luke/ai-podcast
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

Website overhaul: nav, accessibility, shared components, SEO, Reaper silence detection

Browse Source 
Website:
- Add persistent top nav across all pages
- Add skip-to-content links, focus-visible styles, ARIA on audio player
- Fix text contrast for WCAG AA compliance
- Add 600px breakpoint, mobile typography scaling
- Extract shared footer.js, player.js, episode.js components
- Episode pagination (10 + Load More), featured clip dedup
- Worker meta injection for social crawler OG tags
- Unify Plausible analytics proxy across all pages
- Sanitize innerHTML for XSS safety
- Custom 404 page, enhanced llms.txt, fix sitemap
- Bump cache versions to v=4

Reaper:
- Add dual silence threshold: 2.5s for speaker transitions, 6s for same-speaker gaps

Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
This commit is contained in:
tcpsyn
2026-03-16 00:56:29 -06:00
parent c70f83d04a
commit d39cb3f3d4
20 changed files with 870 additions and 678 deletions
Download Patch File Download Diff File Expand all files Collapse all files
reaper/strip_silence_dialog.lua
+47 -9
View File
@@ -8,7 +8,8 @@
-- SETTINGS
---------------------------------------------------------------------------
local SILENCE_DB = -30 -- dBFS — anything below this is "silence"
local MIN_SILENCE_SEC = 6.0 -- only remove silences longer than this
local MIN_SILENCE_SEC = 6.0 -- same-speaker gaps: only remove silences longer than this
local MIN_SILENCE_TRANSITION_SEC = 2.5 -- cross-speaker gaps: shorter threshold for speaker transitions
local MIN_VOICE_SEC = 0.3 -- ignore non-silent bursts shorter than this (filters transients)
local KEEP_PAD_SEC = 0.5 -- leave this much silence on each side of a cut
local BLOCK_SEC = 0.1 -- analysis block size (100ms)
@@ -304,14 +305,34 @@ local function read_block_peak_rms(ta, project_time)
return 0, 0
end
-- find_loudest_track: returns 1-based index of the loudest track at a given time, or 0 if silent
local function find_loudest_track(track_audios, project_time)
local best_peak = 0
local best_idx = 0
for i, ta in ipairs(track_audios) do
local peak, _ = read_block_peak_rms(ta, project_time)
if peak > best_peak then
best_peak = peak
best_idx = i
end
end
if best_peak < THRESHOLD then return 0 end
return best_idx
end
-- find_silences: detects silences and accumulates RMS data
-- Tracks which track was active before/after each silence to distinguish
-- speaker transitions (short threshold) from same-speaker pauses (long threshold).
-- Yields periodically via coroutine for UI responsiveness
-- progress_fn(t): called before each yield with current position
local function find_silences(region, track_audios, rms_acc, progress_fn)
local silences = {}
local in_silence = false
local silence_start = 0
local track_before_silence = 0
local voice_run = 0
local voice_run_track = 0
local last_active_track = 0
local t = region.start_pos
local total_blocks = 0
local silent_blocks = 0
@@ -320,11 +341,13 @@ local function find_silences(region, track_audios, rms_acc, progress_fn)
while t < region.end_pos do
local best_peak = 0
local best_sum = 0
for _, ta in ipairs(track_audios) do
local best_track = 0
for i, ta in ipairs(track_audios) do
local peak, sum_sq = read_block_peak_rms(ta, t)
if peak > best_peak then
best_peak = peak
best_sum = sum_sq
best_track = i
end
end
@@ -332,31 +355,45 @@ local function find_silences(region, track_audios, rms_acc, progress_fn)
total_blocks = total_blocks + 1
if all_silent then silent_blocks = silent_blocks + 1 end
if not all_silent and rms_acc then
rms_acc.sum_sq = rms_acc.sum_sq + best_sum
rms_acc.count = rms_acc.count + BLOCK_SAMPLES
if not all_silent then
last_active_track = best_track
if rms_acc then
rms_acc.sum_sq = rms_acc.sum_sq + best_sum
rms_acc.count = rms_acc.count + BLOCK_SAMPLES
end
end
if in_silence then
if all_silent then
voice_run = 0
voice_run_track = 0
else
if voice_run == 0 then voice_run_track = best_track end
voice_run = voice_run + 1
if voice_run >= MIN_VOICE_BLOCKS then
local voice_start = t - (voice_run - 1) * BLOCK_SEC
local dur = voice_start - silence_start
if dur >= MIN_SILENCE_SEC then
table.insert(silences, {start_pos = silence_start, end_pos = voice_start, duration = dur})
local track_after = voice_run_track
local is_transition = track_before_silence ~= 0 and track_after ~= 0 and track_before_silence ~= track_after
local threshold = is_transition and MIN_SILENCE_TRANSITION_SEC or MIN_SILENCE_SEC
if dur >= threshold then
table.insert(silences, {
start_pos = silence_start, end_pos = voice_start, duration = dur,
is_transition = is_transition,
})
end
in_silence = false
voice_run = 0
voice_run_track = 0
end
end
else
if all_silent then
in_silence = true
silence_start = t
track_before_silence = last_active_track
voice_run = 0
voice_run_track = 0
end
end
@@ -419,7 +456,7 @@ local function phase1_strip_silence(dialog_regions)
end
log("Phase 1: Analyzing using " .. tracks_loaded .. "/" .. #CHECK_TRACKS .. " voice tracks")
log(" threshold=" .. SILENCE_DB .. "dB, min_silence=" .. MIN_SILENCE_SEC .. "s, pad=" .. KEEP_PAD_SEC .. "s")
log(" threshold=" .. SILENCE_DB .. "dB, min_silence=" .. MIN_SILENCE_SEC .. "s (same-speaker), " .. MIN_SILENCE_TRANSITION_SEC .. "s (transition), pad=" .. KEEP_PAD_SEC .. "s")
-- Calculate total duration for progress tracking
local total_duration = 0
@@ -463,7 +500,8 @@ local function phase1_strip_silence(dialog_regions)
end
if not protected then
table.insert(removals, {start_pos = rm_start, end_pos = rm_end})
log(" remove " .. string.format("%.1f", rm_end - rm_start) .. "s at " .. string.format("%.1f", s.start_pos) .. "-" .. string.format("%.1f", s.end_pos))
local tag = s.is_transition and " [transition]" or ""
log(" remove " .. string.format("%.1f", rm_end - rm_start) .. "s at " .. string.format("%.1f", s.start_pos) .. "-" .. string.format("%.1f", s.end_pos) .. tag)
end
end
end