonal
to the color subspace, preventing color shifts during intervention.
Changes:
- calibrate_sharpness() now accepts optional lcs_data parameter
- LCSSharpnessCalibrate node has optional lcs_data input
- Cache files use "_lcs" suffix when LCS is used (different subspace)
facok