Relative (default, unchanged) = per-prompt steering. Absolute = a keyless,
prompt-agnostic "global taste" prior that pulls conditioning toward generally-
liked directions regardless of the prompt; Both layers them.
- New keyless Absolute store (FUNPACK_ABSOLUTE_KEY): pooled liked/disliked
direction + global value function, learned across every rated generation.
- Absolute rating semantics: a rating means "details I like in general" (not
prompt adherence) — reward-driven liked/bad-dir storage, axis attribution
bypassed. Perfect = "more of this everywhere".
- _v2_finalize_conditioning replaces the per-key ascend hook on both returns;
pure Absolute also bypasses per-prompt conditioning memory.
- Scene Chain sampler: embed_guidance_source (relative/absolute) lets the
per-step nudge draw from the global direction/value function.
- Studio popup + V2 node expose Steer mode + Absolute strength.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Adds s_churn/s_tmin/s_tmax to sample_funpack_distilled_flow: optionally
re-noise the latent within [s_tmin, s_tmax] before each eval, then denoise
from the bumped sigma_hat. No extra model calls (still one eval per step,
at a higher noise level). Fights stale/frozen distilled output (more motion)
and can sharpen by letting the model re-resolve. sigma_hat is clamped to the
schedule start so the model never sees a higher-than-trained noise level, the
final jump-to-zero detail step is never churned, and the default 0.5-1.0 band
protects the sub-0.4 detail finish per the empirical sigma findings.
Churn supersedes the legacy s_noise ancestral path while active and is
video-stream-only on packed LTXAV latents. With s_churn=0 the sampler is
byte-identical to the previous deterministic ODE. Wired into the node and
Studio's Distilled Flow config panel.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Port the Hybrid/RF velocity-bias capture/apply and rating-gated reactive
rescue into sample_funpack_distilled_flow, reusing the shared trajectory
memory helpers. Adds velocity_bias_mode/strength/source/key and
rescue_mode/threshold/strength to the FunPackDistilledFlowSampler node and
wires them into Studio's Distilled Flow config panel (same blank/'default'
velocity-key -> refinement-key fallback as Hybrid). Steering is
magnitude-preserving and, on packed LTXAV latents, video-stream only
(audio-safe). With all controls off the sampler is byte-identical to the
previous deterministic ODE.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Document the three undocumented releases (2.7.5 embed guidance + online
value function + mid_scene_guide; 2.7.6 value-function ascent + rating
overhaul; 2.7.7 reactive rescue + Batch Training) and bring the Scene
Chain Sampler, Hybrid Euler-2S, and Studio docs in line with current
parameters and features.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Batch Training confirmed working end to end (controlled batch with transitions, distinct
per-variant seeds, node rating ignored during a batch, per-entry rating + VF/axis learning
on submit). Stripped the temporary trace prints (Batch Training check / Batch check / ACTUAL
scene seeds), keeping clean progress logs. Gitignore the local .testvenv.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
The live log showed the single-scene path: 'packed 3 variant conditioning(s)' with ACTUAL
scene_seeds identical across all 3 (all inherited the base entry's funpack_scene_seed). My
earlier seed fix only covered the multi-scene duplication path. When a prompt resolves to one
scene (common), _v2_build_batch_variants reused the base scene seed for every variant ->
identical noise -> identical videos (only tiny diffs from VF ascent).
Fix: _v2_pack_batch now assigns a fresh, well-spread per-variant scene seed (gen-seeded RNG,
reproducible, per-scene offset) to the packed entries in BOTH paths, after building them.
Verified in venv: single-scene 3 variants -> 3 distinct seeds.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
The 'seed=' in the batch log was the loop index seed (base+idx), not the seed that drives
noise. The real noise seed is each scene's funpack_scene_seed (set distinct per variant by
_v2_pack_batch). Print those so we can tell whether distinct seeds actually reach the
sampler vs. stale/old-code identical seeds.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
1) All batch videos identical: in the transition path each scene carries a
'funpack_scene_seed' the sampler honors over its own seed, so duplicating the scene-set
left every variant with the same seeds. _v2_pack_batch now assigns DISTINCT per-variant
scene seeds via a gen-seeded RNG (well-spread, reproducible per gen). Verified in venv:
4 variants -> 4 distinct seed sets, reproducible.
2) Ignore the node's rating widget while a batch generates: that selection would apply
learning to the PREVIOUS run, but it's stale during a batch (the batch is rated afterward
in the panel). When batch_variants>1, force rating to '-Just forget it-' (skip_learning),
so all learning paths (phrase/conditioning-memory update, VF training, velocity commit)
are skipped. Conditioning APPLICATION still runs (it has no skip_learning guard), so the
batch still uses the learned directions.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Root cause (found via the diagnostics + the venv tests): the producer was gated on
'not split_by_transitions', but the user's workflow is multi-scene chains, so with
transitions on the producer never ran -> no variant markers -> sampler did one pass.
Fix: defer variant packing to the FINAL conditioning, applied right before BOTH returns
(the transition early-return and the normal return), so it wraps whatever output ends up
being. New _v2_pack_batch dispatcher: single-scene output -> _v2_build_batch_variants
(shortcut variants); multi-scene output (transition split) -> duplicate the whole scene
set per variant (seed-only). Each variant tags ALL its scene entries with the same index,
so the sampler groups by marker into N full chains and keeps variants independent.
Verified in the test venv: 3-scene set x 4 -> 12 entries, markers [0,0,0,1,1,1,2,2,2,3,3,3]
-> sampler sees 4 groups of 3 scenes; a single-variant group returns None (chains normally,
no recursion). Marker detection + producer round-trip all green.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Prints (a) the batch_variants value + mode flags the refiner sees and how many packed
entries it produced, and (b) the positive entry count + variant markers + group count the
Chain Sampler sees. Pinpoints whether the break is the arm/settings (batch_variants==1 at
the refiner), the conditioning path (markers stripped before the sampler), or the split
logic. Temporary, to be removed once the cause is found.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
It shouldn't have been added there. The node 'Batch Training' button window is the sole
batch UI now (config when idle, rating list when finished). Removes the loadBatch() panel
and its call from renderRefiner; no leftover references, JS parses clean.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Per feedback: Batch Training is a complete UI thing driven by Studio (the hub), not a
prompt, and it must NOT start generation itself (ComfyUI's Generate does).
- Removed batch_iterations from the Chain Sampler entirely (input + param). The sampler
now enters batch mode ONLY when Studio packs variant markers into positive; the batch
count lives in Studio. _run_batch_training simplified to variant-entries-only.
- The node's 'Batch Training' button now opens a dedicated window (openBatchPanel) that
shows EITHER a config view (set batch size, Arm/Disarm) when no batch is pending, OR the
rating list (preview + per-entry rating + Submit/Forget) when a batch has finished. It
never queues — arming just sets refiner.batch_variants; the user presses Generate.
- Dynamic button label reflects live state: '(off)', '(armed ×N)', '(Processing…)',
'(Finished — rate)'. Driven by execution_start/success/error listeners, which also
one-shot disarm after a batch runs so the next Generate is normal.
The older in-Refiner-tab batch panel remains functional (secondary path); the node-button
window is now the primary entry point.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Submit now does more than train the value function (3a). New refiner method
_v2_learn_from_batch feeds each rated entry's conditioning through the SAME per-axis
direction-memory machinery as a normal rated run (_v2_update_conditioning_memory →
axis directions + liked/bad direction stores), and tallies axis votes across the batch:
because tokens were frozen, an axis rated missing/wrong in >= half the rated entries is a
high-confidence problem (not seed luck). Those consistent axes are written to
active_repair_axes, which the next generation's repair persists until a Perfect — so a
batch where most gens were 'Missing details' makes the next run boost details.
Wiring: batch submit collects (rating, serialized cond) per rated entry, trains the VF,
then calls _v2_learn_from_batch; the resulting insight ('consistent problem axes = ...')
is returned and shown in the Studio submit toast. List route now also returns each item's
variant index + prompt for the panel.
Batch Training is now complete: seed-only + shortcut variants + node button + VF training
+ axis/direction learning, all human-rated.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Producer (Studio owns conditioning, per the boundary law): refine_v2 gains batch_variants.
When >1, _v2_build_batch_variants packs N conditioning entries into the single positive wire,
each tagged 'funpack_batch_variant'. Variant 0 is the fully-processed base; variants 1..N-1
re-resolve shortcuts with seed+i (random option per entry) and get the same conditioning-memory
treatment, deduped by expansion text. No shortcut variability -> all entries resolve identical
and reuse the base cond -> the batch becomes seed-only (sampler varies the seed per entry).
Captures the pre-shortcut prompt to re-resolve. Not combined with split_by_transitions.
Trigger: a 'Batch Training' button on the Studio NODE (next to Open Studio) — prompts for N,
sets refiner.batch_variants, queues one prompt, then auto-reverts to 1 (one-shot). Studio is
the batch trigger (hub principle); no sampler input needed.
Sampler: now enters batch mode when positive carries variant markers, even if batch_iterations
is 1 (markers are the trigger; variant count wins). Each entry already saves its own cond, so
Submit trains the VF per variant — teaching which prompt variant rated better. Panel hints now
point at the button.
End-to-end: press Batch Training -> N -> Studio makes N variant conds -> sampler samples N ->
previews to temp -> rate in Studio panel -> Submit trains VF per entry -> batch cleared.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Per the Studio/sampler boundary law (Studio owns conditioning; sampler only samples), the
sampler no longer encodes anything. It now auto-detects when Studio has packed N shortcut
VARIANTS into positive — each scene entry tagged with 'funpack_batch_variant' = variant
index — via _split_batch_variants: if present, it samples one chain per variant (variant
count wins); otherwise it falls back to seed-only repeats of batch_iterations. Decisions:
'sampler loops, Studio varies' + 'pack into positive + markers'.
Each entry now saves its OWN conditioning (<id>.cond.pt) — identical across a seed-only
batch, distinct per variant — and the manifest records variant index + prompt per item.
Submit (batch_training.py) trains the VF per entry from its own cond, so variant batches
teach the model which prompt variant rated better (not just seed-quality variance). Studio
panel shows the variant tag + prompt per row.
Studio-side variant PRODUCTION (resolve shortcuts N times, encode, pack markers) is the
remaining half — a Studio/refiner feature, tracked next. Sampler side is ready for it.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Phase 1 amended: the engine now also saves the frozen conditioning (cond.pt) per batch
so submit can train on (cond, rating).
Phase 2 (panel + API): new batch_training.py registers /funpack/batch/{list,submit,forget}
(follows the existing /funpack/* PromptServer pattern). Studio gains a 'Batch Training'
section under the refiner: lists the pending batch for the active refinement key with each
entry's filename + seed, a Preview button (loads the temp webp via /view?type=temp), and a
per-entry V2-rating dropdown, plus Submit and 'Forget batch'. Previews served straight from
ComfyUI temp — no custom file route needed.
Phase 3 (learning intake): Submit loads cond.pt and trains the value function on
(cond, reward) for every rated entry — same conditioning, N rewards = variance-reduction
RLHF, which directly sharpens conditioning ascent. '-Just forget it-'/unknown ratings are
skipped. Submit consumes (deletes) the batch; Forget discards it without learning. Either
way the panel clears and the sampler is ready for a new batch.
Human stays the rating oracle throughout (no auto-rating) — 'user over everything'.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
New optional 'batch_iterations' on FunPackLTXAVSceneChainSampler. When >1, runs the
ENTIRE scene chain N times frozen except the seed (seed + i) — conditioning/settings
provably identical because each run reuses sample() itself as the per-generation
primitive with only the seed changed. This gives a clean controlled-batch RLHF signal
(same conditioning, only noise varies) instead of the confounded one-at-a-time loop
where conditioning drifts between rated gens.
Each iteration is persisted under ComfyUI's TEMP dir (funpack_batches/<key>/<stamp>/) so
artifacts are wiped on restart (ephemeral rating material) and are servable via
/view?type=temp for the upcoming Studio panel: latent (.latent.pt) + a downscaled
animated-webp preview + a batch.json manifest (seed, scene prompts, subfolder, rating
slots). LATENT output passes through the last iteration so the downstream decode/save
path is unchanged.
Batch semantics: an entry is the full set of provided scenes — scenes within an entry
chain with overlap as normal; separate entries are independent sample() calls with no
cross-entry carry. Requires refinement_key_input (degrades to a single run without it).
Phase 2 (Studio rating panel + /funpack/batch routes) and Phase 3 (batch-train intake
in the refiner) to follow.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Audit finding: value_guidance defaulted OFF, and it gated value-function TRAINING — so
by default the VF never trained, conditioning ascent + the chain sampler's embed_guidance
were permanently inert, and a user who ran many gens with it off then enabled it started
from zero (no accumulated learning).
Fix decouples learning from application:
- The VF now ALWAYS trains in the background on rated gens (cheap: small MLP, 20 steps,
no diffusion calls) whenever there's a rated previous run + refinement key. The reward
asset accumulates regardless of the toggle.
- value_guidance now controls only APPLICATION of the learned direction (output ascent +
search), via a new apply= param on _v2_ascend_conditioning. Default ON.
- Defaults flipped to True: refine_v2 signature, Studio settings read, JS default +
migration. Studio toggle relabeled 'Apply value guidance' with corrected help (training
always on; applies after 10+ ratings; off = strict prompt fidelity, learning still
accumulates). Fixed stale '5+ ratings' text (MIN_SAMPLES is 10).
Side effect: embed_guidance (chain sampler) now has a trained VF available by default.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Three issues found tracing the full Studio path with all features on:
1) BUG (silent data loss): a custom velocity_refinement_key captured trajectories
under that key but the refiner committed the rating under the effective refinement
key — so custom velocity keys never learned (banks stayed empty, rescue/velocity a
permanent no-op). This undercut per-key cut-routing. Fix: Studio resolves each pass's
velocity key (same logic as _build_one_sampler) and passes them to refine_v2, which
now commits the verdict to the union of {effective key} + velocity keys (dedup).
Backward compatible (callers without _velocity_keys behave as before).
2) Effectiveness: VELOCITY_BIAS_TARGETS (0.90, 0.80) matched only sigma~0.909 on the
standard 8-step LTX schedule, so rescue fired once per gen at a near-noise step and
never at the detail-forming sigmas. Broadened to (0.90, 0.72, 0.42) -> hits 0.909,
0.725, 0.422. Velocity APPLY still structure-only (highest target); rescue now acts
where runs actually go good/bad. (New detail targets fill as gens are rated.)
3) Redundant value: velocity_aspect_bucket was a manual text field never tied to actual
resolution, and the bank key already includes the full latent shape (which encodes
resolution) — so it was doubly redundant. Removed from the key, all sampler plumbing,
INPUT_TYPES, Studio build, and the JS UI/defaults. Persistence loads old stores fine
(the 'aspect' field is simply ignored; key is now (refkey, target, shape)).
Verified: both py files + JS parse clean; all refactored call signatures aligned.
Tests require the ComfyUI venv (torch) to run.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
UX/variability win — collapse the expert knob-soup into one control plus a
legible summary.
- Variability macro (Chill/Spicy/Chaos presets + 0..1 slider) orchestrates the
stochastic knobs: eta (video variability), eta_final (late-noise ramp, only at
high chaos), quality_sharpness (detail recovery as noise rises), velocity spice
(action injection past Spicy, source=nearest), and motion pulses. Deliberately
keeps high_quality_pct>0 (disabling the deterministic finisher under ancestral
noise = mess). Writes the existing hybrid fields; all still individually editable.
- 'Active features (this config)' readout: green/grey rows for ancestral noise,
velocity spice, rescue, quality pass, motion pulses — turns the invisible stack
legible. Notes audio is auto-held deterministic on LTXAV.
Frontend-only; no model/Python changes (all fields already plumb through).
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
The real limiter wasn't the 0.35 slider cap but the internal fixed 4.5%-of-|x|
delta clamp, which saturated almost immediately — the upper slider range was a
no-op. Now the rotation cap scales with strength (0.30*strength, capped 0.95), so
the whole range is meaningful: ~0.15 subtle spice (matches old feel), 0.3-1.0 clear
action crossover, 2-3 approaches full action replacement without wiping the current
gen. Magnitude-preserving renorm still keeps |x|, so high strength rotates rather
than blows up. Raised the param/UI cap to 3.0 and updated the tooltip.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
eta_final was silently ignored whenever the quality pass was off: the decay was
anchored to quality_sigma_start, which collapses to sigma=0 when high_quality_pct=0,
so effective_eta stayed pinned at eta for every step. Combined with the missing
deterministic finisher, full-strength ancestral noise ran through the final detail
steps with nothing to resolve it -> noisy mess.
New _effective_eta() ramps eta -> eta_final across raw schedule progress
(sigma/sigmas[0]), so eta_final works with the quality pass on or off. Used in both
the RF and eps paths. No-op when eta_final == eta (default), so existing runs are
unchanged.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Root cause of consistently bad LTXAV audio: ComfyUI packs the video+audio
latents into one [B,1,N] tensor for sampling (pack_latents, video first), and
the RF sampler's ancestral re-noising (eta/s_noise) + motion pulses + velocity/
rescue steering sprayed the whole packed tensor — including the audio slice.
Ancestral re-noising every step destroys audio coherence (video tolerates it,
and in fact benefits from the scene variability).
Fix: recover comfy's per-stream latent_shapes from the guider conds, build a
[1,1,N] video-region mask, and confine every FunPack perturbation to the video
slice. Audio now rides the clean deterministic euler-RF step:
- ancestral noise branch restructured into x_det (deterministic, audio) vs x_anc
(full ancestral, video), combined via the mask.
- motion pulses, velocity-bias apply, rescue, and quality sharpness masked too.
Single-stream LTXV (mask None) is unaffected — behavior identical to before.
Only the CONST/RF path needs this (LTXAV is always CONST); eps path sees no AV.
Prints a one-line diagnostic when packed AV is detected.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Compensate the slow quality degradation from velocity-bias 'apply' (additive
mean-pull) with four guards on the Hybrid Euler 2S sampler (both eps and
rectified-flow/LTXAV paths):
- B: apply is now magnitude-preserving — rotate x toward (x+delta) and
renormalize to |x| instead of x+delta, so no energy is injected.
- C: apply fires only at the highest (structure) velocity target and decays
strength with the sigma ratio; detail sigmas left untouched.
- E: new quality_sharpness knob — temporal-average unsharp on the x0 prediction
during the quality phase to restore high-frequency detail (free, no extra eval).
Default 0.0 (off).
- D: new velocity_bias_source = mean|nearest — 'nearest' steers toward the single
best-matching prompt cluster (one real good gen's detail) instead of the
washed-out average. Affects both apply and rescue. Default 'mean' (off).
Plumbed through get_sampler/INPUT_TYPES, Studio _build_one_sampler, and the
Studio JS sampler panel (both presets + UI controls).
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
The good/bad trajectory banks were in-memory only, so they died on every
ComfyUI restart — losing valuable rating-learned data. Persist them per key
under refinements/attn_maps/velocity_<hash>.pt (same store as blessed maps),
saved on each commit, loaded once-per-process on first access (rescue/apply/
commit). Directions stored as float16 to halve disk. Now only the deliberate
Studio Session-reset button (clear_velocity_bias_memory) deletes the store; a
restart preserves it. Serialized as weights_only-safe lists/dicts (no tuple
keys); load failures degrade gracefully to empty banks.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Root cause of "rescued toward an Awful gen": velocity capture was
rating-blind — it wrote every sampled trajectory straight into the one
"good" memory, so a fresh key whose only capture was an awful gen made that
awful run the sole steering target. Ratings (which live in the refiner) never
touched the bank.
Now:
- capture STAGES the current gen's trajectory (VELOCITY_BIAS_PENDING) instead
of committing it; staging runs whenever rescue is on (no separate capture
toggle needed).
- the refiner commits the previous gen's staged trajectory by its rating:
like/loved/nailed -> good bank, awful -> bad bank, everything else dropped.
- rescue steers TOWARD the good bank and AWAY from the bad bank (per prompt
cluster), firing when diverged from good OR aligned with bad beyond
threshold; no-op when neither bank has a prompt match.
- session reset / clear_refinement_data now clears good + bad + pending.
velocity-bias "apply" now also reads the rating-gated good bank. Tooltips and
Studio note updated.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
VELOCITY_BIAS_MEMORY is an in-process dict, so reset_session (which only
removed on-disk enhancement files via clear_refinement_data) left stale
good-trajectory memory behind — rescue and velocity bias would keep steering
toward the previous session's content. Add clear_velocity_bias_memory(key)
in samplers.py (drops global averages + prompt clusters for that key) and
call it from clear_refinement_data. Key matches: refine_v2 and the velocity
key default both resolve to the same effective refinement key.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Replaces the minimal RF rescue mirror with _sample_const_rf_full: an
RF-correct port of the hybrid sampler so its features actually run on CONST
models (LTXAV) instead of being swallowed by the euler_ancestral_RF fallback.
- early phase: ancestral RF euler (matches sample_euler_ancestral_RF) with
eta decay, AB2 order-2 denoised extrapolation, motion pulses
- quality phase: deterministic Heun corrector (RF-correct 2nd order; DPM++ 2S
is eps-only) with the progressive correction_blend schedule
- velocity-bias capture/apply + reactive rescue around every model eval
Active by default (matches hybrid defaults) per user request; changes the
LTXAV baseline away from plain euler-ancestral. to_d-based euler is identical
to the RF flow update, so early and quality phases stay consistent.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Root cause of "no rescue output": LTXAV uses CONST (rectified-flow)
sampling, so the hybrid sampler short-circuited to comfy's euler_ancestral
(which routes to euler_ancestral_RF) and skipped the entire custom loop —
velocity bias, motion pulses, and rescue never ran for LTXAV.
Replace the CONST early-return with _sample_const_rf_with_rescue: a faithful
mirror of comfy.sample_euler_ancestral_RF (same noise_scale, downstep/renoise
math) with velocity-bias capture/apply and reactive rescue inserted around
the model eval. With velocity and rescue off it reproduces the stock RF
result step-for-step. Hoist param normalization above the CONST branch so
both paths share it; correct the startup diagnostic accordingly.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
When rescue_mode is on, print at sampler entry: resolved strength/threshold,
whether a prompt signature arrived, whether the model uses CONST sampling
(which bypasses the whole custom loop), and which sigma steps are eligible
velocity targets. Makes it unambiguous why rescue does or does not fire —
no message at all now means the option never reached the hybrid sampler.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
The single velocity-target step (norm sigma ~0.9) lands inside the quality
phase on short distilled schedules (8 steps, high_quality_pct=0.35 ->
quality boundary at sigma 0.909, the only matching step). Capture and rescue
were coded only in the early/non-quality branch, so neither ever ran on the
default LTX schedule — no capture, no rescue, no log output. Hoist
velocity_target above the phase split and run capture+apply+rescue after the
quality-phase model eval too. Capture records raw denoised before rescue
edits it in both branches, so the good-trajectory memory stays clean.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
The velocity_refinement_key (namespace for trajectory capture + rescue
clusters) was a separate field defaulting to "default", easy to drift out
of sync with the refinement key Studio actually uses. Now when it is blank
or left at "default", Studio resolves it to the effective refinement key
(same resolution as seed memory / refine_v2), so capture and rescue share
one bucket automatically. An explicit non-default value is still respected.
JS hint added on the velocity key field.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Rescue previously steered toward a prompt-blind average of all captured
good trajectories, risking false positives that fight a legitimately
different prompt. Now the velocity memory keeps per-prompt clusters: each
tagged with a pooled+normalized conditioning signature. Capture merges into
the nearest cluster (cosine >= 0.92) or starts a new one (cap 8/key, evict
lowest-count). Rescue builds its reference from clusters similar to the
current prompt (cosine >= 0.5, >= 2 matched captures), similarity-weighted;
with no confident prompt match it skips rather than steering to the global
mean. Studio pools its positive conditioning into a signature and threads it
through _build_samplers -> get_sampler -> sampler extra_options. Standalone
node (no conditioning) falls back to the prior global-average behavior.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Studio builds its sampler via FunPackHybridEuler2SSampler.get_sampler but
did not pass the new rescue params. Thread rescue_mode/threshold/strength
through _build_one_sampler, add them to the Studio JS hybrid defaults
(high+low) and the sampler config panel, with a note that rescue needs the
velocity-bias memory captured first.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Detects when a denoising trajectory diverges from the learned
good-trajectory (velocity-bias) memory and rotates the prediction
back toward it on that step only. Reactive, read-only on the memory,
audio-safe (direction-only, magnitude-preserved on the denoised latent).
New sampler widgets: rescue_mode / rescue_threshold / rescue_strength,
threaded through get_sampler extra_options. Shares velocity_refinement_key
and aspect bucket with velocity bias; no-op until that memory is captured.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
New order: positive signals (broad→specific) → negative signals
(broad→specific) → delta cap → VF gate.
Concept-in-context delta moved before bad conditioning repulsion so
positive phrase signal isn't added after negatives are removed.
Pair repulsion stays last among negatives as most targeted.
VF final gate: after all patches + delta cap, if net effect lowered
VF predicted reward by >0.05, linearly roll back toward original
proportionally to how much worse it got (max 75% rollback). Visible
as vf-rollback:X.XX in axis adjustments.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
search() generates 16 random unit-direction perturbations (2.5% of
conditioning norm each), scores each with the VF, returns the best.
Runs after ascend() in _v2_ascend_conditioning — ascent exploits the
known gradient direction, search explores the neighbourhood around it
for any better nearby region. Total cost: 16 VF forward passes ≈ free.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
Learns which phrase combinations in the final encoded prompt correlate with
anatomy/quality failures (reward < -0.10, wrong_appearance, wrong_action+quality).
For each pair of phrases (up to top 8) present in a bad generation, accumulates
the conditioning direction as a per-pair bad slot using the existing
_v2_store_direction mechanism.
At generation time: for each phrase pair in the current final prompt with ≥3
observations, applies a repulsion from the stored bad direction (max 3 pairs,
strength capped at 3.0% of conditioning norm). Visible in axis_actions as
pair-repulsion:N. Prunes to 200 pairs when exceeding 300.
Trigger includes wrong_appearance explicitly (reward=0.0 but signals anatomy
artifacts), missing quality, and all outcomes below -0.10 reward.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
The freeze locked prompt text to the previous Perfect output, overriding
intentional prompt changes when intent similarity stayed above 0.50. With
VF + conditioning memory maintaining quality direction, the freeze is no
longer needed and only causes unwanted video repetition.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
Learns which phrases in the final encoded prompt (not intent, not raw input —
the actual text that went into CLIP) correlate with reward changes when added
or removed between consecutive generations.
Storage: concept_delta_memory per phrase text — add_sum/count, remove_sum/count.
Learning: on rating, diffs previous_run.final_phrase_texts vs prev_final_phrase_texts;
credits reward to added phrases, records it for removed.
Application: before conditioning memory is applied, computes a weighted direction
from phrase embeddings (via phrase_memory) for phrases new to this final prompt
that have ≥3 add observations and avg_reward > 0.10. Nudge capped at 3.5% of
conditioning norm alongside all other deltas.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
Used refinement_key_input/refinement_key directly instead of eff_key
which is assigned ~340 lines later in refine_v2.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
Two new VF-driven behaviours in _v2_apply_conditioning_memory:
1. Global confidence scalar — VF predicted reward scales all conditioning memory
modifications: confidence_scale = (vf_score+1)/2 clamped to [0.4, 1.0].
High score → full strength (direction is good). Low score → dampened
(be cautious, current conditioning direction looks bad). Applies regardless
of which prompt is active — purely VF-signal driven.
2. Gradient-aligned phrase boosting — VF gradient direction (∂reward/∂conditioning)
is computed and compared to each axis direction slot via cosine similarity.
Boost factor [0.5, 1.5] applied to direction strength: phrases/axes pointing
with the reward gradient are amplified, those pointing against are dampened.
Applies to liked_dir, missing axis repair, and satisfied axis preserve.
VF score and boost factors visible in adaptation status output.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
Attention weight accumulation:
- Capture per-token output norm at identity blocks [14,20,21,30,33] during every
generation via the optimized_attention patch (same path as temperature)
- EMA-merge into blessed file on liked ratings (0.8 old + 0.2 new)
- Inject as V-scaling bias on future runs: important tokens get amplified V,
pushing attention toward historically attended positions
- Temp/blessed files cleared on session reset
VF-driven temperature:
- build_enhancements accepts conditioning=None, loads value function, scores it
- _derive_temperature_map(vf_score=) continuously modulates block temperatures:
high VF score → sharpen (0.80×), low → warm/explore (1.20×), regardless of rating
- Rating-driven temperature still applies when reward < 0.8; VF blends with it
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
LTX packs all spatial/temporal dims into a single token [1, 1, C*T*H*W]
during sampling. Per-step latent steering requires hooking into the model
internals, not the k-diffusion callback. Removed: compress_latent(),
_load_latent_value_function(), latent VF training, motion floor callback,
latent_value_guidance toggle in Studio.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
