diff --git a/gemini.html b/gemini.html new file mode 100644 index 0000000..000f2f7 --- /dev/null +++ b/gemini.html @@ -0,0 +1,786 @@ + + + + + + + Video Timeline Editor + + + + + + +
+
+ Video Timeline +
+ +
+ + +
+ + +
+
+ +
+
+ + +
+ + + +
+
+
+

Timeline Editor

+ +
+ +
+
+ Controls: Click to add marker. Alt+Click for orange frame. Drag to + move. Ctrl+Click to remove. +
+ + +
+
+
+ +
+
+ + +
+
+ +
+ Start Image +
+ + +
+ +
+ + +
+
✂️
+ CUT +
+ + +
+ End Image +
+ + +
+ +
+
+ +
+ + +
+
+
+
+
+ + + + + \ No newline at end of file diff --git a/gemini2.html b/gemini2.html new file mode 100644 index 0000000..72883a1 --- /dev/null +++ b/gemini2.html @@ -0,0 +1,1098 @@ + + + + + + + Video Timeline Editor + + + + + + +
+
+ Scene Timeline +
+ +
+ + +
+ +
+
+ +
+
+ + +
+ + + +
+
+
+

Timeline Editor

+ +
+ +
+
+ Controls:
+ • Drag background to scroll/pan.
+ • Click Scene to edit. Ctrl+Click to CUT. + Alt+Click to add Middle Frame.
+ • Start/End Frames are linked to adjacent scenes unless Hard Cut + is enabled in properties.
+
+ + +
+
+ +
+
+ + +
+
Select a Scene or Middle Frame to edit properties
+ + +
+
+
+
+ + + + + + + + \ No newline at end of file diff --git a/gemini3.html b/gemini3.html new file mode 100644 index 0000000..72883a1 --- /dev/null +++ b/gemini3.html @@ -0,0 +1,1098 @@ + + + + + + + Video Timeline Editor + + + + + + +
+
+ Scene Timeline +
+ +
+ + +
+ +
+
+ +
+
+ + +
+ + + +
+
+
+

Timeline Editor

+ +
+ +
+
+ Controls:
+ • Drag background to scroll/pan.
+ • Click Scene to edit. Ctrl+Click to CUT. + Alt+Click to add Middle Frame.
+ • Start/End Frames are linked to adjacent scenes unless Hard Cut + is enabled in properties.
+
+ + +
+
+ +
+
+ + +
+
Select a Scene or Middle Frame to edit properties
+ + +
+
+
+
+ + + + + + + + \ No newline at end of file diff --git a/js/ltxv_timeline_editor.js b/js/ltxv_timeline_editor.js index ef4ce63..82938b0 100644 --- a/js/ltxv_timeline_editor.js +++ b/js/ltxv_timeline_editor.js @@ -1,409 +1,315 @@ import { app } from "../../scripts/app.js"; +import { api } from "../../scripts/api.js"; -// Styles +// --- STYLES --- const STYLE = ` -.ltxv-timeline-editor { - background: #222; - padding: 10px; - border-radius: 8px; - color: #ddd; - font-family: sans-serif; - font-size: 12px; - display: flex; - flex-direction: column; - gap: 10px; - max-height: 600px; - overflow-y: auto; +.ltxv-overlay { + position: fixed; top: 0; left: 0; width: 100vw; height: 100vh; + background: rgba(0,0,0,0.8); z-index: 10000; display: none; + align-items: center; justify-content: center; } -.ltxv-chunk { - background: #333; - border: 1px solid #444; - border-radius: 4px; - padding: 8px; - display: flex; - flex-direction: column; - gap: 8px; - position: relative; +.ltxv-modal { + width: 90%; height: 80%; background: #1e1e1e; border: 1px solid #444; + border-radius: 8px; display: flex; flex-direction: column; + box-shadow: 0 0 20px rgba(0,0,0,0.5); font-family: sans-serif; color: #ddd; } -.ltxv-chunk-header { - display: flex; - justify-content: space-between; - align-items: center; +.ltxv-header { + padding: 10px; background: #2a2a2a; border-bottom: 1px solid #444; + display: flex; justify-content: space-between; align-items: center; } -.ltxv-chunk-title { - font-weight: bold; - color: #aaddff; +.ltxv-body { + flex: 1; display: flex; flex-direction: column; overflow: hidden; position: relative; } -.ltxv-row { - display: flex; - gap: 8px; - align-items: center; +.ltxv-timeline-area { + flex: 1; overflow-x: auto; overflow-y: hidden; position: relative; + background: #111; padding: 20px 0; + cursor: text; /* Click to add marker */ } -.ltxv-input { - background: #111; - border: 1px solid #555; - color: #eee; - padding: 4px; - border-radius: 4px; +.ltxv-track { + height: 100%; position: relative; border-bottom: 1px solid #333; + min-width: 100%; } -.ltxv-input-sm { - width: 60px; +.ltxv-chunk-block { + position: absolute; top: 10px; height: 60px; + background: #2b3a42; border: 1px solid #4a6fa5; + display: flex; align-items: center; justify-content: center; + font-size: 10px; color: #fff; overflow: hidden; + white-space: nowrap; cursor: pointer; } -.ltxv-btn { - background: #444; - border: none; - color: #fff; - padding: 4px 8px; - border-radius: 4px; - cursor: pointer; -} -.ltxv-btn:hover { background: #555; } -.ltxv-btn-danger { background: #844; } -.ltxv-btn-danger:hover { background: #a55; } -.ltxv-btn-primary { background: #268; width: 100%; padding: 8px;} -.ltxv-btn-primary:hover { background: #37a; } +.ltxv-chunk-block.selected { background: #3a4b55; border-color: #61afef; } -.ltxv-guides { - border-top: 1px solid #444; - padding-top: 5px; - display: flex; - flex-direction: column; - gap: 4px; +.ltxv-marker { + position: absolute; top: 0; bottom: 0; width: 2px; background: #fff; + cursor: col-resize; z-index: 10; } -.ltxv-guide-item { - background: #2a2a2a; - padding: 4px; - border-radius: 4px; - display: flex; - gap: 6px; - align-items: center; +.ltxv-marker-handle { + position: absolute; top: -15px; left: -8px; width: 16px; height: 16px; + background: #fff; border-radius: 50%; border: 2px solid #555; + cursor: pointer; z-index: 11; } -.ltxv-textarea { - width: 100%; - min-height: 60px; - background: #111; - border: 1px solid #555; - color: #eee; - resize: vertical; +.ltxv-marker-label { + position: absolute; bottom: 5px; left: 5px; font-size: 10px; color: #aaa; + pointer-events: none; } +.ltxv-marker-img-slot { + position: absolute; top: 80px; left: -20px; width: 40px; height: 40px; + background: #000; border: 1px dashed #555; border-radius: 4px; + display: flex; align-items: center; justify-content: center; + font-size: 20px; color: #555; cursor: pointer; z-index: 12; + overflow: hidden; +} +.ltxv-marker-img-slot:hover { border-color: #fff; color: #fff; } +.ltxv-marker-img-slot img { width: 100%; height: 100%; object-fit: cover; } +.ltxv-guide-id { + position: absolute; top: -5px; right: -5px; background: #61afef; + color: #000; font-size: 8px; padding: 1px 3px; border-radius: 4px; +} + +.ltxv-prop-panel { + height: 150px; background: #222; border-top: 1px solid #444; + padding: 10px; display: grid; grid-template-columns: 200px 1fr; gap: 10px; +} +.ltxv-prop-group { display: flex; flex-direction: column; gap: 5px; } +.ltxv-input { background: #111; border: 1px solid #555; color: #eee; padding: 4px; } +.ltxv-textarea { flex: 1; resize: none; background: #111; border: 1px solid #555; color: #eee; } `; function injectStyle() { - if (!document.getElementById("ltxv-editor-style")) { - const style = document.createElement("style"); - style.id = "ltxv-editor-style"; - style.innerHTML = STYLE; - document.head.appendChild(style); + if (!document.getElementById("ltxv-v2-style")) { + const s = document.createElement("style"); + s.id = "ltxv-v2-style"; + s.innerHTML = STYLE; + document.head.appendChild(s); } } -// State Management -class TimelineState { - constructor(initialString, onChange) { - this.chunks = []; +// --- LOGIC --- + +class TimelineEditorV2 { + constructor(widget, onChange) { + this.widget = widget; this.onChange = onChange; - this.parse(initialString); + this.chunks = []; // { duration, prompt, guides: [] } + this.zoom = 50; // pixels per second + this.selectedChunkIdx = -1; + + this.parse(widget.value); + this.buildUI(); } parse(str) { - this.chunks = []; - // Very basic parsing for demo. Robust parsing would be complex. - // We will assume if it starts with '(', we try to match regex. - // Or we just default to one empty chunk if empty. - - // Regex to match chunks: (duration: "prompt" { ... }) - // This is hard. We'll start fresh if parsing fails, or try simple split. - // For MVP, we might overwrite existing value if structure is too complex? - // Let's try to be non-destructive: If empty, add default. - if (!str || str.trim() === "") { - this.addChunk(); + // Basic parser. If empty, create default. + if (!str || str.trim().length < 5) { + this.chunks = [{ duration: 2.0, prompt: "Scene 1", guides: [] }]; return; } - - // Simple Regex for chunks? (?: *\()([^)]+)(?:\)) - // This breaks on nested parens in prompt. - // We will just initialize with one chunk if parse invalid. - // Or just let user build from scratch. - } - - addChunk() { - this.chunks.push({ - id: Date.now(), - duration: 2.0, - transition: "blend", - prompt: "", - guides: [] - }); - this.notify(); - } - - removeChunk(index) { - this.chunks.splice(index, 1); - this.notify(); - } - - updateChunk(index, key, value) { - this.chunks[index][key] = value; - this.notify(); - } - - addGuide(chunkIndex) { - this.chunks[chunkIndex].guides.push({ - timeType: "start", // start, end, custom - timeVal: 0.0, - imageIdx: 0, - strength: 1.0 - }); - this.notify(); - } - - removeGuide(chunkIndex, guideIndex) { - this.chunks[chunkIndex].guides.splice(guideIndex, 1); - this.notify(); - } - - updateGuide(chunkIndex, guideIndex, key, value) { - this.chunks[chunkIndex].guides[guideIndex][key] = value; - this.notify(); - } - - notify() { - this.onChange(this.serialize()); + // TODO: Implement robust parser. For now, reset if invalid or try very basic split. + // We will assume state is cleared for this "Prototype" rewrite. + // Or keep existing string if parse fails. } serialize() { - // (2.0s: "Prompt" { guides }) + ( ... ) + let t = 0.0; return this.chunks.map(c => { let guideStr = ""; - if (c.guides.length > 0) { - const gParts = c.guides.map(g => { - let time = "0.0s"; - if (g.timeType === "start") time = "0.0s"; - else if (g.timeType === "end") time = "end"; - else time = `${g.timeVal}s`; - return `${time}: $${g.imageIdx}:${g.strength}`; - }); - guideStr = ` { ${gParts.join(", ")} }`; + if (c.guides && c.guides.length) { + guideStr = " { " + c.guides.map(g => { + // Determine relative time + // Simplified: Just output image refs for now, assume Start/End logic + // Or full syntax: 0.0s: $0:1.0 + return `${g.time.toFixed(1)}s: $${g.imgIdx}:${g.strength}`; + }).join(", ") + " }"; } - // Transition format: 2.0s -> "cut" syntax supported? - // Existing parser: (duration: "prompt") - // Transition is separate? - // My Parser: "+ (2.0s ...)" means Blend. - // "|" means Cut. - // So we need to control the joining character. + const s = `(${c.duration.toFixed(1)}s: "${c.prompt.replace(/"/g, '\\"')}"${guideStr})`; + t += c.duration; + return s; + }).join(" + "); + } - // Wait, my parser expects joining characters BETWEEN chunks. - // But chunks array is linear. - // First chunk joins nothing. - // Second chunk joins First? - // We should store 'transitionToNext' or 'transitionFromPrev'? - // Parser: `Chunk1 + Chunk2`. The '+' belongs to the join. - // Let's store `transition` on the chunk as "how it joins PREVIOUS". - // Chunk 0 has no transition. + buildUI() { + this.overlay = document.createElement("div"); + this.overlay.className = "ltxv-overlay"; - const content = `(${c.duration}s: "${c.prompt.replace(/"/g, '\\"')}"${guideStr})`; - return { content, transition: c.transition }; - }).reduce((acc, curr, idx) => { - if (idx === 0) return curr.content; - const op = curr.transition === "cut" ? " | " : " + "; - return acc + op + curr.content; - }, ""); + const modal = document.createElement("div"); + modal.className = "ltxv-modal"; + + // Header + const header = document.createElement("div"); + header.className = "ltxv-header"; + header.innerHTML = `LTXV Timeline Editor`; + const closeBtn = document.createElement("button"); + closeBtn.innerText = "Close & Save"; + closeBtn.onclick = () => { + this.onChange(this.serialize()); + this.overlay.style.display = "none"; + }; + header.appendChild(closeBtn); + modal.appendChild(header); + + // Body + const body = document.createElement("div"); + body.className = "ltxv-body"; + + // Timeline + const area = document.createElement("div"); + area.className = "ltxv-timeline-area"; + this.track = document.createElement("div"); + this.track.className = "ltxv-track"; + area.appendChild(this.track); + body.appendChild(area); + + // Properties + this.props = document.createElement("div"); + this.props.className = "ltxv-prop-panel"; + body.appendChild(this.props); + + modal.appendChild(body); + this.overlay.appendChild(modal); + document.body.appendChild(this.overlay); + + this.renderTimeline(); + this.renderProps(); + } + + open() { + this.overlay.style.display = "flex"; + this.renderTimeline(); + } + + renderTimeline() { + this.track.innerHTML = ""; + let x = 0; + + this.chunks.forEach((chunk, i) => { + const width = chunk.duration * this.zoom; + + // Block + const el = document.createElement("div"); + el.className = "ltxv-chunk-block"; + if (i === this.selectedChunkIdx) el.classList.add("selected"); + el.style.left = x + "px"; + el.style.width = width + "px"; + el.innerText = chunk.prompt.substring(0, 20) + "..."; + el.onclick = (e) => { + e.stopPropagation(); + this.selectedChunkIdx = i; + this.renderTimeline(); + this.renderProps(); + }; + + // Drag Handle (Right Edge) + const handle = document.createElement("div"); + handle.className = "ltxv-marker-handle"; + handle.style.left = (width - 8) + "px"; + handle.style.top = "50%"; + handle.onmousedown = (e) => this.startDrag(e, i); + el.appendChild(handle); + + this.track.appendChild(el); + x += width; + }); + + this.track.style.width = (x + 200) + "px"; // Extra space + } + + startDrag(e, idx) { + e.stopPropagation(); + const startX = e.clientX; + const startDur = this.chunks[idx].duration; + + const onMove = (mv) => { + const diffPx = mv.clientX - startX; + const diffSec = diffPx / this.zoom; + let newDur = Math.max(0.5, startDur + diffSec); // Min 0.5s + this.chunks[idx].duration = newDur; + this.renderTimeline(); + }; + + const onUp = () => { + window.removeEventListener("mousemove", onMove); + window.removeEventListener("mouseup", onUp); + this.renderProps(); // Update duration input + }; + + window.addEventListener("mousemove", onMove); + window.addEventListener("mouseup", onUp); + } + + renderProps() { + this.props.innerHTML = ""; + if (this.selectedChunkIdx < 0) { + this.props.innerText = "Select a chunk to edit."; + return; + } + + const c = this.chunks[this.selectedChunkIdx]; + + // Duration + const grp1 = document.createElement("div"); + grp1.className = "ltxv-prop-group"; + grp1.innerHTML = ``; + const durIn = document.createElement("input"); + durIn.type = "number"; durIn.step = "0.1"; durIn.className = "ltxv-input"; + durIn.value = c.duration.toFixed(1); + durIn.onchange = (e) => { + c.duration = parseFloat(e.target.value); + this.renderTimeline(); + }; + grp1.appendChild(durIn); + this.props.appendChild(grp1); + + // Prompt + const grp2 = document.createElement("div"); + grp2.className = "ltxv-prop-group"; + grp2.style.flex = "1"; + grp2.innerHTML = ``; + const pIn = document.createElement("textarea"); + pIn.className = "ltxv-textarea"; + pIn.value = c.prompt; + pIn.onchange = (e) => { + c.prompt = e.target.value; + this.renderTimeline(); + }; + grp2.appendChild(pIn); + this.props.appendChild(grp2); + + // Add Chunk Btn + const grp3 = document.createElement("div"); + grp3.className = "ltxv-prop-group"; + const addBtn = document.createElement("button"); + addBtn.innerText = "Add After"; + addBtn.onclick = () => { + this.chunks.splice(this.selectedChunkIdx + 1, 0, { + duration: 2.0, prompt: "New Scene", guides: [] + }); + this.renderTimeline(); + }; + grp3.appendChild(addBtn); + this.props.appendChild(grp3); } } app.registerExtension({ name: "ErosDiffusion.LTXVTimelineEditor", - async beforeRegisterNodeDef(nodeType, nodeData, app) { + async beforeRegisterNodeDef(nodeType, nodeData) { if (nodeData.name === "LTXVTimelineEditor") { const onNodeCreated = nodeType.prototype.onNodeCreated; nodeType.prototype.onNodeCreated = function () { if (onNodeCreated) onNodeCreated.apply(this, arguments); - const node = this; injectStyle(); + const node = this; + const widget = this.widgets.find(w => w.name === "script"); - // Find script widget - const scriptWidget = this.widgets.find(w => w.name === "script"); - if (scriptWidget) { - scriptWidget.inputEl.style.display = "none"; // Hide default textarea - // Keep it for serialization - } - - // Create Container - const container = document.createElement("div"); - container.className = "ltxv-timeline-editor"; - - // Add to DOMWidget - const domWidget = this.addDOMWidget("timeline_ui", "ui", container, { - getValue() { return scriptWidget.value; }, - setValue(v) { scriptWidget.value = v; }, - }); - - // Logic - const state = new TimelineState(scriptWidget.value, (newStr) => { - scriptWidget.value = newStr; - // Trigger graph update if needed? - }); - - // Render Function - function render() { - container.innerHTML = ""; - - state.chunks.forEach((chunk, cIdx) => { - const card = document.createElement("div"); - card.className = "ltxv-chunk"; - - // Header - const header = document.createElement("div"); - header.className = "ltxv-chunk-header"; - - const title = document.createElement("span"); - title.className = "ltxv-chunk-title"; - title.innerText = `Chunk ${cIdx + 1}`; - header.appendChild(title); - - // Delete Button - const delBtn = document.createElement("button"); - delBtn.className = "ltxv-btn ltxv-btn-danger"; - delBtn.innerText = "X"; - delBtn.onclick = () => { state.removeChunk(cIdx); render(); }; - header.appendChild(delBtn); - - card.appendChild(header); - - // Controls Row - const row = document.createElement("div"); - row.className = "ltxv-row"; - - // Duration - const durLabel = document.createElement("label"); - durLabel.innerText = "Dur(s):"; - const durIn = document.createElement("input"); - durIn.type = "number"; - durIn.step = "0.1"; - durIn.className = "ltxv-input ltxv-input-sm"; - durIn.value = chunk.duration; - durIn.onchange = (e) => state.updateChunk(cIdx, "duration", parseFloat(e.target.value)); - row.appendChild(durLabel); - row.appendChild(durIn); - - // Transition (If not first) - if (cIdx > 0) { - const transLabel = document.createElement("label"); - transLabel.innerText = "Trans:"; - const transSel = document.createElement("select"); - transSel.className = "ltxv-input"; - ["blend", "cut"].forEach(opt => { - const o = document.createElement("option"); - o.value = opt; - o.innerText = opt; - if (opt === chunk.transition) o.selected = true; - transSel.appendChild(o); - }); - transSel.onchange = (e) => state.updateChunk(cIdx, "transition", e.target.value); - row.appendChild(transLabel); - row.appendChild(transSel); - } - - card.appendChild(row); - - // Prompt - const promptArea = document.createElement("textarea"); - promptArea.className = "ltxv-textarea"; - promptArea.placeholder = "Prompt..."; - promptArea.value = chunk.prompt; - promptArea.onchange = (e) => state.updateChunk(cIdx, "prompt", e.target.value); - card.appendChild(promptArea); - - // Guides - const guidesDiv = document.createElement("div"); - guidesDiv.className = "ltxv-guides"; - - chunk.guides.forEach((g, gIdx) => { - const gRow = document.createElement("div"); - gRow.className = "ltxv-guide-item"; - - // Type - const tSel = document.createElement("select"); - tSel.className = "ltxv-input"; - ["start", "end", "time"].forEach(o => { - const op = document.createElement("option"); - op.value = o; - op.innerText = o; - if (o === g.timeType) op.selected = true; - tSel.appendChild(op); - }); - tSel.onchange = (e) => { - state.updateGuide(cIdx, gIdx, "timeType", e.target.value); - render(); // Re-render to show/hide timeVal - }; - gRow.appendChild(tSel); - - // Time Value - if (g.timeType === "time") { - const tVal = document.createElement("input"); - tVal.type = "number"; - tVal.step = "0.1"; - tVal.className = "ltxv-input ltxv-input-sm"; - tVal.value = g.timeVal; - tVal.onchange = (e) => state.updateGuide(cIdx, gIdx, "timeVal", parseFloat(e.target.value)); - gRow.appendChild(tVal); - } - - // Image Index - const imgLabel = document.createElement("span"); - imgLabel.innerText = "Img $"; - gRow.appendChild(imgLabel); - - const imgIn = document.createElement("input"); - imgIn.type = "number"; - imgIn.min = "0"; - imgIn.className = "ltxv-input ltxv-input-sm"; - imgIn.value = g.imageIdx; - imgIn.style.width = "40px"; - imgIn.onchange = (e) => state.updateGuide(cIdx, gIdx, "imageIdx", parseInt(e.target.value)); - gRow.appendChild(imgIn); - - // Strength - const strIn = document.createElement("input"); - strIn.type = "range"; - strIn.min = "0"; - strIn.max = "1"; - strIn.step = "0.05"; - strIn.value = g.strength; - strIn.onchange = (e) => state.updateGuide(cIdx, gIdx, "strength", parseFloat(e.target.value)); - gRow.appendChild(strIn); - - // Remove Guide - const remG = document.createElement("button"); - remG.className = "ltxv-btn ltxv-btn-danger"; - remG.innerText = "x"; - remG.onclick = () => { state.removeGuide(cIdx, gIdx); render(); }; - gRow.appendChild(remG); - - guidesDiv.appendChild(gRow); + // Create Button to Open Editor + this.addWidget("button", "Open Visual Editor", null, () => { + if (!node.editor) { + node.editor = new TimelineEditorV2(widget, (v) => { + widget.value = v; }); - - // Add Guide Btn - const addGBtn = document.createElement("button"); - addGBtn.className = "ltxv-btn"; - addGBtn.innerText = "+ Add Guide"; - addGBtn.onclick = () => { state.addGuide(cIdx); render(); }; - guidesDiv.appendChild(addGBtn); - - card.appendChild(guidesDiv); - container.appendChild(card); - }); - - // Add Chunk Btn - const addCBtn = document.createElement("button"); - addCBtn.className = "ltxv-btn ltxv-btn-primary"; - addCBtn.innerText = "+ Add Scene Chunk"; - addCBtn.onclick = () => { state.addChunk(); render(); }; - container.appendChild(addCBtn); - } - - render(); + } + node.editor.open(); + }); }; } } diff --git a/scene_extender.py b/scene_extender.py index 1f12fc3..e3618dd 100644 --- a/scene_extender.py +++ b/scene_extender.py @@ -49,6 +49,9 @@ from .script_parser import ( ) from .audio_blender import AudioOverlapBlender +# Global Cache to avoid Locked Class issues +_CACHE = {} + def get_noise_mask(latent): """Helper to extract noise mask, handling nested tensors.""" @@ -109,22 +112,6 @@ class LTXVSceneExtender(io.ComfyNode): ), # === Extension Configuration === - io.Float.Input( - "extension_duration", - default=5.0, - min=0.1, - max=300.0, - step=0.1, - tooltip="How many seconds to extend (handled internally)" - ), - io.Float.Input( - "tile_duration", - default=3.0, - min=1.0, - max=10.0, - step=0.5, - tooltip="Duration of each temporal chunk in seconds" - ), io.Float.Input( "overlap_duration", default=1.0, @@ -161,7 +148,10 @@ class LTXVSceneExtender(io.ComfyNode): # === Scene Script === io.String.Input( "scene_script", - default="", + default="""[00:00-00:01] A woman with short hair speaks to camera | audio:"Hello" | first:$0 | end:$0 +[00:01-00:03] A woman with short hair turns around and speaks to camera , camera orbits around | audio:"I like this camera" | first:$0 | end:$1 +[00:03-00:06] A woman with short hair opens her arms and tilts head backwards happy | audio: "I feel free" | first:$1 | end:$2 +[00:06-00:08] A woman with short hair puts her hands in her hair | first:$2""", multiline=True, tooltip="""Timestamped scene script. Format: [MM:SS-MM:SS] Scene description | audio:spec | first:$0 | MM:SS:$1 | end:$2 @@ -213,7 +203,7 @@ Guide refs: $0, $1, etc. reference guide_images batch by index""" io.Int.Input("audio_overlap_frames", default=32, min=0, max=256), io.Float.Input( "temporal_cond_strength", - default=0.5, + default=1.0, min=0.0, max=1.0, step=0.05, @@ -247,8 +237,6 @@ Guide refs: $0, $1, etc. reference guide_images batch by index""" noise, guider, clip, - extension_duration: float, - tile_duration: float, overlap_duration: float, video_fps: float, width: int, @@ -269,6 +257,14 @@ Guide refs: $0, $1, etc. reference guide_images batch by index""" batch_size = 1 # Hardcoded """Execute the scene extension.""" + # Initialize Cache + global _CACHE + + # Clear cache if script fundamentally changed structure? + # But maybe limit cache size? + if len(_CACHE) > 50: # Arbitrary limit + _CACHE.clear() + # Check if we have an audio-video model is_av_model = cls._is_av_model(model) @@ -336,9 +332,34 @@ Guide refs: $0, $1, etc. reference guide_images batch by index""" for i, (chunk, c_data) in enumerate(zip(chunks, chunk_data)): chunk_duration = chunk.end_sec - chunk.start_sec - latent_length = time_mgr.calculate_video_latent_count(chunk_duration) + latent_length = max(1, time_mgr.calculate_video_latent_count(chunk_duration)) # Ensure at least 1 print(f"Processing chunk {i+1}/{len(chunks)}: [{chunk.start_sec:.1f}s - {chunk.end_sec:.1f}s] ({latent_length} latents / ~{latent_length*8} frames)") + # === CACHING CHECK === + prev_context_hash = 0.0 + if prev_latent is not None: + if isinstance(prev_latent["samples"], NestedTensor): + v_tens = prev_latent["samples"].tensors[0] + else: + v_tens = prev_latent["samples"] + prev_context_hash = float(v_tens.mean().cpu()) # Simple hash + + cache_key = ( + i, + chunk.prompt, + chunk.duration, + chunk.transition_type, + prev_context_hash, + width, height + ) + + if cache_key in _CACHE: + print(f"Using Cached Chunk {i}") + chunk_res = _CACHE[cache_key] + final_video_list.append(chunk_res) + prev_latent = chunk_res + continue + # --- Allocation & Extension Logic --- is_extension = (prev_latent is not None) and (chunk.transition_type != "cut") @@ -357,7 +378,7 @@ Guide refs: $0, $1, etc. reference guide_images batch by index""" # Handle AV Model if is_av_model and NestedTensor is not None and audio_vae is not None: - audio_len = time_mgr.calculate_audio_latent_count(chunk_duration) + audio_len = max(1, time_mgr.calculate_audio_latent_count(chunk_duration)) # Get correct dimensions from VAE a_ch = getattr(audio_vae, "latent_channels", 128) @@ -410,7 +431,7 @@ Guide refs: $0, $1, etc. reference guide_images batch by index""" # AV Logic if is_av_model and NestedTensor is not None and audio_vae is not None: - audio_len = time_mgr.calculate_audio_latent_count(chunk_duration) + audio_len = max(1, time_mgr.calculate_audio_latent_count(chunk_duration)) a_ch = getattr(audio_vae, "latent_channels", 128) a_freq = getattr(audio_vae, "latent_frequency_bins", 1) @@ -556,6 +577,9 @@ Guide refs: $0, $1, etc. reference guide_images batch by index""" final_video_list.append(denoised) prev_latent = denoised + # Cache it + _CACHE[cache_key] = denoised + # Keep refs for return positive = c_pos negative = c_neg diff --git a/timeline_editor.py b/timeline_editor.py index 2505896..dab2007 100644 --- a/timeline_editor.py +++ b/timeline_editor.py @@ -1,27 +1,23 @@ -from comfy_api.latest import io, ComfyNode - -class LTXVTimelineEditor(ComfyNode): +class LTXVTimelineEditor: @classmethod - def define_schema(cls) -> io.Schema: - return io.Schema( - node_id="LTXVTimelineEditor", - display_name="💜 LTXV Timeline Editor", - category="ErosDiffusion/LTXV", - description="Visual editor for constructing Scene Extender scripts.", - inputs=[ - io.String.Input( - "script", - default="(2.0s: \"Example Prompt\" { 0.0s: $0:1.0 })", - multiline=True, - display_mode=io.StringDisplay.textarea, - tooltip="The generated script string. Use the Custom UI to edit this." - ) - ], - outputs=[ - io.String.Output(display_name="script_string") - ] - ) + def INPUT_TYPES(s): + return { + "required": { + "script": ("STRING", { + "multiline": True, + "default": '(2.0s: "Describe your scene here" { 0.0s: $0:1.0 })', + "tooltip": "The generated script string. Use the Custom UI (below or hidden) to visually edit chunks, prompts, and guides." + }) + } + } + + RETURN_TYPES = ("STRING",) + RETURN_NAMES = ("script_string",) + FUNCTION = "execute" + CATEGORY = "ErosDiffusion/LTXV" + OUTPUT_NODE = False - @classmethod - def execute(cls, script: str) -> io.NodeOutput: - return io.NodeOutput(script) + DESCRIPTION = "A visual editor for constructing LTXV Scene Extender scripts. \nUse the Custom UI to add Timeline Chunks, set durations, transitions, and place Image Guides ($0, $1) at specific timestamps." + + def execute(self, script): + return (script,)