feat(tools): 2.0 phase 5 wave 5b - ai pool: ocr-pdf, transcription, background composites (5 tools) (#226)

This commit is contained in:
SnapOtter
2026-06-13 10:19:47 +08:00
parent 6e1b9865f1
commit 51666cdd5f
59 changed files with 5423 additions and 611 deletions
@@ -0,0 +1,135 @@
import { Download } from "lucide-react";
import { useState } from "react";
import { ProgressCard } from "@/components/common/progress-card";
import { useTranslation } from "@/contexts/i18n-context";
import { useToolProcessor } from "@/hooks/use-tool-processor";
import { format } from "@/lib/format";
import { useFileStore } from "@/stores/file-store";
const LANGUAGE_OPTIONS = [
{ value: "auto", labelKey: "autoDetect" },
{ value: "en", label: "English" },
{ value: "de", label: "German" },
{ value: "fr", label: "French" },
{ value: "es", label: "Spanish" },
{ value: "zh", label: "Chinese" },
{ value: "ja", label: "Japanese" },
{ value: "ko", label: "Korean" },
{ value: "id", label: "Indonesian" },
{ value: "th", label: "Thai" },
{ value: "vi", label: "Vietnamese" },
] as const;
const FORMAT_OPTIONS = [
{ value: "srt", label: "SRT" },
{ value: "vtt", label: "WebVTT" },
] as const;
export function AutoSubtitlesSettings() {
const { t } = useTranslation();
const { files } = useFileStore();
const { processFiles, processAllFiles, processing, error, downloadUrl, progress } =
useToolProcessor("auto-subtitles");
const [language, setLanguage] = useState("auto");
const [subtitleFormat, setSubtitleFormat] = useState("srt");
const ts = t.toolSettings["auto-subtitles"];
const hasFile = files.length > 0;
const hasMultiple = files.length > 1;
const handleProcess = () => {
const settings = { language, format: subtitleFormat };
if (hasMultiple) {
processAllFiles(files, settings);
} else {
processFiles(files, settings);
}
};
const handleSubmit = (e: React.FormEvent) => {
e.preventDefault();
handleProcess();
};
return (
<form onSubmit={handleSubmit} className="space-y-4">
{/* Language */}
<div>
<label htmlFor="as-language" className="mb-1.5 block text-sm font-medium">
{ts.language}
</label>
<select
id="as-language"
value={language}
onChange={(e) => setLanguage(e.target.value)}
className="border-border bg-background w-full rounded-md border px-3 py-2 text-sm"
>
{LANGUAGE_OPTIONS.map((opt) => (
<option key={opt.value} value={opt.value}>
{"labelKey" in opt ? (ts as Record<string, string>)[opt.labelKey] : opt.label}
</option>
))}
</select>
</div>
{/* Subtitle Format */}
<div>
<label htmlFor="as-format" className="mb-1.5 block text-sm font-medium">
{ts.format}
</label>
<select
id="as-format"
value={subtitleFormat}
onChange={(e) => setSubtitleFormat(e.target.value)}
className="border-border bg-background w-full rounded-md border px-3 py-2 text-sm"
>
{FORMAT_OPTIONS.map((opt) => (
<option key={opt.value} value={opt.value}>
{opt.label}
</option>
))}
</select>
</div>
{/* Progress / Submit */}
{processing && progress ? (
<ProgressCard
active={processing}
phase={progress.phase === "idle" ? "uploading" : progress.phase}
label={ts.progressLabel}
stage={progress.stage}
percent={progress.percent}
elapsed={progress.elapsed}
/>
) : (
<button
type="submit"
disabled={!hasFile || processing}
className="bg-primary text-primary-foreground hover:bg-primary/90 disabled:bg-muted w-full rounded-md px-4 py-2 text-sm font-medium disabled:cursor-not-allowed"
>
{hasMultiple ? format(ts.submitBatch, { count: files.length }) : ts.submit}
</button>
)}
{/* Error */}
{error && (
<div className="text-destructive rounded-md bg-red-50 p-3 text-sm dark:bg-red-950">
{error}
</div>
)}
{/* Download */}
{downloadUrl && (
<a
href={downloadUrl}
download
className="bg-primary text-primary-foreground hover:bg-primary/90 flex items-center justify-center gap-2 rounded-md px-4 py-2 text-sm font-medium"
>
<Download className="h-4 w-4" />
{t.common.download}
</a>
)}
</form>
);
}
@@ -0,0 +1,103 @@
import { Download } from "lucide-react";
import { useState } from "react";
import { ProgressCard } from "@/components/common/progress-card";
import { useTranslation } from "@/contexts/i18n-context";
import { useToolProcessor } from "@/hooks/use-tool-processor";
import { format } from "@/lib/format";
import { useFileStore } from "@/stores/file-store";
export function BackgroundReplaceSettings() {
const { t } = useTranslation();
const { files } = useFileStore();
const { processFiles, processAllFiles, processing, error, downloadUrl, progress } =
useToolProcessor("background-replace");
const [color, setColor] = useState("#ffffff");
const ts = t.toolSettings["background-replace"];
const hasFile = files.length > 0;
const hasMultiple = files.length > 1;
const handleProcess = () => {
const settings = { color };
if (hasMultiple) {
processAllFiles(files, settings);
} else {
processFiles(files, settings);
}
};
const handleSubmit = (e: React.FormEvent) => {
e.preventDefault();
handleProcess();
};
return (
<form onSubmit={handleSubmit} className="space-y-4">
{/* Color */}
<div>
<label htmlFor="bg-replace-color" className="mb-1.5 block text-sm font-medium">
{ts.color}
</label>
<div className="flex items-center gap-3">
<input
id="bg-replace-color"
type="color"
value={color}
onChange={(e) => setColor(e.target.value)}
className="h-10 w-14 cursor-pointer rounded-md border border-border"
/>
<input
type="text"
value={color}
onChange={(e) => {
const v = e.target.value;
if (/^#[0-9a-fA-F]{0,6}$/.test(v)) setColor(v);
}}
className="border-border bg-background w-28 rounded-md border px-3 py-2 font-mono text-sm"
maxLength={7}
/>
</div>
</div>
{/* Progress / Submit */}
{processing && progress ? (
<ProgressCard
active={processing}
phase={progress.phase === "idle" ? "uploading" : progress.phase}
label={ts.progressLabel}
stage={progress.stage}
percent={progress.percent}
elapsed={progress.elapsed}
/>
) : (
<button
type="submit"
disabled={!hasFile || processing}
className="bg-primary text-primary-foreground hover:bg-primary/90 disabled:bg-muted w-full rounded-md px-4 py-2 text-sm font-medium disabled:cursor-not-allowed"
>
{hasMultiple ? format(ts.submitBatch, { count: files.length }) : ts.submit}
</button>
)}
{/* Error */}
{error && (
<div className="text-destructive rounded-md bg-red-50 p-3 text-sm dark:bg-red-950">
{error}
</div>
)}
{/* Download */}
{downloadUrl && (
<a
href={downloadUrl}
download
className="bg-primary text-primary-foreground hover:bg-primary/90 flex items-center justify-center gap-2 rounded-md px-4 py-2 text-sm font-medium"
>
<Download className="h-4 w-4" />
{t.common.download}
</a>
)}
</form>
);
}
@@ -0,0 +1,93 @@
import { Download } from "lucide-react";
import { useState } from "react";
import { ProgressCard } from "@/components/common/progress-card";
import { useTranslation } from "@/contexts/i18n-context";
import { useToolProcessor } from "@/hooks/use-tool-processor";
import { format } from "@/lib/format";
import { useFileStore } from "@/stores/file-store";
export function BlurBackgroundSettings() {
const { t } = useTranslation();
const { files } = useFileStore();
const { processFiles, processAllFiles, processing, error, downloadUrl, progress } =
useToolProcessor("blur-background");
const [intensity, setIntensity] = useState(50);
const ts = t.toolSettings["blur-background"];
const hasFile = files.length > 0;
const hasMultiple = files.length > 1;
const handleProcess = () => {
const settings = { intensity };
if (hasMultiple) {
processAllFiles(files, settings);
} else {
processFiles(files, settings);
}
};
const handleSubmit = (e: React.FormEvent) => {
e.preventDefault();
handleProcess();
};
return (
<form onSubmit={handleSubmit} className="space-y-4">
{/* Intensity */}
<div>
<label htmlFor="blur-bg-intensity" className="mb-1.5 block text-sm font-medium">
{ts.intensity}: {intensity}%
</label>
<input
id="blur-bg-intensity"
type="range"
min={1}
max={100}
value={intensity}
onChange={(e) => setIntensity(Number(e.target.value))}
className="w-full"
/>
</div>
{/* Progress / Submit */}
{processing && progress ? (
<ProgressCard
active={processing}
phase={progress.phase === "idle" ? "uploading" : progress.phase}
label={ts.progressLabel}
stage={progress.stage}
percent={progress.percent}
elapsed={progress.elapsed}
/>
) : (
<button
type="submit"
disabled={!hasFile || processing}
className="bg-primary text-primary-foreground hover:bg-primary/90 disabled:bg-muted w-full rounded-md px-4 py-2 text-sm font-medium disabled:cursor-not-allowed"
>
{hasMultiple ? format(ts.submitBatch, { count: files.length }) : ts.submit}
</button>
)}
{/* Error */}
{error && (
<div className="text-destructive rounded-md bg-red-50 p-3 text-sm dark:bg-red-950">
{error}
</div>
)}
{/* Download */}
{downloadUrl && (
<a
href={downloadUrl}
download
className="bg-primary text-primary-foreground hover:bg-primary/90 flex items-center justify-center gap-2 rounded-md px-4 py-2 text-sm font-medium"
>
<Download className="h-4 w-4" />
{t.common.download}
</a>
)}
</form>
);
}
@@ -0,0 +1,154 @@
import { Download } from "lucide-react";
import { useState } from "react";
import { ProgressCard } from "@/components/common/progress-card";
import { useTranslation } from "@/contexts/i18n-context";
import { useToolProcessor } from "@/hooks/use-tool-processor";
import { format } from "@/lib/format";
import { useFileStore } from "@/stores/file-store";
const QUALITY_OPTIONS = [
{ value: "fast", label: "Fast" },
{ value: "balanced", label: "Balanced" },
{ value: "best", label: "Best" },
] as const;
const LANGUAGE_OPTIONS = [
{ value: "auto", labelKey: "autoDetect" },
{ value: "en", label: "English" },
{ value: "de", label: "German" },
{ value: "fr", label: "French" },
{ value: "es", label: "Spanish" },
{ value: "zh", label: "Chinese" },
{ value: "ja", label: "Japanese" },
{ value: "ko", label: "Korean" },
] as const;
export function OcrPdfSettings() {
const { t } = useTranslation();
const { files } = useFileStore();
const { processFiles, processAllFiles, processing, error, downloadUrl, progress } =
useToolProcessor("ocr-pdf");
const [quality, setQuality] = useState("balanced");
const [language, setLanguage] = useState("auto");
const [pages, setPages] = useState("all");
const ts = t.toolSettings["ocr-pdf"];
const hasFile = files.length > 0;
const hasMultiple = files.length > 1;
const handleProcess = () => {
const settings = { quality, language, pages };
if (hasMultiple) {
processAllFiles(files, settings);
} else {
processFiles(files, settings);
}
};
const handleSubmit = (e: React.FormEvent) => {
e.preventDefault();
handleProcess();
};
return (
<form onSubmit={handleSubmit} className="space-y-4">
{/* Quality */}
<div>
<label htmlFor="ocrpdf-quality" className="mb-1.5 block text-sm font-medium">
{ts.quality}
</label>
<div className="grid grid-cols-3 gap-1.5">
{QUALITY_OPTIONS.map((opt) => (
<button
key={opt.value}
type="button"
onClick={() => setQuality(opt.value)}
className={`rounded-lg border px-2 py-2 text-xs font-medium transition-colors ${
quality === opt.value
? "border-primary bg-primary/10 text-primary"
: "border-border text-muted-foreground hover:border-primary/50"
}`}
>
{opt.label}
</button>
))}
</div>
</div>
{/* Language */}
<div>
<label htmlFor="ocrpdf-language" className="mb-1.5 block text-sm font-medium">
{ts.language}
</label>
<select
id="ocrpdf-language"
value={language}
onChange={(e) => setLanguage(e.target.value)}
className="border-border bg-background w-full rounded-md border px-3 py-2 text-sm"
>
{LANGUAGE_OPTIONS.map((opt) => (
<option key={opt.value} value={opt.value}>
{"labelKey" in opt ? (ts as Record<string, string>)[opt.labelKey] : opt.label}
</option>
))}
</select>
</div>
{/* Pages */}
<div>
<label htmlFor="ocrpdf-pages" className="mb-1.5 block text-sm font-medium">
{ts.pages}
</label>
<input
id="ocrpdf-pages"
type="text"
value={pages}
onChange={(e) => setPages(e.target.value)}
placeholder={ts.pagesPlaceholder}
className="border-border bg-background w-full rounded-md border px-3 py-2 text-sm"
/>
<p className="mt-1 text-xs text-muted-foreground">{ts.pagesHint}</p>
</div>
{/* Progress / Submit */}
{processing && progress ? (
<ProgressCard
active={processing}
phase={progress.phase === "idle" ? "uploading" : progress.phase}
label={ts.progressLabel}
stage={progress.stage}
percent={progress.percent}
elapsed={progress.elapsed}
/>
) : (
<button
type="submit"
disabled={!hasFile || processing}
className="bg-primary text-primary-foreground hover:bg-primary/90 disabled:bg-muted w-full rounded-md px-4 py-2 text-sm font-medium disabled:cursor-not-allowed"
>
{hasMultiple ? format(ts.submitBatch, { count: files.length }) : ts.submit}
</button>
)}
{/* Error */}
{error && (
<div className="text-destructive rounded-md bg-red-50 p-3 text-sm dark:bg-red-950">
{error}
</div>
)}
{/* Download */}
{downloadUrl && (
<a
href={downloadUrl}
download
className="bg-primary text-primary-foreground hover:bg-primary/90 flex items-center justify-center gap-2 rounded-md px-4 py-2 text-sm font-medium"
>
<Download className="h-4 w-4" />
{t.common.download}
</a>
)}
</form>
);
}
@@ -0,0 +1,136 @@
import { Download } from "lucide-react";
import { useState } from "react";
import { ProgressCard } from "@/components/common/progress-card";
import { useTranslation } from "@/contexts/i18n-context";
import { useToolProcessor } from "@/hooks/use-tool-processor";
import { format } from "@/lib/format";
import { useFileStore } from "@/stores/file-store";
const LANGUAGE_OPTIONS = [
{ value: "auto", labelKey: "autoDetect" },
{ value: "en", label: "English" },
{ value: "de", label: "German" },
{ value: "fr", label: "French" },
{ value: "es", label: "Spanish" },
{ value: "zh", label: "Chinese" },
{ value: "ja", label: "Japanese" },
{ value: "ko", label: "Korean" },
{ value: "id", label: "Indonesian" },
{ value: "th", label: "Thai" },
{ value: "vi", label: "Vietnamese" },
] as const;
const FORMAT_OPTIONS = [
{ value: "txt", labelKey: "plainText" },
{ value: "srt", label: "SRT" },
{ value: "vtt", label: "WebVTT" },
] as const;
export function TranscribeAudioSettings() {
const { t } = useTranslation();
const { files } = useFileStore();
const { processFiles, processAllFiles, processing, error, downloadUrl, progress } =
useToolProcessor("transcribe-audio");
const [language, setLanguage] = useState("auto");
const [outputFormat, setOutputFormat] = useState("txt");
const ts = t.toolSettings["transcribe-audio"];
const hasFile = files.length > 0;
const hasMultiple = files.length > 1;
const handleProcess = () => {
const settings = { language, outputFormat };
if (hasMultiple) {
processAllFiles(files, settings);
} else {
processFiles(files, settings);
}
};
const handleSubmit = (e: React.FormEvent) => {
e.preventDefault();
handleProcess();
};
return (
<form onSubmit={handleSubmit} className="space-y-4">
{/* Language */}
<div>
<label htmlFor="ta-language" className="mb-1.5 block text-sm font-medium">
{ts.language}
</label>
<select
id="ta-language"
value={language}
onChange={(e) => setLanguage(e.target.value)}
className="border-border bg-background w-full rounded-md border px-3 py-2 text-sm"
>
{LANGUAGE_OPTIONS.map((opt) => (
<option key={opt.value} value={opt.value}>
{"labelKey" in opt ? (ts as Record<string, string>)[opt.labelKey] : opt.label}
</option>
))}
</select>
</div>
{/* Output Format */}
<div>
<label htmlFor="ta-output-format" className="mb-1.5 block text-sm font-medium">
{ts.outputFormat}
</label>
<select
id="ta-output-format"
value={outputFormat}
onChange={(e) => setOutputFormat(e.target.value)}
className="border-border bg-background w-full rounded-md border px-3 py-2 text-sm"
>
{FORMAT_OPTIONS.map((opt) => (
<option key={opt.value} value={opt.value}>
{"labelKey" in opt ? (ts as Record<string, string>)[opt.labelKey] : opt.label}
</option>
))}
</select>
</div>
{/* Progress / Submit */}
{processing && progress ? (
<ProgressCard
active={processing}
phase={progress.phase === "idle" ? "uploading" : progress.phase}
label={ts.progressLabel}
stage={progress.stage}
percent={progress.percent}
elapsed={progress.elapsed}
/>
) : (
<button
type="submit"
disabled={!hasFile || processing}
className="bg-primary text-primary-foreground hover:bg-primary/90 disabled:bg-muted w-full rounded-md px-4 py-2 text-sm font-medium disabled:cursor-not-allowed"
>
{hasMultiple ? format(ts.submitBatch, { count: files.length }) : ts.submit}
</button>
)}
{/* Error */}
{error && (
<div className="text-destructive rounded-md bg-red-50 p-3 text-sm dark:bg-red-950">
{error}
</div>
)}
{/* Download */}
{downloadUrl && (
<a
href={downloadUrl}
download
className="bg-primary text-primary-foreground hover:bg-primary/90 flex items-center justify-center gap-2 rounded-md px-4 py-2 text-sm font-medium"
>
<Download className="h-4 w-4" />
{t.common.download}
</a>
)}
</form>
);
}
+2
View File
@@ -71,6 +71,7 @@ import {
MicVocal,
Minimize2,
Music,
PaintBucket,
Palette,
PenLine,
PenTool,
@@ -183,6 +184,7 @@ export const ICON_MAP: Record<string, LucideIcon> = {
MicVocal,
Minimize2,
Music,
PaintBucket,
Palette,
PenLine,
PenTool,
+5
View File
@@ -101,6 +101,11 @@ export const TOOL_DISPLAY_MODES: Record<string, DisplayMode> = {
"transparency-fixer": "before-after",
"content-aware-resize": "side-by-side",
"ai-canvas-expand": "before-after",
"ocr-pdf": "no-comparison",
"transcribe-audio": "no-comparison",
"auto-subtitles": "no-comparison",
"background-replace": "no-comparison",
"blur-background": "no-comparison",
// Video tools
"convert-video": "media-player",
+28
View File
@@ -269,6 +269,9 @@ const UpscaleSettings = lazy(() =>
const OcrSettings = lazy(() =>
import("@/components/tools/ocr-settings").then((m) => ({ default: m.OcrSettings })),
);
const OcrPdfSettings = lazy(() =>
import("@/components/tools/ocr-pdf-settings").then((m) => ({ default: m.OcrPdfSettings })),
);
const BlurFacesSettings = lazy(() =>
import("@/components/tools/blur-faces-settings").then((m) => ({
default: m.BlurFacesSettings,
@@ -344,6 +347,26 @@ const ColorBlindnessSettings = lazy(() =>
default: m.ColorBlindnessSettings,
})),
);
const TranscribeAudioSettings = lazy(() =>
import("@/components/tools/transcribe-audio-settings").then((m) => ({
default: m.TranscribeAudioSettings,
})),
);
const AutoSubtitlesSettings = lazy(() =>
import("@/components/tools/auto-subtitles-settings").then((m) => ({
default: m.AutoSubtitlesSettings,
})),
);
const BackgroundReplaceSettings = lazy(() =>
import("@/components/tools/background-replace-settings").then((m) => ({
default: m.BackgroundReplaceSettings,
})),
);
const BlurBackgroundSettings = lazy(() =>
import("@/components/tools/blur-background-settings").then((m) => ({
default: m.BlurBackgroundSettings,
})),
);
const ConvertVideoSettings = lazy(() =>
import("@/components/tools/convert-video-settings").then((m) => ({
default: m.ConvertVideoSettings,
@@ -952,6 +975,7 @@ const ENTRY_CONFIG: ReadonlyArray<[string, RegistryEntryConfig]> = [
["remove-background", { Settings: RemoveBgSettings }],
["upscale", { Settings: UpscaleSettings }],
["ocr", { Settings: OcrSettings }],
["ocr-pdf", { accept: ".pdf", Settings: OcrPdfSettings }],
["blur-faces", { Settings: BlurFacesSettings }],
["enhance-faces", { Settings: EnhanceFacesSettings }],
["erase-object", { Settings: EraseObjectSettingsWrapper as never }],
@@ -965,6 +989,10 @@ const ENTRY_CONFIG: ReadonlyArray<[string, RegistryEntryConfig]> = [
["transparency-fixer", { Settings: TransparencyFixerSettings }],
["content-aware-resize", { Settings: ContentAwareResizeSettings }],
["ai-canvas-expand", { Settings: AiCanvasExpandSettings }],
["transcribe-audio", { accept: AUDIO_INPUTS.join(","), Settings: TranscribeAudioSettings }],
["auto-subtitles", { accept: VIDEO_INPUTS.join(","), Settings: AutoSubtitlesSettings }],
["background-replace", { Settings: BackgroundReplaceSettings }],
["blur-background", { Settings: BlurBackgroundSettings }],
// Video tools
["convert-video", { accept: VIDEO_INPUTS.join(","), Settings: ConvertVideoSettings }],