.refine audio quality analyzer

This commit is contained in:
afkarxyz
2026-03-25 17:39:43 +07:00
parent 0e6b6f9d39
commit 528bf65771
6 changed files with 871 additions and 339 deletions
+47
View File
@@ -809,6 +809,53 @@ func (a *App) AnalyzeTrack(filePath string) (string, error) {
return string(jsonData), nil
}
func (a *App) AnalyzeSpectrumWithParams(filePath string, fftSize int, windowFunction string) (string, error) {
if filePath == "" {
return "", fmt.Errorf("file path is required")
}
params := backend.SpectrumParams{
FFTSize: fftSize,
WindowFunction: windowFunction,
}
result, err := backend.AnalyzeSpectrumWithParams(filePath, params)
if err != nil {
return "", fmt.Errorf("failed to analyze spectrum: %v", err)
}
jsonData, err := json.Marshal(result)
if err != nil {
return "", fmt.Errorf("failed to encode response: %v", err)
}
return string(jsonData), nil
}
func (a *App) SaveSpectrumImage(audioFilePath string, base64Data string) (string, error) {
if audioFilePath == "" || base64Data == "" {
return "", fmt.Errorf("file path and image data are required")
}
base64Data = strings.TrimPrefix(base64Data, "data:image/png;base64,")
data, err := base64.StdEncoding.DecodeString(base64Data)
if err != nil {
return "", fmt.Errorf("failed to decode base64 image: %v", err)
}
ext := filepath.Ext(audioFilePath)
baseName := strings.TrimSuffix(filepath.Base(audioFilePath), ext)
outPath := filepath.Join(filepath.Dir(audioFilePath), baseName+".png")
err = os.WriteFile(outPath, data, 0644)
if err != nil {
return "", fmt.Errorf("failed to save image to disk: %v", err)
}
return outPath, nil
}
func (a *App) AnalyzeMultipleTracks(filePaths []string) (string, error) {
if len(filePaths) == 0 {
return "", fmt.Errorf("at least one file path is required")
+50 -9
View File
@@ -21,8 +21,23 @@ type TimeSlice struct {
Magnitudes []float64 `json:"magnitudes"`
}
func AnalyzeSpectrum(filepath string) (*SpectrumData, error) {
type SpectrumParams struct {
FFTSize int `json:"fft_size"`
WindowFunction string `json:"window_function"`
}
func DefaultSpectrumParams() SpectrumParams {
return SpectrumParams{
FFTSize: 4096,
WindowFunction: "hann",
}
}
func AnalyzeSpectrum(filepath string) (*SpectrumData, error) {
return AnalyzeSpectrumWithParams(filepath, DefaultSpectrumParams())
}
func AnalyzeSpectrumWithParams(filepath string, params SpectrumParams) (*SpectrumData, error) {
stream, err := flac.ParseFile(filepath)
if err != nil {
return nil, fmt.Errorf("failed to parse FLAC: %w", err)
@@ -42,7 +57,20 @@ func AnalyzeSpectrum(filepath string) (*SpectrumData, error) {
return nil, fmt.Errorf("no audio samples found")
}
return calculateSpectrum(samples, sampleRate), nil
fftSize := params.FFTSize
validSizes := []int{512, 1024, 2048, 4096, 8192}
valid := false
for _, s := range validSizes {
if fftSize == s {
valid = true
break
}
}
if !valid {
fftSize = 4096
}
return calculateSpectrumWithParams(samples, sampleRate, fftSize, params.WindowFunction), nil
}
func readSamples(stream *flac.Stream, channels int) ([]float64, error) {
@@ -75,8 +103,7 @@ func readSamples(stream *flac.Stream, channels int) ([]float64, error) {
return allSamples, nil
}
func calculateSpectrum(samples []float64, sampleRate int) *SpectrumData {
fftSize := 8192
func calculateSpectrumWithParams(samples []float64, sampleRate, fftSize int, windowFunc string) *SpectrumData {
numTimeSlices := 300
duration := float64(len(samples)) / float64(sampleRate)
@@ -98,8 +125,7 @@ func calculateSpectrum(samples []float64, sampleRate int) *SpectrumData {
}
window := samples[startIdx : startIdx+fftSize]
windowedSamples := applyHannWindow(window)
windowedSamples := applyWindow(window, windowFunc)
spectrum := fft(windowedSamples)
@@ -129,18 +155,33 @@ func calculateSpectrum(samples []float64, sampleRate int) *SpectrumData {
}
}
func applyHannWindow(samples []float64) []float64 {
func applyWindow(samples []float64, windowType string) []float64 {
n := len(samples)
windowed := make([]float64, n)
for i := 0; i < n; i++ {
window := 0.5 * (1.0 - math.Cos(2.0*math.Pi*float64(i)/float64(n-1)))
windowed[i] = samples[i] * window
var w float64
switch windowType {
case "hamming":
w = 0.54 - 0.46*math.Cos(2*math.Pi*float64(i)/float64(n-1))
case "blackman":
w = 0.42 - 0.5*math.Cos(2*math.Pi*float64(i)/float64(n-1)) +
0.08*math.Cos(4*math.Pi*float64(i)/float64(n-1))
case "rectangular":
w = 1.0
default:
w = 0.5 * (1.0 - math.Cos(2*math.Pi*float64(i)/float64(n-1)))
}
windowed[i] = samples[i] * w
}
return windowed
}
func applyHannWindow(samples []float64) []float64 {
return applyWindow(samples, "hann")
}
func fft(samples []float64) []complex128 {
n := len(samples)
+103 -58
View File
@@ -1,8 +1,9 @@
import { Card, CardContent, CardHeader } from "@/components/ui/card";
import { Spinner } from "@/components/ui/spinner";
import { Button } from "@/components/ui/button";
import { Activity, Waves, Radio, TrendingUp, FileAudio, Clock, Gauge, HardDrive } from "lucide-react";
import { Activity } from "lucide-react";
import type { AnalysisResult } from "@/types/api";
interface AudioAnalysisProps {
result: AnalysisResult | null;
analyzing: boolean;
@@ -10,19 +11,24 @@ interface AudioAnalysisProps {
showAnalyzeButton?: boolean;
filePath?: string;
}
export function AudioAnalysis({ result, analyzing, onAnalyze, showAnalyzeButton = true, filePath }: AudioAnalysisProps) {
if (analyzing) {
return (<Card>
return (
<Card>
<CardContent className="px-6">
<div className="flex items-center justify-center py-8 gap-3">
<Spinner />
<span className="text-muted-foreground">Analyzing audio quality...</span>
</div>
</CardContent>
</Card>);
</Card>
);
}
if (!result && showAnalyzeButton) {
return (<Card>
return (
<Card>
<CardContent className="px-6">
<div className="flex flex-col items-center justify-center py-8 gap-4">
<Activity className="h-12 w-12 text-primary" />
@@ -32,94 +38,133 @@ export function AudioAnalysis({ result, analyzing, onAnalyze, showAnalyzeButton
Verify the true lossless quality of downloaded files
</p>
</div>
{onAnalyze && (<Button onClick={onAnalyze}>
{onAnalyze && (
<Button onClick={onAnalyze}>
<Activity className="h-4 w-4" />
Analyze Audio
</Button>)}
</Button>
)}
</div>
</CardContent>
</Card>);
</Card>
);
}
if (!result) {
return null;
}
const formatDuration = (seconds: number) => {
const mins = Math.floor(seconds / 60);
const secs = Math.floor(seconds % 60);
return `${mins}:${secs.toString().padStart(2, '0')}`;
return `${mins}:${secs.toString().padStart(2, "0")}`;
};
const formatNumber = (num: number) => {
return num.toFixed(2);
};
const formatFileSize = (bytes: number): string => {
if (bytes === 0)
return "0 B";
if (bytes === 0) return "0 B";
const k = 1024;
const sizes = ["B", "KB", "MB", "GB"];
const i = Math.floor(Math.log(bytes) / Math.log(k));
return parseFloat((bytes / Math.pow(k, i)).toFixed(1)) + " " + sizes[i];
};
const nyquistFreq = result.sample_rate / 2;
return (<Card className="gap-2">
<CardHeader>
{filePath && (<p className="text-sm font-mono break-all">{filePath}</p>)}
return (
<Card className="gap-2">
<CardHeader className="pb-2">
{filePath && (
<p className="text-sm font-mono break-all text-muted-foreground">{filePath}</p>
)}
</CardHeader>
<CardContent className="space-y-2">
<div className="flex flex-wrap items-center gap-x-3 gap-y-1 text-xs">
<div className="flex items-center gap-1">
<Radio className="h-3 w-3 text-muted-foreground"/>
<CardContent>
<div className="grid grid-cols-1 md:grid-cols-2 lg:grid-cols-3 gap-6">
<div className="space-y-2">
<p className="text-xs font-semibold tracking-wider text-muted-foreground uppercase">Format</p>
<ul className="text-sm space-y-1">
<li className="flex justify-between">
<span className="text-muted-foreground">Sample Rate:</span>
<span className="font-semibold">{(result.sample_rate / 1000).toFixed(1)} kHz</span>
</div>
<div className="flex items-center gap-1">
<FileAudio className="h-3 w-3 text-muted-foreground"/>
<span className="font-medium font-mono">{(result.sample_rate / 1000).toFixed(1)} kHz</span>
</li>
<li className="flex justify-between">
<span className="text-muted-foreground">Bit Depth:</span>
<span className="font-semibold">{result.bit_depth}</span>
</div>
<div className="flex items-center gap-1">
<Waves className="h-3 w-3 text-muted-foreground"/>
<span className="font-medium font-mono">{result.bit_depth}</span>
</li>
<li className="flex justify-between">
<span className="text-muted-foreground">Channels:</span>
<span className="font-semibold">{result.channels === 2 ? "Stereo" : result.channels === 1 ? "Mono" : `${result.channels}`}</span>
</div>
<div className="flex items-center gap-1">
<Clock className="h-3 w-3 text-muted-foreground"/>
<span className="font-medium font-mono">{result.channels === 2 ? "Stereo" : result.channels === 1 ? "Mono" : `${result.channels}`}</span>
</li>
<li className="flex justify-between">
<span className="text-muted-foreground">Duration:</span>
<span className="font-semibold">{formatDuration(result.duration)}</span>
</div>
<div className="flex items-center gap-1">
<Gauge className="h-3 w-3 text-muted-foreground"/>
<span className="text-muted-foreground">Nyquist:</span>
<span className="font-semibold">{(nyquistFreq / 1000).toFixed(1)} kHz</span>
</div>
{result.file_size > 0 && (<div className="flex items-center gap-1">
<HardDrive className="h-3 w-3 text-muted-foreground"/>
<span className="font-medium font-mono">{formatDuration(result.duration)}</span>
</li>
{result.file_size > 0 && (
<li className="flex justify-between">
<span className="text-muted-foreground">Size:</span>
<span className="font-semibold">{formatFileSize(result.file_size)}</span>
</div>)}
<span className="font-medium font-mono">{formatFileSize(result.file_size)}</span>
</li>
)}
</ul>
</div>
<div className="flex flex-wrap items-center gap-x-3 gap-y-1 text-xs border-t pt-2">
<div className="flex items-center gap-1">
<TrendingUp className="h-3 w-3 text-muted-foreground"/>
<div className="space-y-2">
<p className="text-xs font-semibold tracking-wider text-muted-foreground uppercase">Signal Analytics</p>
<ul className="text-sm space-y-1">
<li className="flex justify-between">
<span className="text-muted-foreground">Nyquist:</span>
<span className="font-medium font-mono">{(nyquistFreq / 1000).toFixed(1)} kHz</span>
</li>
<li className="flex justify-between">
<span className="text-muted-foreground">Dynamic Range:</span>
<span className="font-semibold">{formatNumber(result.dynamic_range)} dB</span>
<span className="font-medium font-mono">{formatNumber(result.dynamic_range)} dB</span>
</li>
<li className="flex justify-between">
<span className="text-muted-foreground">Peak Amplitude:</span>
<span className="font-medium font-mono">{formatNumber(result.peak_amplitude)} dB</span>
</li>
<li className="flex justify-between">
<span className="text-muted-foreground">RMS Level:</span>
<span className="font-medium font-mono">{formatNumber(result.rms_level)} dB</span>
</li>
<li className="flex justify-between">
<span className="text-muted-foreground">Total Samples:</span>
<span className="font-medium font-mono">{result.total_samples.toLocaleString()}</span>
</li>
</ul>
</div>
<div className="flex items-center gap-1">
<span className="text-muted-foreground">Peak:</span>
<span className="font-semibold">{formatNumber(result.peak_amplitude)} dB</span>
</div>
<div className="flex items-center gap-1">
<span className="text-muted-foreground">RMS:</span>
<span className="font-semibold">{formatNumber(result.rms_level)} dB</span>
</div>
<div className="flex items-center gap-1 ml-auto">
<span className="text-muted-foreground">Samples:</span>
<span className="font-semibold">{result.total_samples.toLocaleString()}</span>
{result.spectrum && (() => {
const frames = result.spectrum.time_slices.length;
const fftSize = result.spectrum.freq_bins * 2;
const freqRes = result.sample_rate / fftSize;
return (
<div className="space-y-2">
<p className="text-xs font-semibold tracking-wider text-muted-foreground uppercase">Spectrum Meta</p>
<ul className="text-sm space-y-1">
<li className="flex justify-between">
<span className="text-muted-foreground">Analysis Frames:</span>
<span className="font-medium font-mono">{frames.toLocaleString()}</span>
</li>
<li className="flex justify-between">
<span className="text-muted-foreground">FFT Size:</span>
<span className="font-medium font-mono">{fftSize.toLocaleString()}</span>
</li>
<li className="flex justify-between">
<span className="text-muted-foreground">Freq Resolution:</span>
<span className="font-medium font-mono">{freqRes.toFixed(2)} Hz/bin</span>
</li>
</ul>
</div>
);
})()}
</div>
</CardContent>
</Card>);
</Card>
);
}
+102 -37
View File
@@ -1,18 +1,51 @@
import { useState, useCallback, useEffect } from "react";
import { useState, useCallback, useEffect, useRef } from "react";
import { Button } from "@/components/ui/button";
import { Upload, ArrowLeft, Trash2 } from "lucide-react";
import { Upload, ArrowLeft, Trash2, Download } from "lucide-react";
import { AudioAnalysis } from "@/components/AudioAnalysis";
import { SpectrumVisualization } from "@/components/SpectrumVisualization";
import { useAudioAnalysis } from "@/hooks/useAudioAnalysis";
import { SelectFile } from "../../wailsjs/go/main/App";
import { SelectFile, SaveSpectrumImage } from "../../wailsjs/go/main/App";
import { toastWithSound as toast } from "@/lib/toast-with-sound";
import { OnFileDrop, OnFileDropOff } from "../../wailsjs/runtime/runtime";
interface AudioAnalysisPageProps {
onBack?: () => void;
}
export function AudioAnalysisPage({ onBack }: AudioAnalysisPageProps) {
const { analyzing, result, analyzeFile, clearResult, selectedFilePath, spectrumLoading } = useAudioAnalysis();
const { analyzing, result, analyzeFile, clearResult, selectedFilePath, spectrumLoading, reAnalyzeSpectrum } =
useAudioAnalysis();
const [isDragging, setIsDragging] = useState(false);
const spectrumRef = useRef<{ getCanvasDataURL: () => string | null; }>(null);
const [isExporting, setIsExporting] = useState(false);
const handleExport = async () => {
if (!selectedFilePath || !spectrumRef.current)
return;
const dataUrl = spectrumRef.current.getCanvasDataURL();
if (!dataUrl) {
toast.error("Export Failed", { description: "Cannot get canvas data" });
return;
}
setIsExporting(true);
try {
const outPath = await SaveSpectrumImage(selectedFilePath, dataUrl);
toast.success("Exported Successfully", {
description: `Saved to: ${outPath}`,
});
}
catch (err) {
toast.error("Export Failed", {
description: err instanceof Error ? err.message : "Failed to save image",
});
}
finally {
setIsExporting(false);
}
};
const handleSelectFile = async () => {
try {
const filePath = await SelectFile();
@@ -26,6 +59,7 @@ export function AudioAnalysisPage({ onBack }: AudioAnalysisPageProps) {
});
}
};
const handleFileDrop = useCallback(async (_x: number, _y: number, paths: string[]) => {
setIsDragging(false);
if (paths.length === 0)
@@ -39,6 +73,7 @@ export function AudioAnalysisPage({ onBack }: AudioAnalysisPageProps) {
}
await analyzeFile(filePath);
}, [analyzeFile]);
useEffect(() => {
OnFileDrop((x, y, paths) => {
handleFileDrop(x, y, paths);
@@ -47,37 +82,55 @@ export function AudioAnalysisPage({ onBack }: AudioAnalysisPageProps) {
OnFileDropOff();
};
}, [handleFileDrop]);
const handleAnalyzeAnother = () => {
clearResult();
};
return (<div className="space-y-6">
const fileName = selectedFilePath
? selectedFilePath.split(/[/\\]/).pop()
: undefined;
return (
<div className="space-y-6">
<div className="flex items-center justify-between">
<div className="flex items-center gap-4">
{onBack && (<Button variant="ghost" size="icon" onClick={onBack}>
{onBack && (
<Button variant="ghost" size="icon" onClick={onBack}>
<ArrowLeft className="h-5 w-5" />
</Button>)}
</Button>
)}
<h1 className="text-2xl font-bold">Audio Quality Analyzer</h1>
</div>
{result && (<Button onClick={handleAnalyzeAnother} variant="outline" size="sm">
<Trash2 className="h-4 w-4"/>
{result && (
<div className="flex gap-2">
<Button
onClick={handleExport}
variant="outline"
size="sm"
disabled={isExporting || spectrumLoading}
>
<Download className="h-4 w-4 mr-1" />
{isExporting ? "Exporting..." : "Export PNG"}
</Button>
<Button onClick={handleAnalyzeAnother} variant="outline" size="sm">
<Trash2 className="h-4 w-4 mr-1" />
Clear
</Button>)}
</Button>
</div>
)}
</div>
{!result && !analyzing && (<div className={`flex flex-col items-center justify-center h-[400px] border-2 border-dashed rounded-lg transition-colors ${isDragging
{!result && !analyzing && (
<div
className={`flex flex-col items-center justify-center h-[400px] border-2 border-dashed rounded-lg transition-colors ${isDragging
? "border-primary bg-primary/10"
: "border-muted-foreground/30"}`} onDragOver={(e) => {
e.preventDefault();
setIsDragging(true);
}} onDragLeave={(e) => {
e.preventDefault();
setIsDragging(false);
}} onDrop={(e) => {
e.preventDefault();
setIsDragging(false);
}} style={{ "--wails-drop-target": "drop" } as React.CSSProperties}>
: "border-muted-foreground/30"}`}
onDragOver={(e) => { e.preventDefault(); setIsDragging(true); }}
onDragLeave={(e) => { e.preventDefault(); setIsDragging(false); }}
onDrop={(e) => { e.preventDefault(); setIsDragging(false); }}
style={{ "--wails-drop-target": "drop" } as React.CSSProperties}
>
<div className="mb-4 flex h-16 w-16 items-center justify-center rounded-full bg-muted">
<Upload className="h-8 w-8 text-primary" />
</div>
@@ -90,24 +143,36 @@ export function AudioAnalysisPage({ onBack }: AudioAnalysisPageProps) {
<Upload className="h-5 w-5" />
Select FLAC File
</Button>
</div>)}
</div>
)}
{analyzing && !result && (<div className="flex flex-col items-center justify-center py-16">
{analyzing && !result && (
<div className="flex flex-col items-center justify-center py-16">
<div className="animate-spin rounded-full h-12 w-12 border-b-2 border-primary mb-4"></div>
<p className="text-sm text-muted-foreground">Analyzing audio file...</p>
</div>)}
</div>
)}
{result && (
<div className="space-y-4">
<AudioAnalysis
result={result}
analyzing={analyzing}
showAnalyzeButton={false}
filePath={selectedFilePath}
/>
{result && (<div className="space-y-4">
<AudioAnalysis result={result} analyzing={analyzing} showAnalyzeButton={false} filePath={selectedFilePath}/>
{spectrumLoading ? (<div className="flex flex-col items-center justify-center py-16 border rounded-lg">
<div className="animate-spin rounded-full h-8 w-8 border-b-2 border-primary mb-2"></div>
<p className="text-sm text-muted-foreground">Loading spectrum data...</p>
</div>) : (<SpectrumVisualization sampleRate={result.sample_rate} bitsPerSample={result.bits_per_sample} duration={result.duration} spectrumData={result.spectrum}/>)}
</div>)}
</div>);
<SpectrumVisualization
ref={spectrumRef}
sampleRate={result.sample_rate}
duration={result.duration}
spectrumData={result.spectrum}
fileName={fileName}
onReAnalyze={reAnalyzeSpectrum}
isAnalyzingSpectrum={spectrumLoading}
/>
</div>
)}
</div>
);
}
+486 -176
View File
@@ -1,193 +1,503 @@
import { useEffect, useRef } from "react";
import { useEffect, useRef, useState, useCallback } from "react";
import type { SpectrumData } from "@/types/api";
import { Label } from "@/components/ui/label";
import { forwardRef, useImperativeHandle } from "react";
import {
Select,
SelectContent,
SelectItem,
SelectTrigger,
SelectValue,
} from "@/components/ui/select";
export interface SpectrumVisualizationHandle {
getCanvasDataURL: () => string | null;
}
interface SpectrumVisualizationProps {
sampleRate: number;
bitsPerSample: number;
duration: number;
spectrumData?: SpectrumData;
fileName?: string;
onReAnalyze?: (fftSize: number, windowFunction: string) => void;
isAnalyzingSpectrum?: boolean;
}
export function SpectrumVisualization({ sampleRate, bitsPerSample, duration, spectrumData, }: SpectrumVisualizationProps) {
type ColorScheme = "spek" | "viridis" | "hot" | "cool" | "grayscale";
function getColor(intensity: number, scheme: ColorScheme): string {
const t = Math.max(0, Math.min(1, intensity));
switch (scheme) {
case "spek":
return spekColor(t);
case "viridis":
return viridisColor(t);
case "hot":
return hotColor(t);
case "cool":
return coolColor(t);
case "grayscale": {
const v = Math.round(t * 255);
return `rgb(${v},${v},${v})`;
}
default:
return spekColor(t);
}
}
function getColorRGB(intensity: number, scheme: ColorScheme): [number, number, number] {
const t = Math.max(0, Math.min(1, intensity));
const css = getColor(t, scheme);
const m = css.match(/\d+/g)!;
return [parseInt(m[0]), parseInt(m[1]), parseInt(m[2])];
}
function spekColor(t: number): string {
if (t < 0.08) {
const v = t / 0.08;
return `rgb(0,0,${Math.round(v * 80)})`;
}
if (t < 0.18) {
const v = (t - 0.08) / 0.10;
return `rgb(${Math.round(v * 50)},${Math.round(v * 30)},${Math.round(80 + v * 175)})`;
}
if (t < 0.28) {
const v = (t - 0.18) / 0.10;
return `rgb(${Math.round(50 + v * 150)},${Math.round(30 - v * 30)},${Math.round(255 - v * 55)})`;
}
if (t < 0.40) {
const v = (t - 0.28) / 0.12;
return `rgb(${Math.round(200 + v * 55)},0,${Math.round(200 - v * 200)})`;
}
if (t < 0.52) {
const v = (t - 0.40) / 0.12;
return `rgb(255,${Math.round(v * 100)},0)`;
}
if (t < 0.65) {
const v = (t - 0.52) / 0.13;
return `rgb(255,${Math.round(100 + v * 80)},0)`;
}
if (t < 0.78) {
const v = (t - 0.65) / 0.13;
return `rgb(255,${Math.round(180 + v * 55)},${Math.round(v * 30)})`;
}
if (t < 0.90) {
const v = (t - 0.78) / 0.12;
return `rgb(255,${Math.round(235 + v * 20)},${Math.round(30 + v * 100)})`;
}
const v = (t - 0.90) / 0.10;
return `rgb(255,255,${Math.round(130 + v * 125)})`;
}
function viridisColor(t: number): string {
const stops: [number, number, number][] = [
[68, 1, 84],
[72, 36, 117],
[62, 74, 137],
[49, 104, 142],
[38, 130, 142],
[31, 158, 137],
[53, 183, 121],
[110, 206, 88],
[181, 222, 43],
[253, 231, 37],
];
const i = t * (stops.length - 1);
const lo = Math.floor(i);
const hi = Math.min(lo + 1, stops.length - 1);
const f = i - lo;
const [r, g, b] = stops[lo].map((v, k) => Math.round(v + (stops[hi][k] - v) * f)) as [number, number, number];
return `rgb(${r},${g},${b})`;
}
function hotColor(t: number): string {
if (t < 0.33) {
return `rgb(${Math.round(t / 0.33 * 255)},0,0)`;
}
if (t < 0.67) {
return `rgb(255,${Math.round((t - 0.33) / 0.34 * 255)},0)`;
}
return `rgb(255,255,${Math.round((t - 0.67) / 0.33 * 255)})`;
}
function coolColor(t: number): string {
if (t < 0.33) {
return `rgb(0,0,${Math.round(128 + t / 0.33 * 127)})`;
}
if (t < 0.67) {
return `rgb(0,${Math.round((t - 0.33) / 0.34 * 255)},255)`;
}
return `rgb(${Math.round((t - 0.67) / 0.33 * 255)},255,255)`;
}
type FreqScale = "linear" | "log2";
const MARGIN = { top: 50, right: 100, bottom: 50, left: 80 };
const CANVAS_W = 1200;
const CANVAS_H = 600;
function renderSpectrogram(
ctx: CanvasRenderingContext2D,
spectrum: SpectrumData,
sampleRate: number,
duration: number,
freqScale: FreqScale,
colorScheme: ColorScheme,
fileName?: string,
) {
const { top, right, bottom, left } = MARGIN;
const pw = CANVAS_W - left - right;
const ph = CANVAS_H - top - bottom;
ctx.fillStyle = "#000";
ctx.fillRect(0, 0, CANVAS_W, CANVAS_H);
const slices = spectrum.time_slices;
if (!slices || slices.length === 0)
return;
const numT = slices.length;
const numF = slices[0].magnitudes.length;
const maxFreq = spectrum.max_freq;
let minDB = Infinity;
let maxDB = -Infinity;
for (const s of slices) {
for (const v of s.magnitudes) {
if (v > maxDB)
maxDB = v;
if (v < minDB && v > -200)
minDB = v;
}
}
minDB = Math.max(minDB, maxDB - 90);
const dbRange = maxDB - minDB;
const img = ctx.createImageData(pw, ph);
const data = img.data;
for (let x = 0; x < pw; x++) {
const tProgress = x / (pw - 1);
const tExact = tProgress * (numT - 1);
const t0 = Math.floor(tExact);
const t1 = Math.min(t0 + 1, numT - 1);
const tf = tExact - t0;
const frame0 = slices[t0].magnitudes;
const frame1 = slices[t1].magnitudes;
for (let y = 0; y < ph; y++) {
let fProgress = (ph - 1 - y) / (ph - 1);
if (freqScale === "log2") {
const minF = 20;
const octaves = Math.log2(maxFreq / minF);
const freq = minF * Math.pow(2, fProgress * octaves);
fProgress = freq / maxFreq;
}
const fExact = fProgress * (numF - 1);
const f0 = Math.floor(fExact);
const f1 = Math.min(f0 + 1, numF - 1);
const ff = fExact - f0;
const m00 = frame0[f0] ?? minDB;
const m01 = frame0[f1] ?? minDB;
const m10 = frame1[f0] ?? minDB;
const m11 = frame1[f1] ?? minDB;
const mag = (m00 * (1 - ff) + m01 * ff) * (1 - tf) + (m10 * (1 - ff) + m11 * ff) * tf;
const norm = Math.max(0, Math.min(1, (mag - minDB) / dbRange));
const [r, g, b] = getColorRGB(norm, colorScheme);
const idx = (y * pw + x) * 4;
data[idx] = r;
data[idx + 1] = g;
data[idx + 2] = b;
data[idx + 3] = 255;
}
}
ctx.putImageData(img, left, top);
ctx.fillStyle = "#ccc";
ctx.font = "12px 'Segoe UI', Arial";
ctx.textAlign = "right";
ctx.textBaseline = "middle";
const freqLabels = buildFreqLabels(maxFreq, freqScale);
for (const freq of freqLabels) {
if (freq > maxFreq)
continue;
let yPos: number;
if (freqScale === "log2") {
const minF = 20;
const norm = Math.log2(freq / minF) / Math.log2(maxFreq / minF);
yPos = top + ph - norm * ph;
} else {
yPos = top + ph - (freq / maxFreq) * ph;
}
const label = freq >= 1000 ? `${freq / 1000}k` : `${freq}`;
ctx.fillText(label, left - 8, yPos);
ctx.strokeStyle = "rgba(255, 255, 255, 0.1)";
ctx.lineWidth = 1;
ctx.beginPath();
ctx.moveTo(left - 4, yPos);
ctx.lineTo(left + pw, yPos);
ctx.stroke();
}
ctx.textAlign = "center";
ctx.textBaseline = "top";
const timeStep = smartTimeStep(duration);
for (let t = 0; t <= duration; t += timeStep) {
const xPos = left + (t / duration) * pw;
const label = timeStep >= 60
? `${Math.floor(t / 60)}m${t % 60 ? (t % 60) + "s" : ""}`
: `${t}s`;
ctx.fillText(label, xPos, top + ph + 8);
ctx.strokeStyle = "rgba(255, 255, 255, 0.1)";
ctx.lineWidth = 1;
ctx.beginPath();
ctx.moveTo(xPos, top);
ctx.lineTo(xPos, top + ph + 4);
ctx.stroke();
}
ctx.fillStyle = "#fff";
ctx.font = "13px 'Segoe UI', Arial";
ctx.textAlign = "center";
ctx.textBaseline = "bottom";
ctx.fillText("Time (seconds)", left + pw / 2, CANVAS_H - 12);
ctx.save();
ctx.translate(24, top + ph / 2);
ctx.rotate(-Math.PI / 2);
ctx.textBaseline = "middle";
ctx.fillText("Frequency (Hz)", 0, 0);
ctx.restore();
ctx.font = "12px 'Segoe UI', Arial";
ctx.textAlign = "left";
ctx.textBaseline = "middle";
ctx.fillStyle = "#fff";
if (fileName)
ctx.fillText(fileName, left, 26);
ctx.textAlign = "right";
ctx.fillText(`Sample Rate: ${sampleRate} Hz`, left + pw, 26);
const cbX = left + pw + 25;
const cbW = 14;
for (let i = 0; i < ph; i++) {
const norm = 1 - i / ph;
ctx.fillStyle = getColor(norm, colorScheme);
ctx.fillRect(cbX, top + i, cbW, 1);
}
ctx.strokeStyle = "rgba(255, 255, 255, 0.5)";
ctx.lineWidth = 1;
ctx.strokeRect(cbX, top, cbW, ph);
ctx.fillStyle = "#fff";
ctx.font = "10px 'Segoe UI', Arial";
ctx.textAlign = "left";
ctx.textBaseline = "middle";
ctx.fillText("High", cbX + cbW + 6, top + 6);
ctx.fillText("Low", cbX + cbW + 6, top + ph - 6);
}
function buildFreqLabels(maxFreq: number, scale: FreqScale): number[] {
if (scale === "log2") {
const labels: number[] = [];
for (let f = 20; f <= maxFreq; f *= 2)
labels.push(f);
for (let f = 100; f <= maxFreq; f *= 10)
labels.push(f);
return [...new Set(labels)].sort((a, b) => a - b);
}
if (maxFreq <= 24000)
return [2000, 4000, 6000, 8000, 10000, 12000, 14000, 16000, 18000, 20000, 22000];
if (maxFreq <= 48000)
return [5000, 10000, 15000, 20000, 25000, 30000, 35000, 40000, 45000];
if (maxFreq <= 96000)
return [10000, 20000, 30000, 40000, 50000, 60000, 70000, 80000, 90000];
return [20000, 40000, 60000, 80000, 100000, 120000, 140000, 160000, 180000];
}
function smartTimeStep(duration: number): number {
if (duration <= 30)
return 5;
if (duration <= 60)
return 10;
if (duration <= 120)
return 15;
if (duration <= 300)
return 30;
if (duration <= 600)
return 60;
return 120;
}
const COLOR_SCHEMES: { value: ColorScheme; label: string; gradient: string; }[] = [
{ value: "spek", label: "Spek", gradient: "linear-gradient(to right, #000050, #1e0080, #4000ff, #8000ff, #ff0080, #ff4000, #ff8000, #ffff00)" },
{ value: "viridis", label: "Viridis", gradient: "linear-gradient(to right, #440154, #31688e, #35b779, #fde725)" },
{ value: "hot", label: "Hot", gradient: "linear-gradient(to right, #000, #f00, #ff0, #fff)" },
{ value: "cool", label: "Cool", gradient: "linear-gradient(to right, #000080, #0000ff, #00ffff, #ffffff)" },
{ value: "grayscale", label: "Grayscale", gradient: "linear-gradient(to right, #000, #fff)" },
];
export const SpectrumVisualization = forwardRef<SpectrumVisualizationHandle, SpectrumVisualizationProps>(({
sampleRate,
duration,
spectrumData,
fileName,
onReAnalyze,
isAnalyzingSpectrum,
}, ref) => {
const canvasRef = useRef<HTMLCanvasElement>(null);
useImperativeHandle(ref, () => ({
getCanvasDataURL: () => {
if (!canvasRef.current)
return null;
return canvasRef.current.toDataURL("image/png");
}
}));
const [freqScale, setFreqScale] = useState<FreqScale>("linear");
const [colorScheme, setColorScheme] = useState<ColorScheme>("spek");
const [fftSize, setFftSize] = useState<string>(() => {
if (spectrumData && spectrumData.freq_bins) {
return String(spectrumData.freq_bins * 2);
}
return "4096";
});
const [windowFunction, setWindowFunction] = useState<string>("hann");
useEffect(() => {
if (spectrumData && spectrumData.freq_bins) {
setFftSize(String(spectrumData.freq_bins * 2));
}
}, [spectrumData]);
const handleReAnalyze = (newFftSize: string, newWindowFunc: string) => {
setFftSize(newFftSize);
setWindowFunction(newWindowFunc);
if (onReAnalyze) {
onReAnalyze(parseInt(newFftSize), newWindowFunc);
}
};
const draw = useCallback(() => {
const canvas = canvasRef.current;
if (!canvas)
return;
const ctx = canvas.getContext("2d");
if (!ctx)
return;
const width = canvas.width;
const height = canvas.height;
const marginLeft = 70;
const marginRight = 70;
const marginTop = 30;
const marginBottom = 65;
const plotWidth = width - marginLeft - marginRight;
const plotHeight = height - marginTop - marginBottom;
ctx.fillStyle = "#000000";
ctx.fillRect(0, 0, width, height);
const nyquistFreq = sampleRate / 2;
if (spectrumData) {
drawRealSpectrum(ctx, marginLeft, marginTop, plotWidth, plotHeight, spectrumData);
}
drawAxesAndLabels(ctx, marginLeft, marginTop, plotWidth, plotHeight, nyquistFreq, duration, sampleRate);
drawColorBar(ctx, marginLeft + plotWidth + 15, marginTop, 20, plotHeight);
}, [sampleRate, bitsPerSample, duration, spectrumData]);
const drawRealSpectrum = (ctx: CanvasRenderingContext2D, x: number, y: number, width: number, height: number, spectrum: SpectrumData) => {
const timeSlices = spectrum.time_slices;
if (timeSlices.length === 0)
return;
const freqBins = timeSlices[0].magnitudes.length;
const nyquistFreq = spectrum.max_freq;
let minDB = 0;
let maxDB = -200;
timeSlices.forEach((slice) => {
slice.magnitudes.forEach((db) => {
if (db > maxDB)
maxDB = db;
if (db < minDB && db > -200)
minDB = db;
});
});
minDB = Math.max(minDB, maxDB - 90);
const dbRange = maxDB - minDB;
const sliceWidth = Math.ceil(width / timeSlices.length);
for (let t = 0; t < timeSlices.length; t++) {
const slice = timeSlices[t];
const xPos = x + (t / timeSlices.length) * width;
for (let f = 0; f < freqBins && f < slice.magnitudes.length; f++) {
const db = slice.magnitudes[f];
const freq = (f / freqBins) * nyquistFreq;
const freqRatio = freq / nyquistFreq;
const yPos = y + height - (freqRatio * height);
const nextFreq = ((f + 1) / freqBins) * nyquistFreq;
const nextFreqRatio = nextFreq / nyquistFreq;
const nextYPos = y + height - (nextFreqRatio * height);
const binHeight = Math.max(1, Math.abs(yPos - nextYPos) + 1);
const intensity = Math.max(0, Math.min(1, (db - minDB) / dbRange));
const color = getSpekColor(intensity);
ctx.fillStyle = color;
ctx.fillRect(xPos, nextYPos, sliceWidth, binHeight);
}
}
};
const getSpekColor = (intensity: number): string => {
if (intensity < 0.08) {
const t = intensity / 0.08;
return `rgb(0, 0, ${Math.floor(t * 80)})`;
}
else if (intensity < 0.18) {
const t = (intensity - 0.08) / 0.10;
return `rgb(${Math.floor(t * 50)}, ${Math.floor(t * 30)}, ${Math.floor(80 + t * 175)})`;
}
else if (intensity < 0.28) {
const t = (intensity - 0.18) / 0.10;
return `rgb(${Math.floor(50 + t * 150)}, ${Math.floor(30 - t * 30)}, ${Math.floor(255 - t * 55)})`;
}
else if (intensity < 0.40) {
const t = (intensity - 0.28) / 0.12;
return `rgb(${Math.floor(200 + t * 55)}, 0, ${Math.floor(200 - t * 200)})`;
}
else if (intensity < 0.52) {
const t = (intensity - 0.40) / 0.12;
return `rgb(255, ${Math.floor(t * 100)}, 0)`;
}
else if (intensity < 0.65) {
const t = (intensity - 0.52) / 0.13;
return `rgb(255, ${Math.floor(100 + t * 80)}, 0)`;
}
else if (intensity < 0.78) {
const t = (intensity - 0.65) / 0.13;
return `rgb(255, ${Math.floor(180 + t * 55)}, ${Math.floor(t * 30)})`;
}
else if (intensity < 0.90) {
const t = (intensity - 0.78) / 0.12;
return `rgb(255, ${Math.floor(235 + t * 20)}, ${Math.floor(30 + t * 100)})`;
}
else {
const t = (intensity - 0.90) / 0.10;
return `rgb(255, 255, ${Math.floor(130 + t * 125)})`;
}
};
const drawAxesAndLabels = (ctx: CanvasRenderingContext2D, x: number, y: number, width: number, height: number, nyquistFreq: number, duration: number, sampleRate: number) => {
ctx.fillStyle = "#CCCCCC";
ctx.font = "12px Arial";
ctx.textAlign = "right";
ctx.textBaseline = "middle";
const freqLabels = generateFreqLabels(nyquistFreq);
freqLabels.forEach(freq => {
if (freq <= nyquistFreq) {
const freqRatio = freq / nyquistFreq;
const yPos = y + height - (freqRatio * height);
const label = freq >= 1000 ? `${freq / 1000}k` : `${freq}`;
ctx.fillText(label, x - 8, yPos);
}
});
ctx.fillText("0", x - 8, y + height);
renderSpectrogram(ctx, spectrumData, sampleRate, duration, freqScale, colorScheme, fileName);
} else {
ctx.fillStyle = "#000";
ctx.fillRect(0, 0, CANVAS_W, CANVAS_H);
ctx.fillStyle = "#444";
ctx.font = "16px Arial";
ctx.textAlign = "center";
ctx.textBaseline = "top";
const timeStep = getTimeStep(duration);
for (let t = 0; t <= duration; t += timeStep) {
const xPos = x + (t / duration) * width;
ctx.fillText(`${Math.round(t)}s`, xPos, y + height + 8);
}
ctx.fillStyle = "#FFFFFF";
ctx.font = "13px Arial";
ctx.save();
ctx.translate(12, y + height / 2);
ctx.rotate(-Math.PI / 2);
ctx.textAlign = "center";
ctx.fillText("Frequency (Hz)", 0, 0);
ctx.restore();
ctx.textAlign = "center";
ctx.fillText("Time (seconds)", x + width / 2, y + height + 35);
ctx.textAlign = "right";
ctx.fillStyle = "#CCCCCC";
ctx.font = "12px Arial";
ctx.fillText(`Sample Rate: ${sampleRate} Hz`, x + width - 5, y - 3);
};
const generateFreqLabels = (nyquistFreq: number): number[] => {
if (nyquistFreq <= 24000) {
return [2000, 4000, 6000, 8000, 10000, 12000, 14000, 16000, 18000, 20000, 22000];
}
else if (nyquistFreq <= 48000) {
return [5000, 10000, 15000, 20000, 25000, 30000, 35000, 40000, 45000];
}
else if (nyquistFreq <= 96000) {
return [10000, 20000, 30000, 40000, 50000, 60000, 70000, 80000, 90000];
}
else {
return [20000, 40000, 60000, 80000, 100000, 120000, 140000, 160000, 180000];
}
};
const getTimeStep = (duration: number): number => {
if (duration <= 60)
return 15;
if (duration <= 120)
return 30;
if (duration <= 300)
return 30;
if (duration <= 600)
return 60;
return 60;
};
const drawColorBar = (ctx: CanvasRenderingContext2D, x: number, y: number, width: number, height: number) => {
for (let i = 0; i < height; i++) {
const intensity = 1 - (i / height);
const color = getSpekColor(intensity);
ctx.fillStyle = color;
ctx.fillRect(x, y + i, width, 1);
}
ctx.strokeStyle = "#666666";
ctx.lineWidth = 1;
ctx.strokeRect(x, y, width, height);
ctx.fillStyle = "#FFFFFF";
ctx.font = "11px Arial";
ctx.textAlign = "left";
ctx.textBaseline = "middle";
ctx.fillText("High", x + width + 5, y + 10);
ctx.fillText("Low", x + width + 5, y + height - 10);
};
return (<div className="border border-white/10 rounded-lg overflow-hidden bg-black shadow-xl">
<canvas ref={canvasRef} width={1200} height={600} className="w-full h-auto" style={{ imageRendering: "auto" }}/>
</div>);
ctx.fillText("No spectrum data", CANVAS_W / 2, CANVAS_H / 2);
}
}, [spectrumData, sampleRate, duration, freqScale, colorScheme, fileName]);
useEffect(() => { draw(); }, [draw]);
useEffect(() => { draw(); }, [draw]);
return (
<div className="space-y-4">
<div className="flex flex-wrap items-center gap-3 sm:gap-4 p-1">
<div className="flex items-center gap-2">
<Label className="whitespace-nowrap text-sm font-medium">Color Scheme:</Label>
<Select value={colorScheme} onValueChange={(v) => setColorScheme(v as ColorScheme)} disabled={isAnalyzingSpectrum}>
<SelectTrigger className="h-8 w-[130px] text-sm">
<SelectValue />
</SelectTrigger>
<SelectContent>
{COLOR_SCHEMES.map((s) => (
<SelectItem key={s.value} value={s.value}>
<div className="flex items-center gap-2">
<div
className="h-4 w-4 rounded-sm border opacity-90"
style={{ backgroundImage: s.gradient }}
/>
<span>{s.label}</span>
</div>
</SelectItem>
))}
</SelectContent>
</Select>
</div>
<div className="h-6 w-px bg-border hidden sm:block mx-1"></div>
<div className="flex items-center gap-2">
<Label className="whitespace-nowrap text-sm font-medium">Freq Scale:</Label>
<Select value={freqScale} onValueChange={(v) => { if (v) setFreqScale(v as FreqScale); }} disabled={isAnalyzingSpectrum}>
<SelectTrigger className="h-8 w-[90px] text-sm">
<SelectValue />
</SelectTrigger>
<SelectContent>
<SelectItem value="linear">Linear</SelectItem>
<SelectItem value="log2">Log2</SelectItem>
</SelectContent>
</Select>
</div>
<div className="flex items-center gap-2">
<Label className="whitespace-nowrap text-sm font-medium">FFT Size:</Label>
<Select value={fftSize} onValueChange={(v) => handleReAnalyze(v, windowFunction)} disabled={isAnalyzingSpectrum}>
<SelectTrigger className="h-8 w-[90px] text-sm">
<SelectValue />
</SelectTrigger>
<SelectContent>
<SelectItem value="512">512</SelectItem>
<SelectItem value="1024">1024</SelectItem>
<SelectItem value="2048">2048</SelectItem>
<SelectItem value="4096">4096</SelectItem>
</SelectContent>
</Select>
</div>
<div className="flex items-center gap-2">
<Label className="whitespace-nowrap text-sm font-medium">Window:</Label>
<Select value={windowFunction} onValueChange={(v) => handleReAnalyze(fftSize, v)} disabled={isAnalyzingSpectrum}>
<SelectTrigger className="h-8 w-[115px] text-sm capitalize">
<SelectValue />
</SelectTrigger>
<SelectContent>
<SelectItem value="hann">Hann</SelectItem>
<SelectItem value="hamming">Hamming</SelectItem>
<SelectItem value="blackman">Blackman</SelectItem>
<SelectItem value="rectangular">Rectangular</SelectItem>
</SelectContent>
</Select>
</div>
</div>
<div className="relative border border-white/10 rounded-lg overflow-hidden bg-black shadow-xl">
{isAnalyzingSpectrum && (
<div className="absolute inset-0 bg-black/60 flex flex-col items-center justify-center z-10 backdrop-blur-sm">
<div className="animate-spin rounded-full h-8 w-8 border-b-2 border-primary mb-2"></div>
<p className="text-sm text-foreground">Re-analyzing spectrum...</p>
</div>
)}
<canvas
ref={canvasRef}
width={CANVAS_W}
height={CANVAS_H}
className="w-full h-auto"
style={{ imageRendering: "auto" }}
/>
</div>
</div>
);
});
+25 -1
View File
@@ -1,5 +1,5 @@
import { useState, useCallback, useEffect } from "react";
import { AnalyzeTrack } from "../../wailsjs/go/main/App";
import { AnalyzeTrack, AnalyzeSpectrumWithParams } from "../../wailsjs/go/main/App";
import type { AnalysisResult } from "@/types/api";
import { logger } from "@/lib/logger";
import { toastWithSound as toast } from "@/lib/toast-with-sound";
@@ -135,6 +135,29 @@ export function useAudioAnalysis() {
}
};
}, [result, selectedFilePath, spectrumLoading]);
const reAnalyzeSpectrum = useCallback(async (fftSize: number, windowFunction: string) => {
if (!selectedFilePath || !result)
return;
setSpectrumLoading(true);
try {
const response = await AnalyzeSpectrumWithParams(selectedFilePath, fftSize, windowFunction);
const spectrumData = JSON.parse(response);
setResult(prev => prev ? { ...prev, spectrum: spectrumData } : null);
setSpectrumCache(selectedFilePath, spectrumData);
}
catch (err) {
const errorMessage = err instanceof Error ? err.message : "Failed to re-analyze spectrum";
logger.error(`Spectrum re-analysis error: ${errorMessage}`);
toast.error("Spectrum Analysis Failed", {
description: errorMessage,
});
}
finally {
setSpectrumLoading(false);
}
}, [selectedFilePath, result]);
return {
analyzing,
result,
@@ -142,6 +165,7 @@ export function useAudioAnalysis() {
selectedFilePath,
spectrumLoading,
analyzeFile,
reAnalyzeSpectrum,
clearResult,
};
}