dsh-vidi 0.9.1 → 0.9.2

package/lib/client.js

@@ -358,6 +358,13 @@ const Icon = {
 			rx: "2"
 		})
 	}),
+	merge: () => /* @__PURE__ */ (0, react_jsx_runtime.jsx)("svg", {
+		viewBox: "0 0 24 24",
+		width: "16",
+		height: "16",
+		...P,
+		children: /* @__PURE__ */ (0, react_jsx_runtime.jsx)("path", { d: "M7 4v5a5 5 0 0 0 5 5 5 5 0 0 1 5 5v1M17 4v5a5 5 0 0 1-3 4.6M12 17l-3 3M12 17l3 3" })
+	}),
 	timer: () => /* @__PURE__ */ (0, react_jsx_runtime.jsxs)("svg", {
 		viewBox: "0 0 24 24",
 		width: "18",
@@ -35416,9 +35423,15 @@ function History({ t, ui, onOpen }) {
 			}),
 			/* @__PURE__ */ (0, react_jsx_runtime.jsx)("div", {
 				className: "hl",
-				children: items.map((v) => /* @__PURE__ */ (0, react_jsx_runtime.jsxs)("button", {
+				children: items.map((v) => /* @__PURE__ */ (0, react_jsx_runtime.jsxs)("div", {
+					role: "button",
+					tabIndex: 0,
 					className: "hi",
 					onClick: () => onOpen(v),
+					onKeyDown: (e) => {
+						if (e.key === "Enter") onOpen(v);
+					},
+					title: v.live && v.source === "url" && !v.running ? t("hist.goLive") : void 0,
 					children: [
 						/* @__PURE__ */ (0, react_jsx_runtime.jsxs)("span", {
 							className: "th",
@@ -35450,6 +35463,14 @@ function History({ t, ui, onOpen }) {
 						v.running && /* @__PURE__ */ (0, react_jsx_runtime.jsx)("span", {
 							className: "chip",
 							children: t("hist.running")
+						}),
+						v.live && v.source === "url" && !v.running && /* @__PURE__ */ (0, react_jsx_runtime.jsx)("button", {
+							className: "dshl-link replay",
+							onClick: (e) => {
+								e.stopPropagation();
+								onOpen(v, true);
+							},
+							children: t("hist.replay")
 						})
 					]
 				}, v.id))
@@ -35930,6 +35951,20 @@ function LiveView(props) {
 		}).catch(() => job));
 	};
 	/** Switching modes while a video is open takes effect right away: no delay stops transcribing, the others (re)start it. */
+	const switchEngine = async (id) => {
+		setEngine(id);
+		if (!job || subs.state === "off" || job.status === "error") return;
+		try {
+			setJob(await post("subs", {
+				...options(),
+				provider: id,
+				jobId: job.id,
+				action: "start"
+			}));
+		} catch (e) {
+			setError(e.message);
+		}
+	};
 	const changeMode = (m) => {
 		if (m === mode) return;
 		setMode(m);
@@ -36209,8 +36244,13 @@ function LiveView(props) {
 			return "camera";
 		}
 	};
-	const openPast = async (v) => {
+	const openPast = async (v, replay = false) => {
 		setError("");
+		if (!replay && v.live && v.source === "url" && v.url && !v.running) {
+			if (v.conversationId !== sessionId && props.onOpenVideo && await props.onOpenVideo(v.conversationId, v.id, v.url)) return;
+			start(v.url, "url");
+			return;
+		}
 		try {
 			if (v.conversationId !== sessionId && props.onOpenVideo && await props.onOpenVideo(v.conversationId, v.id)) return;
 			const j = await post("reopen", {
@@ -36237,7 +36277,7 @@ function LiveView(props) {
 			pageMode && /* @__PURE__ */ (0, react_jsx_runtime.jsx)(History, {
 				t,
 				ui,
-				onOpen: (v) => void openPast(v)
+				onOpen: (v, replay) => void openPast(v, replay)
 			}),
 			srcTab === "push" && armed?.rtmp && /* @__PURE__ */ (0, react_jsx_runtime.jsx)("div", {
 				className: "dshl-pushinfo",
@@ -36313,6 +36353,34 @@ function LiveView(props) {
 		openBench: setBench
 	} : void 0;
 	/** Subtitle lines and glossary: under the player on the video page, beside it in a conversation. */
+	const [splitting, setSplitting] = (0, react.useState)(null);
+	const lineKey = (k) => k === "src" || k === spokenBase ? "src" : k;
+	const mergeLine = async (id) => {
+		try {
+			await post("lines", {
+				jobId: job.id,
+				action: "merge",
+				id
+			});
+		} catch (e) {
+			setError(e.message);
+		}
+	};
+	const splitLine = async (id, key, at) => {
+		setSplitting(null);
+		if (at <= 0) return;
+		try {
+			await post("lines", {
+				jobId: job.id,
+				action: "split",
+				id,
+				key,
+				at
+			});
+		} catch (e) {
+			setError(e.message);
+		}
+	};
 	const transcriptBox = /* @__PURE__ */ (0, react_jsx_runtime.jsxs)(react_jsx_runtime.Fragment, { children: [
 		/* @__PURE__ */ (0, react_jsx_runtime.jsxs)("div", {
 			className: "dshl-sidetabs",
@@ -36377,9 +36445,45 @@ function LiveView(props) {
 					children: [shown.length === 0 && /* @__PURE__ */ (0, react_jsx_runtime.jsx)("div", {
 						className: "dshl-muted pad",
 						children: emptyLines
-					}), shown.map((c) => /* @__PURE__ */ (0, react_jsx_runtime.jsxs)("button", {
+					}), shown.map((c, idx) => splitting?.id === c.id ? /* @__PURE__ */ (0, react_jsx_runtime.jsxs)("div", {
+						className: "dshl-line splitting",
+						children: [
+							/* @__PURE__ */ (0, react_jsx_runtime.jsx)("span", {
+								className: "ts",
+								children: CLOCK(c.start)
+							}),
+							/* @__PURE__ */ (0, react_jsx_runtime.jsxs)("span", {
+								className: "txt",
+								children: [/* @__PURE__ */ (0, react_jsx_runtime.jsx)("span", {
+									className: "main cutline",
+									children: Array.from((splitting.key === "src" ? c.src : c.tr?.[splitting.key]) ?? "").map((ch, i) => /* @__PURE__ */ (0, react_jsx_runtime.jsx)("span", {
+										className: "ch",
+										onClick: () => void splitLine(c.id, splitting.key, i),
+										children: ch
+									}, i))
+								}), /* @__PURE__ */ (0, react_jsx_runtime.jsxs)("span", {
+									className: "hint",
+									children: [
+										t("lines.splitHint"),
+										" ",
+										/* @__PURE__ */ (0, react_jsx_runtime.jsx)("button", {
+											className: "dshl-link",
+											onClick: () => setSplitting(null),
+											children: t("player.cancel")
+										})
+									]
+								})]
+							}),
+							/* @__PURE__ */ (0, react_jsx_runtime.jsx)("span", {})
+						]
+					}, c.id) : /* @__PURE__ */ (0, react_jsx_runtime.jsxs)("div", {
+						role: "button",
+						tabIndex: 0,
 						className: `dshl-line ${c.id === activeId ? "active" : ""} ${c.polished ? "ok" : "raw"}`,
 						onClick: () => player.current?.seek(c.start),
+						onKeyDown: (e) => {
+							if (e.key === "Enter") player.current?.seek(c.start);
+						},
 						title: c.polished ? t("tr.polished") : t("tr.raw"),
 						children: [
 							/* @__PURE__ */ (0, react_jsx_runtime.jsx)("span", {
@@ -36400,6 +36504,27 @@ function LiveView(props) {
 								className: "mark",
 								"aria-hidden": "true",
 								children: c.polished && /* @__PURE__ */ (0, react_jsx_runtime.jsx)(Icon.check, {})
+							}),
+							c.final && /* @__PURE__ */ (0, react_jsx_runtime.jsxs)("span", {
+								className: "tools",
+								onClick: (e) => e.stopPropagation(),
+								children: [idx < shown.length - 1 && shown[idx + 1]?.final && /* @__PURE__ */ (0, react_jsx_runtime.jsx)("button", {
+									className: "ghost",
+									title: t("lines.merge"),
+									onClick: () => void mergeLine(c.id),
+									children: /* @__PURE__ */ (0, react_jsx_runtime.jsx)(Icon.merge, {})
+								}), (() => {
+									const k = shownLines.find((x) => (lineText(c, x) ?? "").length > 1) ?? "src";
+									return Array.from(lineText(c, k) ?? c.src).length > 1 && /* @__PURE__ */ (0, react_jsx_runtime.jsx)("button", {
+										className: "ghost",
+										title: t("lines.split"),
+										onClick: () => setSplitting({
+											id: c.id,
+											key: lineKey(k)
+										}),
+										children: /* @__PURE__ */ (0, react_jsx_runtime.jsx)(Icon.scissors, {})
+									});
+								})()]
 							})
 						]
 					}, c.id))]
@@ -36739,30 +36864,52 @@ function LiveView(props) {
 										t
 									}) })
 								},
-								...!job.vod && job.status === "running" ? [{
-									id: "mode",
+								{
+									id: "how",
 									icon: /* @__PURE__ */ (0, react_jsx_runtime.jsx)(Icon.timer, {}),
-									title: t("set.mode"),
-									label: t(`mode.${mode}`),
-									content: /* @__PURE__ */ (0, react_jsx_runtime.jsx)("div", {
+									title: t("how.title"),
+									label: [!job.vod ? t(`mode.${mode}`) : "", engineName(usableEngine())].filter(Boolean).join(" · "),
+									content: /* @__PURE__ */ (0, react_jsx_runtime.jsxs)("div", {
 										className: "dshl-modemenu",
 										children: [
-											"live",
-											"realtime",
-											"polished"
-										].map((m) => /* @__PURE__ */ (0, react_jsx_runtime.jsxs)("button", {
-											className: `mm ${mode === m ? "on" : ""}`,
-											onClick: () => changeMode(m),
-											children: [/* @__PURE__ */ (0, react_jsx_runtime.jsxs)("span", {
-												className: "n",
-												children: [t(`mode.${m}`), mode === m && /* @__PURE__ */ (0, react_jsx_runtime.jsx)(Icon.check, {})]
-											}), /* @__PURE__ */ (0, react_jsx_runtime.jsx)("span", {
-												className: "d",
-												children: t(`mode.${m}.tip`)
-											})]
-										}, m))
+											!job.vod && /* @__PURE__ */ (0, react_jsx_runtime.jsxs)(react_jsx_runtime.Fragment, { children: [/* @__PURE__ */ (0, react_jsx_runtime.jsx)("div", {
+												className: "mh",
+												children: t("set.mode")
+											}), [
+												"live",
+												"realtime",
+												"polished"
+											].map((m) => /* @__PURE__ */ (0, react_jsx_runtime.jsxs)("button", {
+												className: `mm ${mode === m ? "on" : ""}`,
+												onClick: () => changeMode(m),
+												children: [/* @__PURE__ */ (0, react_jsx_runtime.jsxs)("span", {
+													className: "n",
+													children: [t(`mode.${m}`), mode === m && /* @__PURE__ */ (0, react_jsx_runtime.jsx)(Icon.check, {})]
+												}), /* @__PURE__ */ (0, react_jsx_runtime.jsx)("span", {
+													className: "d",
+													children: t(`mode.${m}.tip`)
+												})]
+											}, m))] }),
+											/* @__PURE__ */ (0, react_jsx_runtime.jsx)("div", {
+												className: "mh",
+												children: t("set.engine")
+											}),
+											/* @__PURE__ */ (0, react_jsx_runtime.jsx)(Pick, {
+												value: usableEngine(),
+												onChange: (id) => void switchEngine(id),
+												options: providers.filter((p) => p.ready).map((p) => ({
+													value: p.id,
+													label: engineName(p.id),
+													hint: p.model
+												}))
+											}),
+											/* @__PURE__ */ (0, react_jsx_runtime.jsx)("p", {
+												className: "mt",
+												children: t("how.more")
+											})
+										]
 									})
-								}] : [],
+								},
 								{
 									id: "restream",
 									icon: /* @__PURE__ */ (0, react_jsx_runtime.jsx)(Icon.push, {}),
@@ -37599,6 +37746,13 @@ const en = {
 	"dev.blocked": "No picture or sound from the device: DeepSeek Harness needs the system’s permission. Open System Settings, switch DeepSeek Harness on, then retry.",
 	"dev.denied": "DeepSeek Harness needs permission to use the {d}. Open System Settings, switch it on in the list, then retry.",
 	"dev.and": " and ",
+	"how.title": "How subtitles are made",
+	"how.more": "More services and keys are in Settings.",
+	"lines.merge": "Join with the next line",
+	"lines.split": "Split this line",
+	"lines.splitHint": "Click where the line should break.",
+	"hist.goLive": "Watch live again",
+	"hist.replay": "Replay",
 	"hist.title": "Recent videos",
 	"hist.untitled": "Untitled video",
 	"hist.live": "live",
@@ -38053,6 +38207,13 @@ const zh = {
 	"dev.blocked": "设备没有传来画面和声音：DeepSeek Harness 需要系统授权。点「打开系统设置」，在列表里打开 DeepSeek Harness，再点重试。",
 	"dev.denied": "DeepSeek Harness 需要{d}权限。点「打开系统设置」，在列表里打开它，再点重试。",
 	"dev.and": "和",
+	"how.title": "字幕怎么做",
+	"how.more": "更多服务和 key 在设置里。",
+	"lines.merge": "和下一句合并",
+	"lines.split": "拆成两句",
+	"lines.splitHint": "点一下要断开的位置。",
+	"hist.goLive": "回到直播",
+	"hist.replay": "回放",
 	"hist.title": "最近的视频",
 	"hist.untitled": "未命名视频",
 	"hist.live": "直播",
@@ -38743,6 +38904,21 @@ html[data-platform='darwin'] .dshl.page .dshl-top :is(button, input, label, a, .
 .dshl-history .mm { overflow: hidden; text-overflow: ellipsis; white-space: nowrap; color: var(--faint); font-size: 12px; }
 .dshl-history .chip { flex: none; padding: 2px 8px; border-radius: 999px; background: color-mix(in srgb, var(--live) 14%, transparent); color: var(--live); font-size: 11.5px; font-weight: 600; }
 .dshl-history .more { margin: 8px 0 0 6px; font-size: 12.5px; }
+.dshl-modemenu .mh { margin: 6px 4px 2px; color: rgba(255,255,255,.55); font-size: 11.5px; }
+.dshl-modemenu .mh:first-child { margin-top: 0; }
+.dshl-modemenu .dshl-select { width: 100%; }
+.dshl-modemenu .mt { margin: 2px 4px 0; color: rgba(255,255,255,.45); font-size: 11.5px; }
+.dshl-line { position: relative; }
+.dshl-line .tools { position: absolute; top: 4px; right: 4px; display: none; gap: 2px; padding: 2px; border-radius: 7px; background: var(--bg); box-shadow: 0 1px 4px rgba(0,0,0,.12); }
+.dshl-line:hover .tools, .dshl-line:focus-within .tools { display: flex; }
+.dshl-line .tools .ghost { display: grid; place-items: center; width: 26px; height: 24px; border: 0; border-radius: 5px; background: none; color: var(--muted); cursor: pointer; }
+.dshl-line .tools svg { width: 15px; height: 15px; }
+.dshl-line .tools .ghost:hover { background: var(--hover); color: var(--ink); }
+.dshl-line.splitting { cursor: default; background: color-mix(in srgb, var(--accent) 8%, transparent); }
+.dshl-line .cutline .ch { cursor: col-resize; border-left: 1px solid transparent; }
+.dshl-line .cutline .ch:hover { border-left-color: var(--accent); background: color-mix(in srgb, var(--accent) 12%, transparent); }
+.dshl-line .hint { color: var(--faint); font-size: 12px; }
+.dshl-history .replay { flex: none; font-size: 12.5px; }
 .dshl-modemenu { display: grid; gap: 4px; width: 300px; max-width: 100%; }
 .dshl-modemenu .mm { display: grid; gap: 3px; padding: 9px 10px; border: 0; border-radius: 8px; background: none; color: inherit; text-align: left; cursor: pointer; }
 .dshl-modemenu .mm:hover { background: rgba(255,255,255,.08); }
@@ -39187,9 +39363,9 @@ function videoHandlers(sessions, workspaces, sessionId, fallbackTitle) {
 			}
 		},
 		/** Open a past video in its own conversation; false when that conversation is gone. */
-		onOpenVideo: async (conversationId, jobId) => {
+		onOpenVideo: async (conversationId, jobId, pending) => {
 			if (!conversationId || !sessions.has(conversationId)) return false;
-			await fetch("/live/api/reopen", {
+			if (!pending) await fetch("/live/api/reopen", {
 				method: "POST",
 				headers: { "content-type": "application/json" },
 				body: JSON.stringify({
@@ -39199,7 +39375,8 @@ function videoHandlers(sessions, workspaces, sessionId, fallbackTitle) {
 			}).catch(() => {});
 			await openPage(sessions, {
 				sessionId: conversationId,
-				workspaces
+				workspaces,
+				pending
 			});
 			return true;
 		},

package/lib/index.mjs

@@ -1990,6 +1990,32 @@ const AHEAD_MAX_LINES = 10;
 const seekAfter = (skip) => skip > .01 ? ["-ss", skip.toFixed(3)] : [];
 const SOI = Buffer.from([255, 216]);
 const EOI = Buffer.from([255, 217]);
+const CJK_CHAR = /[\u3040-\u30ff\u3400-\u9fff\uac00-\ud7af\uff00-\uffef\u3000-\u303f]/;
+/** Two pieces of one language as one line: no space between CJK text, a space between words otherwise. */
+function joinText(a, b) {
+	const x = a.trim(), y = b.trim();
+	if (!x || !y) return x || y;
+	return CJK_CHAR.test(x.at(-1)) || CJK_CHAR.test(y[0]) ? x + y : `${x} ${y}`;
+}
+/** Where to cut a text near `ratio` of its length: after a pause mark if one is close, else between words. */
+function cutNear(cs, ratio) {
+	const n = cs.length;
+	if (n < 2) return n;
+	const target = Math.round(n * ratio);
+	let best = -1, bestScore = Infinity;
+	for (let i = 1; i < n; i++) {
+		const prev = cs[i - 1], here = cs[i];
+		const pause = /[，。、；：！？,.;:!?]/.test(prev);
+		const word = here === " " || CJK_CHAR.test(prev) || CJK_CHAR.test(here);
+		if (!pause && !word) continue;
+		const score = Math.abs(i - target) - (pause ? n * .12 : 0);
+		if (score < bestScore) {
+			bestScore = score;
+			best = i;
+		}
+	}
+	return best > 0 ? best : Math.max(1, Math.min(n - 1, target));
+}
 var LiveJob = class LiveJob extends EventEmitter {
 	opts;
 	deps;
@@ -2052,7 +2078,7 @@ var LiveJob = class LiveJob extends EventEmitter {
 			targets: () => this.targets(),
 			still: (at) => this.still(at),
 			onLine: (cue, src, tr, speaker) => {
-				if (!this.cues.includes(cue)) return;
+				if (!this.cues.includes(cue) || cue.edited) return;
 				this.deps.glossary.notice(src);
 				cue.src = src;
 				if (speaker) cue.speaker = speaker;
@@ -2860,6 +2886,76 @@ var LiveJob = class LiveJob extends EventEmitter {
 		})).join("\n") + "\n");
 		this.emit("cues", this.cues);
 	}
+	/** Join a finished line with the next one (the user's own line breaks). */
+	mergeCue(id) {
+		const i = this.cues.findIndex((c) => c.id === id);
+		const a = this.cues[i], b = this.cues[i + 1];
+		if (!a || !b || !a.final || !b.final) return false;
+		a.src = joinText(a.src, b.src);
+		const keys = /* @__PURE__ */ new Set([...Object.keys(a.tr ?? {}), ...Object.keys(b.tr ?? {})]);
+		if (keys.size) a.tr = Object.fromEntries([...keys].map((k) => [k, joinText(a.tr?.[k] ?? "", b.tr?.[k] ?? "")]));
+		a.end = Math.max(a.end, b.end);
+		a.polished = !!a.polished && !!b.polished;
+		a.edited = true;
+		this.cues.splice(i + 1, 1);
+		appendFileSync(join(this.deps.dir, "cues.jsonl"), JSON.stringify({
+			id: b.id,
+			dropped: true
+		}) + "\n");
+		this.record(a);
+		this.emit("cues", this.cues);
+		return true;
+	}
+	/**
+	* Split a finished line in two where the user pointed: `key` is the text they pointed in ('src' or a
+	* language) and `at` the character position. The other texts are cut at the matching place (at a
+	* pause or a word nearby), and the time in proportion.
+	*/
+	splitCue(id, key, at) {
+		const i = this.cues.findIndex((c) => c.id === id);
+		const a = this.cues[i];
+		if (!a || !a.final) return false;
+		const text = key === "src" ? a.src : a.tr?.[key];
+		if (!text) return false;
+		const chars = Array.from(text);
+		const cut = Math.max(1, Math.min(chars.length - 1, Math.round(at)));
+		const ratio = cut / chars.length;
+		const pieces = (t, own) => {
+			const cs = Array.from(t);
+			const k = own ? cut : cutNear(cs, ratio);
+			return [cs.slice(0, k).join("").trim(), cs.slice(k).join("").trim()];
+		};
+		const [srcA, srcB] = pieces(a.src, key === "src");
+		if (!srcA || !srcB) return false;
+		const trA = {}, trB = {};
+		for (const [k, v] of Object.entries(a.tr ?? {})) {
+			const [x, y] = pieces(v, k === key);
+			trA[k] = x;
+			trB[k] = y;
+		}
+		const mid = a.start + (a.end - a.start) * ratio;
+		const b = {
+			id: this.nextId++,
+			start: mid,
+			end: a.end,
+			src: srcB,
+			tr: Object.keys(trB).length ? trB : void 0,
+			lang: a.lang,
+			final: true,
+			polished: a.polished,
+			speaker: a.speaker,
+			edited: true
+		};
+		a.src = srcA;
+		if (Object.keys(trA).length) a.tr = trA;
+		a.end = mid;
+		a.edited = true;
+		this.cues.splice(i + 1, 0, b);
+		this.record(a);
+		this.record(b);
+		this.emit("cues", this.cues);
+		return true;
+	}
 	/** Append a finished cue (with its translation, if any) to cues.jsonl. */
 	record(cue) {
 		appendFileSync(join(this.deps.dir, "cues.jsonl"), JSON.stringify(cue) + "\n");
@@ -6774,6 +6870,8 @@ function apply(ctx, config) {
 				if (!job) return send(res, 404, { error: "no such job" });
 				if (b.action === "pause") await job.pauseSubs();
 				else {
+					const cur = job.summary();
+					if ((typeof b.provider === "string" && b.provider !== cur.provider || typeof b.mode === "string" && modeOf(b.mode) !== cur.mode) && cur.subs?.state === "running") await job.pauseSubs();
 					const opts = {};
 					if (Array.isArray(b.tgtLangs)) opts.tgtLangs = b.tgtLangs.map(String).slice(0, 4);
 					if (typeof b.srcLang === "string") opts.srcLang = b.srcLang;
@@ -7050,6 +7148,13 @@ function apply(ctx, config) {
 				createReadStream(file).pipe(res);
 				return;
 			}
+			if (route === "lines" && req.method === "POST") {
+				const b = await readJson(req);
+				const job = jobs.get(String(b.jobId));
+				if (!job) return send(res, 404, { error: "no such job" });
+				const ok = b.action === "merge" ? job.mergeCue(Number(b.id)) : b.action === "split" ? job.splitCue(Number(b.id), String(b.key ?? "src"), Number(b.at)) : false;
+				return send(res, ok ? 200 : 400, ok ? { ok: true } : { error: "cannot change that line" });
+			}
 			if (route === "reopen" && req.method === "POST") {
 				const b = await readJson(req);
 				const job = loadJob(String(b.jobId));

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "dsh-vidi",
-  "version": "0.9.1",
+  "version": "0.9.2",
   "description": "A video tool for DeepSeek Harness: watch live streams, videos and podcasts with real-time subtitles in any language, then clip and download",
   "type": "module",
   "license": "PolyForm-Noncommercial-1.0.0",