refactor: fix peer review findings in audio, encoder, renderer, and CLI

.github/ISSUE_TEMPLATE/config.yml

@@ -1,7 +1,7 @@
 blank_issues_enabled: false
 contact_links:
   - name: 🛟 Support
-    url: https://github.com/linuxmatters/jivefire/blob/HEAD/SUPPORT.md
+    url: https://github.com/linuxmatters/jive-visualiser/blob/HEAD/SUPPORT.md
     about: Check the support guide before opening an issue.
 #  - name: Code of Conduct Report
 #    url: https://yourproject.org/community-report/

.github/workflows/builder.yml

@@ -32,11 +32,11 @@ jobs:
       - run: go vet ./...
       - name: Check cyclomatic complexity
         run: |
-          go install github.com/fzipp/gocyclo/cmd/gocyclo@latest
+          go install github.com/fzipp/gocyclo/cmd/gocyclo@v0.6.0
           gocyclo -top 20 -ignore '_test\.go$' -avg .
       - name: Check ineffectual assignments
         run: |
-          go install github.com/gordonklaus/ineffassign@latest
+          go install github.com/gordonklaus/ineffassign@v0.2.0
           ineffassign ./...
       - uses: golangci/golangci-lint-action@v9
 
@@ -139,7 +139,7 @@ jobs:
           comment-summary-in-pr: on-failure
 
       - name: Run govulncheck
-        uses: golang/govulncheck-action@master
+        uses: golang/govulncheck-action@v1.0.4
         with:
           repo-checkout: false
           go-version-file: go.mod
@@ -264,10 +264,6 @@ jobs:
           chmod +x jive-visualiser-darwin-arm64
           sudo mv jive-visualiser-darwin-arm64 /usr/local/bin/jive-visualiser
           ```
-
-          ## Checksums
-
-          SHA256 checksums are provided below for verification.
           NOTES
 
           cat CHANGELOG.md

AGENTS.md

@@ -19,20 +19,23 @@
 - Memory-efficient: ~50MB footprint for 30-minute audio vs 600MB for single-pass
 
 ### Key Modules
-- `cmd/jive-visualiser/main.go` — CLI entry, 2-pass coordinator
+- `cmd/jive-visualiser/main.go` - CLI entry and Pass 1 coordinator
+- `cmd/jive-visualiser/pass2.go` - Pass 2 rendering and encoding runner
 - `internal/audio/` — `StreamingReader` (reader.go) chunk-based decode, FFT analysis
-- `internal/encoder/` — ffmpeg-statigo wrapper, RGB→YUV conversion, FIFO buffer
-- `internal/yuv/` — YCbCr coefficients, `RGBToY`/`RGBToCb`/`RGBToCr`, `ParallelRows`
+- `internal/encoder/` - ffmpeg-statigo muxer facade, video/audio helpers, RGB→YUV conversion, FIFO buffer
+- `internal/yuv/` - YCbCr coefficients, `RGBToY`/`RGBToCb`/`RGBToCr`, `ParallelRows`, `RowPool`
 - `internal/renderer/` — Frame generation, bar drawing, thumbnail
-- `internal/ui/` — Bubbletea v2 TUI (unified progress.go for both passes)
+- `internal/ui/` - Bubbletea v2 TUI state, spectrum, preview, and summary views
 - `internal/config/` — Constants (dimensions, FFT params, colours)
+- `internal/theme/` - Terminal colour theme
+- `internal/cli/` - Kong CLI helpers and styled help
 
 ## FFmpeg Integration
 
 - All FFmpeg access through `third_party/ffmpeg-statigo` submodule (FFmpeg 8.0 static bindings)
 - `*.gen.go` files in submodule are auto-generated — do not edit
 - Audio decoding: `internal/audio/reader.go` — `NewStreamingReader` returns `*StreamingReader`
-- Video/audio encoding: `internal/encoder/encoder.go` wraps libx264/AAC
+- Video/audio encoding: `internal/encoder/encoder.go` exposes the public facade; `audio_encoder.go` and `video_encoder.go` own codec setup
 
 ## Charm TUI (v2)
 
@@ -47,13 +50,13 @@
 ## Audio Processing
 
 - FFT size: 2048 samples (Hanning window)
-- 64 frequency bars with log-scale binning
+- 64 frequency bars with linear (uniform) frequency binning; logarithmic scaling applies to amplitude only
 - Harmonica spring peak-hold bar dynamics: each bar rises instantly to any new peak, then springs back toward the live level. Spring params: frequency `6.0`, damping `1.0`, delta `1/FPS`, gain `2.0` (replaces the amplitude lift the old CAVA integrator provided)
-- Audio frame size mismatch handled by FFmpeg's `AVAudioFifo` (in `internal/encoder/encoder.go`; FFT needs 2048, AAC expects 1024)
+- Audio frame size mismatch handled by FFmpeg's `AVAudioFifo` (in `internal/encoder/audio_encoder.go`; FFT needs 2048, AAC expects 1024)
 
 ## Performance Patterns
 
-- RGB→YUV conversion in `encoder/frame.go` parallelised across CPU cores via `yuv.ParallelRows` (8.4× faster than swscale)
+- RGB→YUV conversion in `encoder/frame.go` parallelised across CPU cores via `yuv.RowPool` (13.2× faster than swscale)
 - `convertRGBAToYUV` (YUV420P) and `convertRGBAToNV12` (NV12) in `encoder/frame.go` are intentionally kept as separate functions despite near-identical structure — the hot-path duplication avoids a callback/interface indirection that would hurt throughput; do not refactor into a shared helper (shared low-level primitives live in `internal/yuv`)
 - Frame rendering uses symmetric mirroring (draw 1/4 pixels, mirror 3×)
 - Pre-computed intensity/colour tables in `renderer/frame.go`
@@ -84,14 +87,19 @@
 - Gradient/alpha tables: pre-computed in `NewFrame()`
 
 ### Changing UI output
-- Unified progress UI: `internal/ui/progress.go` (handles both passes)
+- Progress UI state: `internal/ui/progress.go` (handles both passes)
+- Spectrum and completion helpers: `internal/ui/spectrum.go`, `internal/ui/summary.go`
 - Message types: `AnalysisProgress`, `AnalysisComplete`, `RenderProgress`, `RenderComplete`
 - Audio profile display persists from Pass 1 through Pass 2
 - Video preview: `internal/ui/preview.go`
 
+## Releases
+
+- Do not add checksum or hash generation to the release workflow. GitHub shows SHA256 digests for release assets by default.
+
 ## Environment
 
 - NixOS development shell via `flake.nix`
 - Fish shell for terminal commands
 - CGO required (`CGO_ENABLED=1` in build)
-- Go 1.24.0 minimum
+- Go 1.26 minimum

README.md

@@ -14,14 +14,14 @@ Your podcast audio deserves more than a static image on YouTube. Jive Visualiser
 
 - 🖼️ **Thumbnail generator** YouTube-style PNG with your title, saved alongside the video
 - 🎬 **1280×720 @ 30fps** H.264/AAC YouTube-ready MP4, no questions asked
-  - 🎚️ **64 frequency bars** that actually look discrete (not that smeared spectrum nonsense)
+  - 🎚️ **64 frequency bars** that look discrete (not that smeared spectrum nonsense)
   - 🪞 **Symmetric mirroring** above and below centre, doubles the visual impact
-  - 🔬 **FFT-based analysis** 2048-point Hanning window, log scale frequency binning
+  - 🔬 **FFT-based analysis** 2048-point Hanning window, linear frequency binning, log-scaled amplitude
   - ✨ **Spring-driven bar dynamics** bars snap up instantly, spring back down via harmonica peak-hold
 - 🚀 **Stupidly fast** streaming pipeline, parallel RGB→YUV conversion
   - ⚡ **GPU acceleration** auto-detected: NVENC, Vulkan, VA-API, QuickSync, VideoToolbox
 - 📦 **Single binary** No Python. No FFmpeg install required. Just drop and render
-  - 🐧 **Linux** (amd64 and aarch64)
+  - 🐧 **Linux** (amd64 and AArch64)
   - 🍏 **macOS** (x86 and Apple Silicon)
 
 ## Usage
@@ -36,7 +36,7 @@ Your podcast audio deserves more than a static image on YouTube. Jive Visualiser
 ./jive-visualiser --episode=42 --title="Linux Matters" input.wav output.mp4
 ```
 
-### Without Episode Number (unnumbered audio)
+### Without Episode Number (Unnumbered Audio)
 ```bash
 ./jive-visualiser --title="Linux Matters" input.wav output.mp4
 ```
@@ -53,7 +53,7 @@ Your podcast audio deserves more than a static image on YouTube. Jive Visualiser
 
 ## Build
 
-Jive Visualiser uses [ffmpeg-statigo](https://github.com/linuxmatters/ffmpeg-statigo) for FFmpeg static bindings.
+Jive Visualiser uses [FFmpeg statigo](https://github.com/linuxmatters/ffmpeg-statigo) for FFmpeg static bindings.
 
 ```bash
 # Setup or update ffmpeg-statigo submodule and library
@@ -69,6 +69,6 @@ just test-encoder # Test encoder
 
 FFmpeg's audio visualisation filters (`showfreqs`, `showspectrum`) render continuous frequency spectra, not discrete bars. No amount of FFmpeg filter chain kung-fu can achieve the discrete 64-bar aesthetic required for Linux Matters branding. Solution: Do the FFT analysis and bar rendering in Go, pipe frames to FFmpeg for encoding.
 
-**Why Go over Python?** The original `djfun/audio-visualizer-python` tool is a moribund Qt5 GUI with significant tech debt. For our podcast production needs we wanted multi-archtitecture tools that's that can integrate into automation pipelines.
+**Why Go over Python?** The original `djfun/audio-visualizer-python` tool is a moribund Qt5 GUI with significant tech debt. For our podcast production needs we wanted multi-architecture tools that can integrate into automation pipelines.
 
-The Jive Visualiser architecture, such as it is, is available in the [ARCHITECTURE.md](docs/ARCHITECTURE.md) document.
+The Jive Visualiser architecture, such as it is, is available in the [architecture document](docs/ARCHITECTURE.md).

SECURITY.md

@@ -8,7 +8,7 @@ Only the latest release is supported. Security fixes are released as patch versi
 
 **Do not open a public issue for security vulnerabilities.**
 
-[Report vulnerabilities privately](https://github.com/linuxmatters/jivefire/security/advisories/new). Include:
+[Report vulnerabilities privately](https://github.com/linuxmatters/jive-visualiser/security/advisories/new). Include:
 
 - Steps to reproduce
 - Affected versions