Audio Prompt Guide: per-segment / per-time / global notes

The notes box now routes each line to a target: 'seg2: fast' (by number, also '2:'
/ 'S2:'), '10s-15s: drop' (by time range / '12.3s:'), or unprefixed = global. Segment
notes are attached to the matching segment in the summary (marked NOTE, overrides the
energy default) and drawn on the waveform image; time notes attach to overlapping
segments (strict end boundary); globals listed separately. README documents the syntax.

Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>

README.md

@@ -86,6 +86,15 @@ The model then gets the audio's timing/energy + your motion notes and writes bea
 camera moves, and deltas that escalate with the music. `librosa` is optional (BPM/beats);
 without it you still get the energy envelope + segments.
 
+**Notes can target a segment** (the note overrides that segment's energy default, and is drawn
+on the waveform). Syntax in the `notes` box, one per line:
+
+```
+seg1: slow dreamy intro          # by segment number (also "1:" or "S1:")
+10s-15s: explosive drop          # by time range (also "12.3s:")
+cinematic, moody grade           # no prefix = global (applies throughout)
+```
+
 ## Performance / speed
 
 This node runs models through **transformers `.generate()`** — the simplest path, but the

nodes/audio_guide.py

@@ -19,11 +19,49 @@ with it you also get tempo (BPM) and beat times.
 
 from __future__ import annotations
 
+import re
+
 import numpy as np
 import torch
 from PIL import Image, ImageDraw
 
 
+def _split_notes(notes: str):
+    """Parse the notes box into per-segment / per-time / global notes. Lines like:
+        seg2: fast motion        (or  2: fast motion  /  S2: ...)   -> segment 2
+        5s: build                (or  0-5s: ...  /  12.3s: ...)     -> by time
+        cinematic, moody          (no prefix)                       -> global
+    Returns (seg_map {num: [notes]}, time_notes [(t0,t1,text)], global_notes [str])."""
+    seg_map, time_notes, glob = {}, [], []
+    for raw in notes.splitlines():
+        line = raw.strip()
+        if not line:
+            continue
+        mt = re.match(r"(?i)^(\d+(?:\.\d+)?)\s*s(?:\s*[-–to]+\s*(\d+(?:\.\d+)?)\s*s?)?\s*[:)\-]\s*(.+)$", line)
+        ms = (re.match(r"(?i)^(?:seg(?:ment)?|s)\s*(\d+)\s*[:)\-]\s*(.+)$", line)
+              or re.match(r"^(\d+)\s*[:)]\s*(.+)$", line))
+        if mt:
+            t0 = float(mt.group(1)); t1 = float(mt.group(2)) if mt.group(2) else t0
+            time_notes.append((t0, t1, mt.group(3).strip()))
+        elif ms:
+            seg_map.setdefault(int(ms.group(1)), []).append(ms.group(2).strip())
+        else:
+            glob.append(line)
+    return seg_map, time_notes, glob
+
+
+def _attach_notes(segs, notes):
+    """Attach the parsed notes to each segment (by number or overlapping time).
+    Returns the leftover global notes."""
+    seg_map, time_notes, glob = _split_notes(notes)
+    for s in segs:
+        s0, s1 = s["start_s"], s["start_s"] + s["duration_s"]
+        found = list(seg_map.get(s["segment"], []))
+        found += [txt for (t0, t1, txt) in time_notes if t0 < s1 and t1 > s0]
+        s["note"] = "; ".join(found)
+    return glob
+
+
 def _to_mono(audio) -> tuple[np.ndarray, int]:
     """ComfyUI AUDIO dict -> (mono float32 samples, sample_rate)."""
     wf = audio["waveform"]
@@ -108,12 +146,15 @@ def _render(rms_n, times, duration, beats, bounds, segs):
     for bd in bounds:                                 # segment boundaries (white)
         d.line([(X(bd), 0), (X(bd), H)], fill=(240, 240, 240), width=1)
     for s in segs:
-        d.text((X(s["start_s"]) + 4, 4), f"S{s['segment']} {s['energy']}", fill=(255, 255, 255))
+        x = X(s["start_s"]) + 4
+        d.text((x, 4), f"S{s['segment']} {s['energy']}", fill=(255, 255, 255))
+        if s.get("note"):                             # per-segment note (amber)
+            d.text((x, 18), s["note"][:30], fill=(255, 210, 110))
     arr = np.asarray(img, dtype=np.float32) / 255.0
     return torch.from_numpy(arr)[None, ...]           # [1, H, W, 3]
 
 
-def _summary(duration, sr, bpm, beats, segs, notes):
+def _summary(duration, sr, bpm, beats, segs, global_notes):
     lines = ["AUDIO GUIDE — align the video beats to this audio:",
              f"- duration: {duration:.2f}s  |  sample_rate: {sr} Hz"]
     if bpm:
@@ -121,21 +162,23 @@ def _summary(duration, sr, bpm, beats, segs, notes):
     if beats:
         preview = ", ".join(f"{b:.2f}" for b in beats[:24])
         lines.append(f"- beat times (s): {preview}{' ...' if len(beats) > 24 else ''}")
-    lines.append("- suggested segments (use these durations/frames and energy):")
+    lines.append("- suggested segments (use these durations/frames and energy; honor the NOTE):")
     for s in segs:
         peak = f", energy peak @ {s['peak_s']}s" if s["energy"] == "high" else ""
+        note = f"  <<< NOTE: {s['note']}" if s.get("note") else ""
         lines.append(
             f"    seg{s['segment']}: {s['start_s']}–{s['start_s'] + s['duration_s']:.2f}s, "
             f"{s['duration_s']}s, {s['frames']} frames, energy {s['energy'].upper()} "
-            f"({s['stage_hint']}){peak}")
+            f"({s['stage_hint']}){peak}{note}")
     lines.append("")
     lines.append("Guidance: higher energy -> faster motion and bigger camera moves; "
                  "lower energy -> slower, settle. Put escalation on rising energy and the "
-                 "release on the final segment. Snap each beat's frames to 8n+1.")
+                 "release on the final segment. Snap each beat's frames to 8n+1. Where a "
-    if notes.strip():
+                 "segment has a NOTE, that instruction OVERRIDES the energy default.")
+    if global_notes:
         lines.append("")
-        lines.append("USER NOTES (map these onto the matching segments/times):")
+        lines.append("GLOBAL NOTES (apply throughout):")
-        lines.append(notes.strip())
+        lines.extend(f"  - {n}" for n in global_notes)
     return "\n".join(lines)
 
 
@@ -162,8 +205,9 @@ class AudioPromptGuide:
         rms_n, times = _rms_envelope(y, sr)
         bpm, beats = _tempo_beats(y, sr)
         segs, bounds = _segments(rms_n, times, duration, fps, max_segments, beats)
+        global_notes = _attach_notes(segs, notes)     # per-segment / per-time / global
         image = _render(rms_n, times, duration, beats, bounds, segs)
-        summary = _summary(duration, sr, bpm, beats, segs, notes)
+        summary = _summary(duration, sr, bpm, beats, segs, global_notes)
         return (image, summary)