midiviz: the focus column holds instead of breathing; the wordmark sharpens

PLN: 'its weird moving a djf makes the col grow then shrink over time' —
width on a decay curve was motion with no meaning; the played column now
HOLDS 2x while in use, hands cause the only transitions, uniform after
FOCUS_HOLD of silence. Wordmark: two tight glow rings instead of a wide
fog, tighter tracking, condense floored at 0.75 — nine ExtraBold glyphs
are ~9.4em and three columns ~3.9em, so the rest goes to height. oscfeed
says 'first trigger received (dN)' once, so 'is Tidal even sending' is a
journalctl away.

tools/bridge/midiviz.py

@@ -484,6 +484,7 @@ WAVE_ALPHA = 0.12
 # vendored out of the built site for the same standalone reason as the wave.
 SYNE_PATH = Path(__file__).resolve().parent / "ui" / "syne-extrabold.ttf"
 WORD_ALPHA = 0.8
+WORD_CONDENSE_FLOOR = 0.75          # never squash the bloc thinner than this
 
 # ── the HL feed ────────────────────────────────────────────────────────────
 # PLN, 2026-09-24: "please do indeed sub from the events, to indeed feed
@@ -494,6 +495,8 @@ WORD_ALPHA = 0.8
 # default at runtime (--no-hl for the off switch): a lens that only shows
 # the controls was half the picture.
 HL_ON_BY_DEFAULT = True
+# How long the focus column keeps its 2× after the last touch of it.
+FOCUS_HOLD = 4.0
 
 # ── the size belongs to the window manager, the density to the window ──────
 # PLN, 2026-09-22: "only drag-> grab atm, can we have corner grabs to make it
@@ -1706,12 +1709,17 @@ def build_widget(port_label: str, reader: "Reader | None", scale: float = 1.0,
             return self._wave_pm
 
         def _wave_font(self, target_w: float, target_h: float):
-            """Syne ExtraBold, letter-spaced like the site's hero (0.15em),
-            sized to fit the D1-D4 span exactly. The vendored ttf is the
-            variable font INSTANCED at 800, not the family default — Qt
-            resolves a variable font's default instance (Regular) however
-            loudly you ask for a weight, which is how the first cut read as
-            thin as a system sans."""
+            """Syne ExtraBold, sized to fill the row height as far as the
+            span allows at a 0.75 condense floor — PLN, 2026-09-24: "letters
+            could be a bit taller, like atm they occupy flattened rectangles
+            boxes, can be more square?" The arithmetic is honest: nine
+            ExtraBold Syne glyphs are ~9.4 em wide, three columns ~3.9 em, so
+            square letters would need either a 0.42× squash (its own
+            distortion) or a wider span. The compromise: condense at most to
+            0.75, and let the type take whatever height still fits. The
+            vendored ttf is the variable font INSTANCED at 800 — Qt resolves
+            a variable font's default instance (Regular) however loudly you
+            ask for a weight, which is how the first cut read thin."""
             key = (round(target_w), round(target_h))
             if self._word_font_key == key:
                 return self._word_font
@@ -1719,26 +1727,34 @@ def build_widget(port_label: str, reader: "Reader | None", scale: float = 1.0,
             if fam_id < 0:
                 return None
             fam = QtGui.QFontDatabase.applicationFontFamilies(fam_id)[0]
+            probe = QtGui.QFont(fam)
+            probe.setPixelSize(40)
+            probe.setLetterSpacing(QtGui.QFont.SpacingType.PercentageSpacing,
+                                   106.0)          # +0.06em — tighter tracking
+            fmr = QtGui.QFontMetricsF(probe)
+            word = "PARVAGUES"
+            em_ratio = (sum(fmr.horizontalAdvance(c) for c in word)
+                        + fmr.horizontalAdvance(" ") * 0.06 * (len(word) - 1)
+                        ) / 40.0
+            px = min(target_h * 0.92,
+                     target_w / (WORD_CONDENSE_FLOOR * em_ratio))
             f = QtGui.QFont(fam)
-            f.setPixelSize(max(6, target_h * 0.62))
+            f.setPixelSize(max(6, px))
             f.setLetterSpacing(QtGui.QFont.SpacingType.PercentageSpacing,
-                               115.0)               # +0.15em, as tracked
-            fm = QtGui.QFontMetricsF(f)
-            word = "PARVAGUES"
-            total = sum(fm.horizontalAdvance(c) for c in word) \
-                + fm.horizontalAdvance(" ") * 0.15 * (len(word) - 1)
-            f.setPixelSize(max(6, target_h * 0.62
-                               * target_w / max(1.0, total)))
+                               106.0)
             self._word_font, self._word_font_key = f, key
             return f
 
         def _word_pixmap(self, span_w: float, span_h: float):
             """The wordmark BAKED — glyphs, tracking and neon glow rendered
             once per geometry, not per frame. The glow is the site hero's
-            `textShadow: 0 0 60px rgba(217,0,255,0.15)` in miniature: rings
-            of translucent magenta under a solid core. A per-frame fake of
-            this would cost two dozen drawText calls; baked, the frame pays
-            one blit."""
+            `textShadow` in miniature, but TIGHT — PLN: "a little less
+            blurry? esp made fullscreen it gets very blurry" — because a wide
+            ring halo at 60px of type is indistinguishable from an out-of-
+            focus lens. Two close rings only; the core carries the colour.
+            Width overrun is taken by a vector condense (a scaled painter,
+            still antialiased), never by raster scaling — so fullscreen
+            rebakes stay sharp."""
             key = (round(span_w), round(span_h))
             if self._word_pm_key == key:
                 return self._word_pm
@@ -1747,10 +1763,14 @@ def build_widget(port_label: str, reader: "Reader | None", scale: float = 1.0,
                 return None
             fm = QtGui.QFontMetricsF(f)
             word = "PARVAGUES"
-            total = sum(fm.horizontalAdvance(c) for c in word) \
-                + fm.horizontalAdvance(" ") * 0.15 * (len(word) - 1)
-            pad = max(2, int(f.pixelSize() * 0.35))
-            img = QtGui.QImage(round(total) + pad * 2, round(span_h) + pad * 2,
+            def measure():
+                return sum(fm.horizontalAdvance(c) for c in word) \
+                    + fm.horizontalAdvance(" ") * 0.06 * (len(word) - 1)
+            total = measure()
+            sx = min(1.0, span_w / max(1.0, total))   # the condense factor
+            pad = max(2, int(f.pixelSize() * 0.22))
+            img = QtGui.QImage(round(span_w * sx) + pad * 2,
+                               round(span_h) + pad * 2,
                                QtGui.QImage.Format.Format_ARGB32_Premultiplied)
             img.fill(0)
             qp = QtGui.QPainter(img)
@@ -1758,25 +1778,23 @@ def build_widget(port_label: str, reader: "Reader | None", scale: float = 1.0,
                 qp.setRenderHint(QtGui.QPainter.RenderHint.TextAntialiasing,
                                  True)
                 qp.setFont(f)
+                qp.scale(sx, 1.0)
                 asc = fm.ascent()
                 base = pad + (span_h + asc - fm.descent()) / 2.0
-                x0 = pad
                 def draw_word(pen):
                     qp.setPen(pen)
-                    x = x0
+                    x = float(pad)
                     for c in word:
                         qp.drawText(QtCore.QPointF(x, base), c)
                         x += fm.horizontalAdvance(c) \
-                            + fm.horizontalAdvance(" ") * 0.15
-                # the glow: three rings of translucent neon, widest first —
-                # the cheap cousin of a blur, and per-frame it is free
-                r0 = max(1.0, f.pixelSize() * 0.16)
-                for rad, alpha in ((r0 * 1.8, 28), (r0, 44), (r0 * 0.5, 70)):
+                            + fm.horizontalAdvance(" ") * 0.06
+                # the glow: two CLOSE rings, faint — a rim of neon, not a fog
+                r0 = max(1.0, f.pixelSize() * 0.09)
+                for rad, alpha in ((r0, 40), (r0 * 0.45, 60)):
                     glow = QtGui.QPen(QtGui.QColor(217, 0, 255, alpha))
-                    glow.setWidthF(max(2.0, f.pixelSize() * 0.10))
-                    n = 12
-                    for i in range(n):
-                        a = 2 * math.pi * i / n
+                    glow.setWidthF(max(1.5, f.pixelSize() * 0.07))
+                    for i in range(10):
+                        a = 2 * math.pi * i / 10
                         qp.save()
                         qp.translate(math.cos(a) * rad, math.sin(a) * rad)
                         draw_word(glow)
@@ -2168,26 +2186,27 @@ def build_widget(port_label: str, reader: "Reader | None", scale: float = 1.0,
                            self.lut[FAM_FX2][14])
 
         def _col_geo(self, now):
-            """Per-column (x, width) with a recency-weighted focus column.
+            """Per-column (x, width) with a STEADY focus column.
 
             PLN, 2026-09-24: "can we have 2x wider matrix column, shrinking
             slightly other cols, and algorithmically minimize overlaps when
-            moving several controls together, for readability". The algorithm
-            reuses the column glow's OWN decay as the focus weight: width
-            w(col) = 1 + charge(col), charge = the TAU_COL level normalized to
-            0..1. Full charge is exactly 2× a cold column, the others ease
-            narrower to pay for it, several columns moving together share the
-            extra room smoothly (no snaps — the widths ride the same decay the
-            glow does), and the resting matrix is perfectly uniform again.
-            Total width is invariant, so nothing outside the matrix moves.
+            moving several controls together, for readability" — then, of the
+            first cut's decay-driven widths: "its weird moving a djf makes
+            the e.g. col1 grow then shrink over time?" He was right: a width
+            that breathes on a decay curve is motion with no meaning. Now the
+            most recently touched column HOLDS 2× while it is in use and the
+            others give up a little width; the focus switches when another
+            column is touched and everything relaxes to uniform after
+            FOCUS_HOLD of silence. The only transitions are caused by hands.
             """
-            weights = []
+            focus, last = None, 0.0
             for col in range(1, 9):
-                if self.col_t[col]:
-                    lv = self._lv(now - self.col_t[col], tau=TAU_COL)
-                    weights.append(1.0 + lv / (LEVELS - 1))
-                else:
-                    weights.append(1.0)
+                if self.col_t[col] > last:
+                    focus, last = col, self.col_t[col]
+            if focus is None or now - last > FOCUS_HOLD:
+                weights = [1.0] * 8
+            else:
+                weights = [2.0 if c == focus else 1.0 for c in range(1, 9)]
             total = self.mw_px
             scale = total / sum(weights)
             out, x = [], self.mx

tools/bridge/oscfeed.py

@@ -19,6 +19,7 @@ from __future__ import annotations
 
 import socket
 import struct
+import sys
 import threading
 import time
 from collections import deque
@@ -182,6 +183,12 @@ class HLFeed:
                     self._q.append({"orbit": o, "t": now,
                                     "params": msg["args"]})
                     self.total += 1
+                    first = self.total == 1
+            if first:
+                # One line, once: the question "is Tidal even sending?" is
+                # answered by the journal, not by staring at a silent grid.
+                print("⚓ HL feed: first trigger received (d%d)" % o,
+                      file=sys.stderr, flush=True)
 
     def alive(self) -> bool:
         return self._sock is not None and self.error is None