Free, open-source AI video generator for Mac. Run LTX-2.5 and MiniMax H3 (Hailuo) locally on Apple Silicon: text-to-video and image-to-video with sound, lip-sync, character LoRA training, songs from lyrics (YuE2) and image editing (Qwen-Image-Edit). One-click install via Pinokio. Offline: no cloud, no API key, no NVIDIA GPU.
Python ai-filmmakingai-videoai-video-generatorapple-siliconhailuoimage-to-videolip-synclocal-ailora-trainingltx-2ltx-2-5ltx-videomacosminimax-h3mlxmusic-generationoffline-aipinokiotext-to-videovideo-generation
feat(form-pane): composer-led redesign — hero composer card, quick settings strip, sticky action footer
mrbizarro9c14eb03a7cb
diff --git a/mlx_ltx_panel.py b/mlx_ltx_panel.pyindex bf2f73f..dacbcb9 100644--- a/mlx_ltx_panel.py+++ b/mlx_ltx_panel.py@@ -8447,6 +8447,364 @@ HTML = r"""<!doctype html>details summary:hover { color: var(--text); }details[open] summary { margin-bottom: 6px; }+ /* ============================================================+ REDESIGN — composer-led form-pane (Y2.000)+ Goal: collapse the noisy h2-stack into a clear hierarchy:+ 1. Mode bar (compact segmented strip, no label)+ 2. Composer card (refs + prompt + inline tools, hero element)+ 3. Quick settings strip (Quality + Duration + Seed in one row)+ 4. Customize+Advanced disclosure (folded — power users only)+ 5. LoRAs disclosure (kept)+ 6. Sticky action footer (Generate + queue chips)+ Uses tokens; no hardcoded colors.+ ============================================================ */++ /* Mode segmented bar — replaces <h2>Mode</h2> + pill-group cols-5.+ Sits at the top of the form. Compact, equal-weight, tactile. The+ active chip uses the accent fill so the user always knows what mode+ they're composing in. Subtitle text sits inside each chip — kept+ small so the bar height stays low. */+ .mode-bar {+ display: grid;+ grid-template-columns: repeat(5, 1fr);+ gap: 4px;+ padding: 4px;+ background: var(--panel);+ border: 1px solid var(--border);+ border-radius: var(--r-md);+ margin-bottom: 14px;+ }+ .mode-bar .mode-chip {+ width: 100%;+ display: flex; flex-direction: column; align-items: center; gap: 1px;+ padding: 8px 4px;+ background: transparent;+ border: 1px solid transparent;+ border-radius: var(--r-sm);+ color: var(--muted);+ font-size: 12px; font-weight: 600;+ cursor: pointer;+ transition: var(--t-fast);+ letter-spacing: 0;+ text-align: center;+ }+ .mode-bar .mode-chip:hover {+ color: var(--text);+ background: rgba(255,255,255,0.03);+ }+ .mode-bar .mode-chip.active {+ background: var(--accent-dim);+ border-color: var(--accent);+ color: var(--accent-bright);+ box-shadow: var(--shadow-1);+ }+ .mode-bar .mode-chip .mc-sub {+ font-size: 9.5px;+ font-weight: 500;+ color: var(--muted);+ letter-spacing: 0.02em;+ text-transform: none;+ opacity: 0.85;+ }+ .mode-bar .mode-chip.active .mc-sub {+ color: var(--accent-bright);+ opacity: 0.7;+ }++ /* Composer card — the hero of the form. Wraps the reference picker(s)+ (mode-conditional) + prompt textarea + an inline tools footer+ (Enhance, HDR, No-music, Avoid disclosure). One elevated surface so+ the prompt reads as the primary input. */+ .composer-card {+ background: var(--panel);+ border: 1px solid var(--border-strong);+ border-radius: var(--r-lg);+ padding: 12px 12px 10px;+ margin-bottom: 14px;+ box-shadow: var(--shadow-1);+ transition: border-color var(--t-base), box-shadow var(--t-base);+ }+ .composer-card:focus-within {+ border-color: var(--accent);+ box-shadow: var(--ring), var(--shadow-1);+ }+ /* Slot for reference pickers; sits above the prompt. Mode-only blocks+ hide entirely when not relevant (T2V), so the composer card only+ carries refs when the mode needs them. The wrapper applies a+ bottom-margin only when at least one .mode-only.show is inside —+ use :has() (Safari ≥ 15.4 supports it) so T2V mode collapses the+ slot to zero with no orphan margin. */+ .composer-refs { margin-bottom: 0; }+ .composer-refs:has(> .mode-only.show) { margin-bottom: 12px; }+ .composer-refs > .mode-only.show + .mode-only.show { margin-top: 8px; }+ /* Prompt textarea inside the composer — hero treatment. Larger min+ height, no hardcoded border (the card carries it), placeholder color+ softer, font slightly larger so it reads as the main input. */+ .composer-card .composer-prompt {+ width: 100%;+ min-height: 110px;+ padding: 4px 4px 6px;+ font-size: 14px; line-height: 1.5;+ background: transparent;+ border: 0;+ color: var(--text);+ resize: vertical;+ font-family: inherit;+ }+ .composer-card .composer-prompt::placeholder {+ color: var(--muted);+ opacity: 0.7;+ }+ .composer-card .composer-prompt:focus { outline: none; box-shadow: none; }+ /* Tools strip inside the composer — Enhance + toggles + Avoid+ disclosure. Sits flush at the bottom with a hairline separator. */+ .composer-tools {+ margin-top: 8px;+ padding-top: 10px;+ border-top: 1px solid var(--border);+ display: flex; align-items: center; gap: 8px; flex-wrap: wrap;+ }+ .composer-tools .ct-spacer { flex: 1 1 auto; }+ /* Avoid disclosure — collapsed by default. When opened, the+ textarea sits outside the composer-card (in its own panel below)+ so the prompt stays clean. */+ .avoid-row {+ margin-top: 10px;+ display: none;+ }+ .avoid-row.show { display: block; }+ .avoid-row .avoid-label {+ display: flex; align-items: center; gap: 8px;+ margin: 0 0 5px 2px;+ font-size: 10.5px; font-weight: 600;+ color: var(--muted);+ text-transform: uppercase; letter-spacing: 0.08em;+ }+ .avoid-row .avoid-label .avoid-hint {+ font-weight: 400; font-size: 10px;+ letter-spacing: 0; text-transform: none;+ color: var(--muted); opacity: 0.7;+ }+ /* Tiny icon-style toggle inside the composer tools row. Same idiom as+ .toggle-pill but icon-first so it stays compact. */+ .ct-link {+ width: auto;+ padding: 4px 9px;+ background: transparent;+ border: 1px solid transparent;+ color: var(--muted);+ font-size: 11px; font-weight: 500;+ border-radius: var(--r-xs);+ cursor: pointer;+ transition: var(--t-fast);+ }+ .ct-link:hover { color: var(--text); background: rgba(255,255,255,0.04); }+ .ct-link.active {+ color: var(--accent-bright);+ background: var(--accent-dim);+ border-color: var(--accent);+ }++ /* Quiet h2 — used inside the composer card / customize body where the+ loud uppercase mono label would be excessive. */+ .composer-card h2,+ .customize-section h2,+ .composer-refs h2 {+ font-size: 10px;+ margin: 0 0 6px;+ color: var(--muted);+ text-transform: uppercase;+ letter-spacing: 0.08em;+ font-weight: 600;+ }++ /* Quick-settings row — Quality picker + Duration + Seed in one strip.+ The Quality picker shrinks to a compact 4-up segmented control; the+ two number fields sit on the right with inline labels above them. */+ .quick-settings {+ display: grid;+ grid-template-columns: 1fr;+ gap: 12px;+ margin-bottom: 14px;+ }+ .quick-settings .qs-label {+ display: flex; align-items: baseline; justify-content: space-between;+ margin-bottom: 6px;+ }+ .quick-settings .qs-label .qs-name {+ font-size: 10px; font-weight: 600;+ color: var(--muted);+ text-transform: uppercase; letter-spacing: 0.08em;+ }+ .quick-settings .qs-label .qs-meta {+ font-size: 11px;+ color: var(--muted);+ font-family: ui-monospace, SFMono-Regular, Menlo, monospace;+ }+ /* Compact quality strip — looks like the mode bar but tighter, 4 cols.+ Each chip carries a name on top of an optional secondary spec. */+ .quality-strip {+ display: grid;+ grid-template-columns: repeat(4, 1fr);+ gap: 4px;+ padding: 4px;+ background: var(--panel);+ border: 1px solid var(--border);+ border-radius: var(--r-md);+ }+ .quality-strip .q-chip {+ display: flex; flex-direction: column; align-items: center; gap: 2px;+ padding: 7px 4px;+ background: transparent;+ border: 1px solid transparent;+ border-radius: var(--r-sm);+ color: var(--muted);+ font-size: 12px; font-weight: 600;+ cursor: pointer;+ transition: var(--t-fast);+ width: 100%;+ }+ .quality-strip .q-chip:hover:not(.disabled) {+ color: var(--text);+ background: rgba(255,255,255,0.03);+ }+ .quality-strip .q-chip.active {+ background: var(--accent-dim);+ border-color: var(--accent);+ color: var(--accent-bright);+ }+ .quality-strip .q-chip.disabled {+ opacity: 0.4;+ cursor: not-allowed;+ }+ .quality-strip .q-chip .q-spec {+ font-size: 9.5px; font-weight: 500;+ color: inherit; opacity: 0.7;+ font-family: ui-monospace, SFMono-Regular, Menlo, monospace;+ letter-spacing: 0;+ }++ /* Inline mini-fields — Duration / Frames / Seed. Smaller padding,+ label sits inside the field's label slot above. */+ .mini-fields {+ display: grid;+ grid-template-columns: repeat(3, 1fr);+ gap: 8px;+ }+ .mini-fields .mf-cell {+ display: flex; flex-direction: column; gap: 4px;+ }+ .mini-fields .mf-label {+ font-size: 10px; font-weight: 600;+ color: var(--muted);+ text-transform: uppercase; letter-spacing: 0.08em;+ }+ .mini-fields input {+ width: 100%;+ padding: 7px 9px;+ font-size: 13px;+ background: var(--panel-2);+ border: 1px solid var(--border);+ border-radius: var(--r-sm);+ color: var(--text);+ }+ .mini-fields input:focus {+ outline: none;+ border-color: var(--accent);+ background: var(--bg-2);+ box-shadow: var(--ring);+ }++ /* Sticky action footer — Generate stays at the bottom of the form-pane+ no matter how far the user scrolls. The form scrolls under it. The+ queue chip strip sits ABOVE the Generate button so it's never the+ primary touch target. */+ .form-pane {+ /* allow the footer to overlap the scroll content; padding keeps the+ last form section visible above the footer. */+ padding-bottom: 132px;+ }+ .form-action-footer {+ position: sticky;+ bottom: 0;+ left: 0; right: 0;+ /* break out horizontally so the footer spans edge-to-edge of the+ form-pane (form-pane has padding 16px); vertically use auto top+ margin so the form content above it sets natural spacing. */+ margin: 16px -16px 0;+ padding: 14px 16px 16px;+ /* Soft fade into the form content above so the sticky footer+ doesn't read as a hard bar — content scrolling under it gradually+ becomes invisible. The transparent top + opaque bottom means+ scrolling content fades out as it passes under the footer. */+ background: linear-gradient(+ 180deg,+ rgba(0, 6, 26, 0) 0%,+ rgba(0, 6, 26, 0.85) 22%,+ var(--bg) 50%,+ var(--bg) 100%+ );+ border-top: 1px solid var(--border);+ z-index: 5;+ backdrop-filter: blur(8px);+ -webkit-backdrop-filter: blur(8px);+ }+ .form-action-footer .actions {+ display: grid;+ grid-template-columns: 1fr auto;+ gap: 8px;+ margin: 0;+ }+ .form-action-footer .actions .primary {+ padding: 12px 14px;+ font-size: 14px; font-weight: 600;+ border-radius: var(--r-md);+ }+ .form-action-footer .actions .danger {+ padding: 12px 18px;+ border-radius: var(--r-md);+ width: auto;+ }+ /* Queue chip strip — sits ABOVE the Generate button. Compact icon-led+ chips for batch / pause / clear so they don't compete with Generate. */+ .form-action-footer .queue-strip {+ display: flex; gap: 6px;+ margin-bottom: 10px;+ flex-wrap: wrap;+ align-items: center;+ }+ .form-action-footer .queue-strip .qchip {+ width: auto;+ padding: 4px 10px;+ font-size: 11px;+ background: transparent;+ border: 1px solid var(--border);+ color: var(--muted);+ border-radius: var(--r-pill);+ cursor: pointer;+ transition: var(--t-fast);+ display: inline-flex; align-items: center; gap: 5px;+ font-weight: 500;+ }+ .form-action-footer .queue-strip .qchip:hover {+ color: var(--text);+ border-color: var(--accent);+ background: rgba(47,129,247,0.06);+ }+ .form-action-footer .queue-strip .qchip-spacer { flex: 1 1 auto; }+ .form-action-footer .queue-strip .qchip-derived {+ font-size: 10.5px;+ color: var(--muted);+ font-family: ui-monospace, SFMono-Regular, Menlo, monospace;+ padding: 4px 0;+ }+ .form-action-footer .queue-strip .qchip-derived strong {+ color: var(--accent-bright); font-weight: 600;+ }++ /* When form-pane is in agent mode, drop the sticky footer padding. */+ body[data-workflow="agent"] .form-pane { padding-bottom: 0; }+/* ===== STAGE (PLAYER + CAROUSEL) ===== */.stage-pane {background: linear-gradient(180deg, #0a0d12 0%, var(--panel) 100%);@@ -13625,60 +13983,25 @@ HTML = r"""<!doctype html><a class="models-inline-link" onclick="openModelsModal()">Manage all models →</a></div>- <!-- Mode picker — lifted OUT of #genForm so it stays visible when- the user switches to Image Studio (which hides #genForm). Without- this, once a user clicks Studio they had no way back to T2V/I2V- short of refreshing. The hidden #mode input stays inside #genForm- (further below) so the video form's FormData picks it up.- Used to be inside #genForm at the top of the form. -->- <h2 style="margin-top:0">Mode</h2>- <!-- 5 mode pills on ONE row. The previous markup carried a stale- `cols-2` class plus an inline 4-col override, which left the- 5th pill (Studio) on its own line and visually disconnected.- Switching to `cols-5` keeps all 5 pills equal-width on the- same row so Studio reads as a peer of T2V/I2V/FFLF/Extend. -->- <div class="pill-group cols-5" id="modeGroup">- <button type="button" class="pill-btn" data-mode="t2v"><span>Text</span><span class="sub">prompt → video</span></button>- <button type="button" class="pill-btn" data-mode="i2v"><span>Image</span><span class="sub">image + prompt</span></button>- <button type="button" class="pill-btn" data-mode="keyframe"><span>FFLF</span><span class="sub">first + last frame</span></button>- <button type="button" class="pill-btn" data-mode="extend"><span>Extend</span><span class="sub">continue a clip</span></button>- <button type="button" class="pill-btn" data-mode="image" title="Open Image Studio — generate stills (Qwen-Image-Edit-2509 multi-ref + others)"><span>Studio</span><span class="sub">stills + library</span></button>+ <!-- Mode bar — segmented chip strip. Lifted OUT of #genForm so it+ stays visible when the user switches to Image Studio (which hides+ #genForm). Without this, once a user clicks Studio they had no+ way back to T2V/I2V short of refreshing. The hidden #mode input+ stays inside #genForm (further below) so the video form's+ FormData picks it up. The element keeps id=modeGroup and each+ chip keeps data-mode + .pill-btn (alongside .mode-chip) so the+ existing setMode() click wiring keeps working unchanged. -->+ <div class="mode-bar pill-group" id="modeGroup">+ <button type="button" class="mode-chip pill-btn" data-mode="t2v">Text<span class="mc-sub sub">prompt → video</span></button>+ <button type="button" class="mode-chip pill-btn" data-mode="i2v">Image<span class="mc-sub sub">image + prompt</span></button>+ <button type="button" class="mode-chip pill-btn" data-mode="keyframe">FFLF<span class="mc-sub sub">first + last</span></button>+ <button type="button" class="mode-chip pill-btn" data-mode="extend">Extend<span class="mc-sub sub">continue a clip</span></button>+ <button type="button" class="mode-chip pill-btn" data-mode="image" title="Open Image Studio — generate stills (Qwen-Image-Edit-2509 multi-ref + others)">Studio<span class="mc-sub sub">stills + library</span></button></div><form id="genForm"><input type="hidden" name="preset_label" id="preset_label" value=""><input type="hidden" name="mode" id="mode" value="t2v">-- <!-- Quality picker (Y1.013): one decision instead of four. Each- button bundles dimensions + model + step count + tier- recommendation, with the actual specs visible on the pill so- power users can read what they're getting. Aspect, custom- W/H, and the experimental Speed setting moved into the- Customize disclosure below. Beginners pick a button; power- users open Customize. -->- <h2>Quality</h2>- <div class="pill-group cols-2 quality-row" id="qualityGroup">- <button type="button" class="pill-btn pill-quality" data-quality="quick">- <span class="ql-name">Quick</span>- <span class="sub ql-spec">640×480 · ~2 min</span>- <span class="ql-tier">Q4 · any Mac</span>- </button>- <button type="button" class="pill-btn pill-quality active" data-quality="balanced">- <span class="ql-name">Balanced</span>- <span class="sub ql-spec">1024×576 → 720p</span>- <span class="ql-tier">Q4 · ~5 min · no crop</span>- </button>- <button type="button" class="pill-btn pill-quality" data-quality="standard">- <span class="ql-name">Standard</span>- <span class="sub ql-spec">1280×704 · ~7 min</span>- <span class="ql-tier">Q4 · standard tier+</span>- </button>- <button type="button" class="pill-btn pill-quality disabled" data-quality="high" id="qualityHigh">- <span class="ql-name">High</span>- <span class="sub ql-spec" id="highSpec">1280×704 · ~12 min</span>- <span class="ql-tier" id="highSub">Q8 not installed</span>- </button>- </div><input type="hidden" name="quality" id="quality" value="balanced"><input type="hidden" name="accel" id="accel" value="off"><input type="hidden" name="temporal_mode" id="temporal_mode" value="native">@@ -13686,157 +14009,220 @@ HTML = r"""<!doctype html><div id="warnBanner" class="warn-banner"></div>- <!-- Mode-specific: image (I2V).- New picker: a clickable / drag-drop tile + a "Recent uploads"- strip below. The raw path input is gone — paths are still set- via the hidden field for form submission, but never typed by the- user. Same component is reused for FFLF Start / End below. -->- <div class="mode-only" id="imageSection">- <h2>Reference image</h2>- <div class="picker" data-key="image">- <div class="picker-drop" id="picker_drop_image">- <div class="picker-empty">- <div class="picker-icon">🖼</div>- <div class="picker-cta">Drop image here, or <strong>click to browse</strong></div>- <div class="hint">PNG / JPG / WEBP — auto cover-crop to model size</div>+ <!-- ============== COMPOSER CARD ==============+ Hero element of the form. Carries the reference picker(s)+ (mode-conditional), the prompt textarea, and an inline tools+ strip (Enhance + HDR + No-music + Avoid disclosure). One+ elevated surface so the prompt reads as the primary input. -->+ <div class="composer-card">+ <!-- Slot for the mode-conditional reference pickers. Each+ .mode-only block is shown/hidden by setMode(); when none are+ active (T2V), the slot collapses (CSS :empty/:not(.show)). -->+ <div class="composer-refs">+ <!-- I2V — single image ref. -->+ <div class="mode-only" id="imageSection">+ <h2>Reference image</h2>+ <div class="picker" data-key="image">+ <div class="picker-drop" id="picker_drop_image">+ <div class="picker-empty">+ <div class="picker-icon">🖼</div>+ <div class="picker-cta">Drop image here, or <strong>click to browse</strong></div>+ <div class="hint">PNG / JPG / WEBP — auto cover-crop to model size</div>+ </div>+ <img class="picker-preview" id="picker_preview_image" alt="" style="display:none">+ <button type="button" class="picker-clear" id="picker_clear_image" title="Clear" style="display:none">×</button>+ </div>+ <input type="file" id="picker_file_image" accept="image/*" style="display:none">+ <input type="hidden" name="image" id="image" value="">+ <div class="picker-recent" id="picker_recent_image_wrap" style="display:none">+ <div class="picker-recent-label">Recent uploads · click to use</div>+ <div class="picker-recent-strip" id="picker_recent_image"></div>+ </div></div>- <img class="picker-preview" id="picker_preview_image" alt="" style="display:none">- <button type="button" class="picker-clear" id="picker_clear_image" title="Clear" style="display:none">×</button>- </div>- <input type="file" id="picker_file_image" accept="image/*" style="display:none">- <input type="hidden" name="image" id="image" value="">- <div class="picker-recent" id="picker_recent_image_wrap" style="display:none">- <div class="picker-recent-label">Recent uploads · click to use</div>- <div class="picker-recent-strip" id="picker_recent_image"></div></div>- </div>- </div>- <!-- Mode-specific: keyframe (FFLF). Two pickers, same component. -->- <div class="mode-only" id="keyframeSection">- <h2>Start frame (frame 0)</h2>- <div class="picker" data-key="start_image">- <div class="picker-drop" id="picker_drop_start_image">- <div class="picker-empty">- <div class="picker-icon">🎬</div>- <div class="picker-cta">Drop the <strong>first frame</strong>, or <strong>click to browse</strong></div>- <div class="hint">This image opens the clip — its aspect picks the output dimensions.</div>+ <!-- FFLF — start + end pickers. -->+ <div class="mode-only" id="keyframeSection">+ <h2>Start frame (frame 0)</h2>+ <div class="picker" data-key="start_image">+ <div class="picker-drop" id="picker_drop_start_image">+ <div class="picker-empty">+ <div class="picker-icon">🎬</div>+ <div class="picker-cta">Drop the <strong>first frame</strong>, or <strong>click to browse</strong></div>+ <div class="hint">This image opens the clip — its aspect picks the output dimensions.</div>+ </div>+ <img class="picker-preview" id="picker_preview_start_image" alt="" style="display:none">+ <button type="button" class="picker-clear" id="picker_clear_start_image" title="Clear" style="display:none">×</button>+ </div>+ <input type="file" id="picker_file_start_image" accept="image/*" style="display:none">+ <input type="hidden" name="start_image" id="start_image" value="">+ <div class="picker-recent" id="picker_recent_start_image_wrap" style="display:none">+ <div class="picker-recent-label">Recent uploads · click to use</div>+ <div class="picker-recent-strip" id="picker_recent_start_image"></div>+ </div></div>- <img class="picker-preview" id="picker_preview_start_image" alt="" style="display:none">- <button type="button" class="picker-clear" id="picker_clear_start_image" title="Clear" style="display:none">×</button>- </div>- <input type="file" id="picker_file_start_image" accept="image/*" style="display:none">- <input type="hidden" name="start_image" id="start_image" value="">- <div class="picker-recent" id="picker_recent_start_image_wrap" style="display:none">- <div class="picker-recent-label">Recent uploads · click to use</div>- <div class="picker-recent-strip" id="picker_recent_start_image"></div>- </div>- </div>- <h2>End frame (last frame)</h2>- <div class="picker" data-key="end_image">- <div class="picker-drop" id="picker_drop_end_image">- <div class="picker-empty">- <div class="picker-icon">🎯</div>- <div class="picker-cta">Drop the <strong>last frame</strong>, or <strong>click to browse</strong></div>- <div class="hint">A close-up here anchors face identity through the clip.</div>+ <h2 style="margin-top:10px">End frame (last frame)</h2>+ <div class="picker" data-key="end_image">+ <div class="picker-drop" id="picker_drop_end_image">+ <div class="picker-empty">+ <div class="picker-icon">🎯</div>+ <div class="picker-cta">Drop the <strong>last frame</strong>, or <strong>click to browse</strong></div>+ <div class="hint">A close-up here anchors face identity through the clip.</div>+ </div>+ <img class="picker-preview" id="picker_preview_end_image" alt="" style="display:none">+ <button type="button" class="picker-clear" id="picker_clear_end_image" title="Clear" style="display:none">×</button>+ </div>+ <input type="file" id="picker_file_end_image" accept="image/*" style="display:none">+ <input type="hidden" name="end_image" id="end_image" value="">+ <div class="picker-recent" id="picker_recent_end_image_wrap" style="display:none">+ <div class="picker-recent-label">Recent uploads · click to use</div>+ <div class="picker-recent-strip" id="picker_recent_end_image"></div>+ </div></div>- <img class="picker-preview" id="picker_preview_end_image" alt="" style="display:none">- <button type="button" class="picker-clear" id="picker_clear_end_image" title="Clear" style="display:none">×</button>++ <div class="hint" style="margin-top:8px">FFLF needs Q8 (auto-selects High quality). The model interpolates between the two frames you provide.</div></div>- <input type="file" id="picker_file_end_image" accept="image/*" style="display:none">- <input type="hidden" name="end_image" id="end_image" value="">- <div class="picker-recent" id="picker_recent_end_image_wrap" style="display:none">- <div class="picker-recent-label">Recent uploads · click to use</div>- <div class="picker-recent-strip" id="picker_recent_end_image"></div>++ <!-- Extend — source-video select + extend-by + direction ++ extend-mode preset. Lives inside the composer card so the+ prompt sits below the source picker, mirroring the I2V flow. -->+ <div class="mode-only" id="extendSection">+ <h2>Source video</h2>+ <select id="extendSrcSelect" onchange="document.getElementById('video_path').value=this.value"></select>+ <input name="video_path" id="video_path" placeholder="/path/to/source.mp4" style="margin-top:6px">+ <div class="row" style="margin-top:8px">+ <div>+ <label class="lbl">Extend by (seconds)</label>+ <!-- min must align to step from value=2: with min=0.4 step=0.5 the+ valid sequence is 0.4, 0.9, 1.4, 1.9, 2.4… — `2` is OFF the+ grid. Chrome blocks the entire form submission silently when+ ANY input fails validation, even hidden ones in inactive+ modes. Generate appeared dead until cocktailpeanut diagnosed+ it. With min=0.5 the sequence is 0.5, 1.0, 1.5, 2.0… and+ value=2 is valid. -->+ <input id="extend_seconds" type="number" value="2" min="0.5" max="10" step="0.5">+ <input type="hidden" name="extend_frames" id="extend_frames" value="6">+ <div class="hint" id="extendDurationHint" style="margin-top:4px">≈ 2.0 s of new content (6 latent frames × 8 video frames at 24 fps)</div>+ </div>+ <div>+ <label class="lbl">Direction</label>+ <select name="extend_direction" id="extend_direction">+ <option value="after" selected>After</option>+ <option value="before">Before</option>+ </select>+ </div>+ </div>+ <label class="lbl" style="margin-top:10px">Extend mode</label>+ <div class="pill-group cols-2" id="extendModeGroup">+ <button type="button" class="pill-btn active" data-extend-mode="fast"><span>Fast</span><span class="sub">12 steps · no CFG · 64 GB safe</span></button>+ <button type="button" class="pill-btn" data-extend-mode="quality"><span>Quality</span><span class="sub">30 steps · CFG 3.0 · 96+ GB</span></button>+ </div>+ <input type="hidden" name="extend_steps" id="extend_steps" value="12">+ <input type="hidden" name="extend_cfg" id="extend_cfg" value="1.0">+ <div class="hint">Each latent ≈ 8 frames (~0.33s). Quality mode runs the upstream defaults but pushes 1280×704 into swap on 64 GB Macs (~2 hr/render). Stick with Fast unless you've got more RAM.</div></div></div>- <div class="hint">FFLF needs Q8 (auto-selects High quality). The model interpolates between the two frames you provide.</div>+ <!-- Prompt — hero textarea. No h2 (the composer-card itself+ announces "this is where you compose"). Audio-cue placeholder+ is the only nudge users get to describe the soundscape;+ documented in the LTX 2.3 paper but unobvious. -->+ <textarea name="prompt" id="prompt" class="composer-prompt"+ placeholder="Describe the scene AND the sound — e.g. wizard in a forest clearing, fireflies spiraling up · low whispered chant, ember crackle, distant owl. Audio is generated jointly with video; without sound cues the model outputs near-silent ambient."></textarea>++ <!-- Tools strip — Enhance + HDR + No-music + Avoid disclosure.+ All inline so they read as part of composing the prompt.+ Avoid is now a toggle-disclosure (collapsed by default) — it's+ optional and most users won't touch it. -->+ <div class="composer-tools">+ <button type="button" class="ghost-btn" id="enhanceBtn" onclick="enhancePrompt()" title="Use Gemma to rewrite your prompt in the style LTX 2.3 was trained on">✨ Enhance with Gemma</button>+ <button type="button" class="ct-link" id="avoidToggleBtn" onclick="toggleAvoidRow()" title="Add 'avoid' / negative prompt — things the model should NOT generate">+ <span id="avoidToggleLabel">Avoid +</span>+ </button>+ <span class="ct-spacer"></span>+ <label class="toggle-pill" id="hdrPill"+ title="Boost dynamic range and color depth. Implemented as the official Lightricks HDR LoRA fused into the transformer. First HDR job spawns a one-time download (~120 MB) of the LoRA weights from Hugging Face, then renders share the cache.">+ <input type="checkbox" id="hdr" name="hdr">+ <span class="toggle-dot"></span>+ <span>HDR</span>+ </label>+ <label class="toggle-pill" id="noMusicPill"+ title="When on, the prompt is augmented with: 'Audio: voice and ambient sounds only, no music, no soundtrack, no score.' Useful for clips you'll score yourself in post — music can't be cleanly removed afterwards.">+ <input type="checkbox" id="noMusic" name="no_music">+ <span class="toggle-dot"></span>+ <span>No music</span>+ </label>+ </div>+ </div>++ <!-- Avoid (negative prompt) — collapsed by default; toggled by the+ Avoid + button inside the composer-tools strip. Lives outside+ the composer card so the textarea has its own framed surface+ and the composer card stays focused on the positive prompt. -->+ <div class="avoid-row" id="avoidRow">+ <div class="avoid-label" for="negative_prompt">+ <span>Avoid</span>+ <span class="avoid-hint">things the model should NOT generate</span>+ </div>+ <textarea class="avoid-textarea" name="negative_prompt" id="negative_prompt" placeholder="blurry hands, distorted fingers, extra fingers, smeared face, warped text"></textarea></div>- <!-- Mode-specific: extend -->- <div class="mode-only" id="extendSection">- <h2>Source video</h2>- <select id="extendSrcSelect" onchange="document.getElementById('video_path').value=this.value"></select>- <input name="video_path" id="video_path" placeholder="/path/to/source.mp4" style="margin-top:6px">- <div class="row" style="margin-top:8px">- <div>- <label class="lbl">Extend by (seconds)</label>- <!-- min must align to step from value=2: with min=0.4 step=0.5 the- valid sequence is 0.4, 0.9, 1.4, 1.9, 2.4… — `2` is OFF the- grid. Chrome blocks the entire form submission silently when- ANY input fails validation, even hidden ones in inactive- modes. Generate appeared dead until cocktailpeanut diagnosed- it. With min=0.5 the sequence is 0.5, 1.0, 1.5, 2.0… and- value=2 is valid. -->- <input id="extend_seconds" type="number" value="2" min="0.5" max="10" step="0.5">- <input type="hidden" name="extend_frames" id="extend_frames" value="6">- <div class="hint" id="extendDurationHint" style="margin-top:4px">≈ 2.0 s of new content (6 latent frames × 8 video frames at 24 fps)</div>+ <!-- ============== QUICK SETTINGS ==============+ Quality picker (compact 4-up segmented strip) + Duration ++ Frames + Seed in one dense block. Everything more advanced+ (aspect, dims, speed, long-clips, export, audio source,+ open-when-done) is folded into the Customize disclosure below. -->+ <div class="quick-settings">+ <div>+ <div class="qs-label">+ <span class="qs-name">Quality</span>+ <span class="qs-meta" id="qualityMeta"></span></div>- <div>- <label class="lbl">Direction</label>- <select name="extend_direction" id="extend_direction">- <option value="after" selected>After</option>- <option value="before">Before</option>- </select>+ <!-- Compact 4-col strip. Each chip carries name + a single spec+ line. Same data-quality wiring as before so setQuality()+ keeps working unchanged. The .pill-btn class is preserved+ for the click handler that filters by it. -->+ <div class="quality-strip pill-group" id="qualityGroup">+ <button type="button" class="q-chip pill-btn pill-quality" data-quality="quick">+ <span class="ql-name">Quick</span>+ <span class="q-spec ql-spec sub">640×480</span>+ <span class="ql-tier" hidden>Q4 · any Mac</span>+ </button>+ <button type="button" class="q-chip pill-btn pill-quality active" data-quality="balanced">+ <span class="ql-name">Balanced</span>+ <span class="q-spec ql-spec sub">1024×576</span>+ <span class="ql-tier" hidden>Q4 · ~5 min · no crop</span>+ </button>+ <button type="button" class="q-chip pill-btn pill-quality" data-quality="standard">+ <span class="ql-name">Standard</span>+ <span class="q-spec ql-spec sub">1280×704</span>+ <span class="ql-tier" hidden>Q4 · standard tier+</span>+ </button>+ <button type="button" class="q-chip pill-btn pill-quality disabled" data-quality="high" id="qualityHigh">+ <span class="ql-name">High</span>+ <span class="q-spec ql-spec sub" id="highSpec">1280×704</span>+ <span class="ql-tier" id="highSub" hidden>Q8 not installed</span>+ </button></div></div>- <!-- Speed/quality preset for extend.- Fast: cfg_scale=1.0 (no CFG, ~half the activation memory) +- 12 steps. Fits comfortably on 64 GB at 1280×704. ~3-5 min.- Quality: cfg_scale=3.0 + 30 steps. The upstream defaults. Will- swap on 64 GB at 1280×704 (peak ~47 GB resident + 12 GB- swap) — only pick this on a 96+ GB machine or below 768- max-side. Pinokio 64 GB users should leave this on Fast. -->- <label class="lbl" style="margin-top:10px">Extend mode</label>- <div class="pill-group cols-2" id="extendModeGroup">- <button type="button" class="pill-btn active" data-extend-mode="fast"><span>Fast</span><span class="sub">12 steps · no CFG · 64 GB safe</span></button>- <button type="button" class="pill-btn" data-extend-mode="quality"><span>Quality</span><span class="sub">30 steps · CFG 3.0 · 96+ GB</span></button>+ <div class="mode-only show" id="quickMetricsRow">+ <div class="mini-fields">+ <div class="mf-cell">+ <span class="mf-label">Duration (s)</span>+ <input id="duration" value="5" type="number" min="1" max="20" step="1">+ </div>+ <div class="mf-cell">+ <span class="mf-label">Frames</span>+ <input name="frames" id="frames" value="121" type="number" min="1" title="8k+1">+ </div>+ <div class="mf-cell">+ <span class="mf-label">Seed</span>+ <input name="seed" id="seed" value="-1" title="-1 = random">+ </div>+ </div></div>- <input type="hidden" name="extend_steps" id="extend_steps" value="12">- <input type="hidden" name="extend_cfg" id="extend_cfg" value="1.0">- <div class="hint">Each latent ≈ 8 frames (~0.33s). Quality mode runs the upstream defaults but pushes 1280×704 into swap on 64 GB Macs (~2 hr/render). Stick with Fast unless you've got more RAM.</div>- </div>-- <h2>Prompt</h2>- <!-- Audio-guidance hint in the placeholder. LTX 2.3 generates audio- jointly but is CONDITIONED on prompt cues — visual-only prompts- produce near-silent room tone (peaks at -37 dB on test runs).- Most users assume "no sound" = bug. Hint nudges them to describe- the soundscape too. Documented in the LTX 2.3 paper but unobvious. -->- <textarea name="prompt" id="prompt" placeholder="Describe the scene AND the sound: e.g. 'wizard in a forest clearing, fireflies spiraling up — low whispered chant, ember crackle, distant owl'. Audio is generated jointly with video; without sound cues the model outputs near-silent ambient."></textarea>- <label class="lbl" for="negative_prompt">Avoid</label>- <textarea class="avoid-textarea" name="negative_prompt" id="negative_prompt" placeholder="Optional: blurry hands, distorted fingers, extra fingers, smeared face, warped text"></textarea>- <!-- Gemma-driven prompt enhancement (upstream's `ltx-2-mlx enhance`).- Rewrites your prompt with the structure/keywords LTX 2.3 trained- on. ~12-15s on cold start (Gemma needs to load), ~5s warm.-- "No music" checkbox: appends an audio constraint to the prompt- on submit so users can keep voice + ambient and skip the model's- default soundtrack tendency. Music is annoying in editing because- it can't be cleanly removed without affecting the dialogue track.- Recommended for clips you plan to score yourself in post. -->- <!-- Three controls on one row: Enhance button on the left, then- HDR + No-music toggles pushed to the right. HDR is exposed as- a plain toggle even though it's implemented as a curated- Lightricks LoRA — most users don't care that it's a LoRA, they- just want HDR or not. -->- <div class="row-actions" style="margin-top:8px; display: flex; gap: 8px; align-items: center; flex-wrap: wrap">- <button type="button" class="ghost-btn" id="enhanceBtn" onclick="enhancePrompt()" title="Use Gemma to rewrite your prompt in the style LTX 2.3 was trained on">✨ Enhance with Gemma</button>- <label class="toggle-pill" id="hdrPill" style="margin-left:auto"- title="Boost dynamic range and color depth. Implemented as the official Lightricks HDR LoRA fused into the transformer. First HDR job spawns a one-time download (~120 MB) of the LoRA weights from Hugging Face, then renders share the cache.">- <input type="checkbox" id="hdr" name="hdr">- <span class="toggle-dot"></span>- <span>HDR</span>- </label>- <label class="toggle-pill" id="noMusicPill"- title="When on, the prompt is augmented with: 'Audio: voice and ambient sounds only, no music, no soundtrack, no score.' Useful for clips you'll score yourself in post — music can't be cleanly removed afterwards.">- <input type="checkbox" id="noMusic" name="no_music">- <span class="toggle-dot"></span>- <span>No music</span>- </label></div><!-- LoRA picker — UNIFIED across video form + Image Studio.@@ -13909,34 +14295,21 @@ HTML = r"""<!doctype html>and attaches a `loras` field manually. --><input type="hidden" id="lorasJson" name="loras" value="">- <!-- Advanced — power-user options. We trimmed two things in cleanup:- 1. Removed the "Enhance prompt" checkbox: it was labeled "CLI only,- ignored by helper" — actual dead code. The Enhance button next- to the prompt textarea is the real thing.- 2. The I2V audio mode (mux external audio over LTX-generated video)- only applies in I2V mode; the panel auto-hides it elsewhere via- `.mode-only` on the wrapper. -->- <details>- <summary>Advanced</summary>- <div class="mode-only" id="i2vAudioModeSection">- <label class="lbl">I2V audio source</label>- <select id="i2vMode">- <option value="i2v" selected>Joint audio (LTX generates audio synced with the visual)</option>- <option value="i2v_clean_audio">Use external audio file (mux it onto LTX video)</option>- </select>- <div class="mode-only" id="audioSection">- <label class="lbl">Audio file path</label>- <input name="audio" id="audio" placeholder="/path/to/your/track.wav">- </div>- </div>- <label class="check" style="margin-top:6px">- <input type="checkbox" name="open_when_done" id="open_when_done"> Open file when done- </label>- </details>-- <!-- Sizing for non-extend modes. The headline Quality picker above- sets sensible defaults; this disclosure lets power users override- aspect, exact dimensions, and the experimental sampler speed. -->+ <input type="hidden" name="steps" id="steps" value="8">++ <!-- ============== UNIFIED CUSTOMIZE DISCLOSURE ==============+ Folds together the old "Customize" (aspect, dims, speed,+ long-clips, export, method) and "Advanced" (audio source,+ open-when-done) into ONE collapsed disclosure. Power users+ open it; beginners ignore it. The summary line shows the+ effective state so power users can scan at a glance.++ Lives at form level (not mode-conditional) — its inner rows+ manage their own visibility (e.g. dimsRow hides in I2V flows;+ i2vAudioModeSection only shows for I2V; method only shows+ when an upscale is active). The wrapping #sizingSection+ class kept for the mode-only show/hide that still toggles+ it for Extend (Extend supplies its own duration/etc). --><div class="mode-only" id="sizingSection"><details id="customizeDetails" class="customize-section"><summary class="cz-summary">@@ -14027,16 +14400,33 @@ HTML = r"""<!doctype html></div><input type="hidden" name="upscale_method" id="upscale_method" value="lanczos"></div>- </div>- </details>- <div class="row3" style="margin-top:10px">- <div><label class="lbl">Duration (s)</label><input id="duration" value="5" type="number" min="1" max="20" step="1"></div>- <div><label class="lbl">Frames (8k+1)</label><input name="frames" id="frames" value="121" type="number" min="1"></div>- <div><label class="lbl">Seed (-1 random)</label><input name="seed" id="seed" value="-1"></div>- </div>+ <!-- Audio source (I2V only) — folded in from the old separate+ Advanced disclosure. Picker auto-hides when the current+ mode isn't I2V (see updateDerived's mode-aware visibility). -->+ <div class="cz-control mode-only" id="i2vAudioModeSection">+ <div class="cz-label">I2V audio source+ <span class="cz-label-hint">joint or external</span>+ </div>+ <select id="i2vMode">+ <option value="i2v" selected>Joint audio (LTX generates audio synced with the visual)</option>+ <option value="i2v_clean_audio">Use external audio file (mux it onto LTX video)</option>+ </select>+ <div class="mode-only" id="audioSection" style="margin-top:8px">+ <label class="lbl">Audio file path</label>+ <input name="audio" id="audio" placeholder="/path/to/your/track.wav">+ </div>+ </div>- <input type="hidden" name="steps" id="steps" value="8">+ <!-- Open file when done — minor convenience. Was its own row+ in the old Advanced disclosure; now folded inline. -->+ <div class="cz-control">+ <label class="check" style="margin:0">+ <input type="checkbox" name="open_when_done" id="open_when_done"> Open file when done+ </label>+ </div>+ </div>+ </details><div class="derived" id="derived"></div></div>@@ -14053,14 +14443,23 @@ HTML = r"""<!doctype html></label></div>- <div class="actions">- <button type="submit" class="primary" id="genBtn">Generate</button>- <button type="button" class="danger" onclick="api('/stop', 'POST').then(poll)">Stop</button>- </div>- <div class="row-actions" style="margin-top:8px">- <button type="button" class="small" onclick="openBatch()">Batch paste</button>- <button type="button" class="small" id="pauseBtn" onclick="togglePause()">Pause queue</button>- <button type="button" class="small" onclick="api('/queue/clear','POST').then(poll)">Clear queue</button>+ <!-- ============== STICKY ACTION FOOTER ==============+ Generate stays anchored at the bottom of the form-pane no+ matter how far the user scrolls. The queue chip strip sits+ above it (Batch / Pause / Clear) — compact pills, never the+ primary touch target. -->+ <div class="form-action-footer" id="formActionFooter">+ <div class="queue-strip">+ <button type="button" class="qchip" onclick="openBatch()" title="Paste many prompts at once and queue them">⊞ Batch</button>+ <button type="button" class="qchip" id="pauseBtn" onclick="togglePause()" title="Pause / resume the render queue">⏸ Pause queue</button>+ <button type="button" class="qchip" onclick="api('/queue/clear','POST').then(poll)" title="Clear all queued jobs (running job continues)">✕ Clear</button>+ <span class="qchip-spacer"></span>+ <span class="qchip-derived" id="derivedFooter"></span>+ </div>+ <div class="actions">+ <button type="submit" class="primary" id="genBtn">Generate</button>+ <button type="button" class="danger" onclick="api('/stop', 'POST').then(poll)">Stop</button>+ </div></div></form>@@ -14079,39 +14478,59 @@ HTML = r"""<!doctype html>download — driven by /image/engine_status (server-side cacheprobe via _repo_hf_cache_dir). --><div class="mode-only studio-pane" id="studioSection">- <!-- The Mode picker above this <div> already announces "Studio" so- a separate "Image Studio" h3 just adds noise. The hint moves- inline as a one-liner below the prompt placeholder if needed. -->-- <h2 style="margin-top:0">Prompt</h2>- <textarea id="imgStudioPrompt" rows="4" placeholder="A cinematic medium close-up of a woman in a sunlit kitchen, soft morning light through blinds, shallow depth of field, photorealistic"></textarea>-- <h2>Reference images- <span class="h2-hint">multi-ref · Qwen-Image-Edit-2511 / FLUX.2 Klein-Edit only</span>- </h2>- <div class="studio-ref-grid" id="imgStudioRefs">- <div class="studio-ref-slot" data-slot="0">- <span class="ref-tag">Primary</span>- </div>- <div class="studio-ref-slot" data-slot="1">- <span class="ref-tag">Multi-ref</span>- </div>- <div class="studio-ref-slot" data-slot="2">- <span class="ref-tag">Multi-ref</span>+ <!-- ============== STUDIO COMPOSER CARD ==============+ Same composer-card chrome as the video form. Refs (3 multi-ref+ slots) sit at the top, the prompt is the hero, and the engine /+ aspect / n / seed live in the Quick Settings strip below.+ Generate sticks at the bottom (sticky footer) so it's always+ one click away regardless of how far the user scrolls. -->+ <div class="composer-card">+ <div class="composer-refs">+ <div class="show">+ <h2>Reference images+ <span class="h2-hint">multi-ref · Qwen-Image-Edit-2511 / FLUX.2 Klein-Edit only</span>+ </h2>+ <div class="studio-ref-grid" id="imgStudioRefs">+ <div class="studio-ref-slot" data-slot="0">+ <span class="ref-tag">Primary</span>+ </div>+ <div class="studio-ref-slot" data-slot="1">+ <span class="ref-tag">Multi-ref</span>+ </div>+ <div class="studio-ref-slot" data-slot="2">+ <span class="ref-tag">Multi-ref</span>+ </div>+ </div>+ <!-- Mirrors the video I2V picker's "Recent uploads · click to use"+ strip. Click a thumb → fills the next empty ref slot. Hidden+ when /uploads is empty. -->+ <div class="studio-ref-recent" id="imgStudioRecentWrap" style="display:none">+ <div class="studio-ref-recent-label">Recent uploads · click to use</div>+ <div class="studio-ref-recent-strip" id="imgStudioRecentStrip"></div>+ </div>+ </div></div>++ <textarea id="imgStudioPrompt" class="composer-prompt" rows="4"+ placeholder="A cinematic medium close-up of a woman in a sunlit kitchen, soft morning light through blinds, shallow depth of field, photorealistic"></textarea></div>- <!-- Mirrors the video I2V picker's "Recent uploads · click to use"- strip. Click a thumb → fills the next empty ref slot. Hidden- when /uploads is empty. -->- <div class="studio-ref-recent" id="imgStudioRecentWrap" style="display:none">- <div class="studio-ref-recent-label">Recent uploads · click to use</div>- <div class="studio-ref-recent-strip" id="imgStudioRecentStrip"></div>++ <!-- Validation / status messages (e.g. "pick a ref image"). Lives+ OUTSIDE the composer card so the card chrome stays focused on+ composing. Empty by default. -->+ <div class="studio-status-row" style="margin: 6px 0 -4px;">+ <span id="imgStudioStatus" class="hint"></span></div>- <h2>Settings</h2>- <div class="studio-field-grid">- <div class="studio-field studio-field-wide">- <label class="lbl">Engine</label>+ <!-- Quick settings — Engine on its own row (long labels), then a+ dense 3-up row for Aspect / Candidates / Seed. Wall-time+ estimate sits below the strip in the muted derived style. -->+ <div class="quick-settings">+ <div>+ <div class="qs-label">+ <span class="qs-name">Engine</span>+ <span class="qs-meta" id="imgStudioWallEstimate"></span>+ </div><div class="studio-engine-row"><select id="imgStudioEngine" onchange="imgStudioUpdateValidity();imgStudioRefreshEngineStatus();imgStudioUpdateEstimate();if(typeof renderLorasList==='function')renderLorasList()"><option value="auto" selected>Auto (use Settings)</option>@@ -14129,24 +14548,29 @@ HTML = r"""<!doctype html>onclick="imgStudioOnPillClick()">…</span></div></div>- <div class="studio-field">- <label class="lbl">Aspect</label>- <select id="imgStudioAspect" onchange="imgStudioUpdateEstimate()">- <option value="16:9" selected>16:9 — 1280×720</option>- <option value="4:3">4:3 — 1024×768</option>- <option value="1:1">1:1 — 1024×1024</option>- <option value="9:16">9:16 — 720×1280</option>- <option value="3:4">3:4 — 768×1024</option>- <option value="21:9">21:9 — 1280×544</option>- </select>- </div>- <div class="studio-field">- <label class="lbl">Candidates (n)</label>- <input type="number" id="imgStudioN" min="1" max="8" value="4" oninput="imgStudioUpdateEstimate()">- </div>- <div class="studio-field">- <label class="lbl">Seed (-1 random)</label>- <input type="number" id="imgStudioSeed" value="-1">++ <div>+ <div class="mini-fields">+ <div class="mf-cell">+ <span class="mf-label">Aspect</span>+ <select id="imgStudioAspect" onchange="imgStudioUpdateEstimate()" style="padding:7px 9px;font-size:13px;">+ <option value="16:9" selected>16:9 — 1280×720</option>+ <option value="4:3">4:3 — 1024×768</option>+ <option value="1:1">1:1 — 1024×1024</option>+ <option value="9:16">9:16 — 720×1280</option>+ <option value="3:4">3:4 — 768×1024</option>+ <option value="21:9">21:9 — 1280×544</option>+ </select>+ </div>+ <div class="mf-cell">+ <span class="mf-label">Candidates (n)</span>+ <input type="number" id="imgStudioN" min="1" max="8" value="4" oninput="imgStudioUpdateEstimate()">+ </div>+ <div class="mf-cell">+ <span class="mf-label">Seed</span>+ <input type="number" id="imgStudioSeed" value="-1" title="-1 = random">+ </div>+ </div></div></div>@@ -14158,39 +14582,23 @@ HTML = r"""<!doctype html>chips that don't match the current engine get a warning badge. --><div id="loraPickerStudioSlot"></div>- <!-- Wall-time estimate sits on its own muted line above the action- row, mirroring the video form's "Duration 5.00s @ 24fps · …"- derived row. Updated by imgStudioUpdateEstimate as the user- changes engine / aspect / n. -->- <div class="derived" id="imgStudioWallEstimateRow">- <span id="imgStudioWallEstimate"></span>- </div>-- <!-- Image Studio's action row uses the EXACT same .actions grid- (Generate=2fr, Stop=1fr) the video form uses, so the chrome- reads as the same panel in two modes instead of two different- panels. Same /stop endpoint works for image jobs (they go- through the shared queue worker), so the Stop button is- literally identical to the video form's. -->- <div class="actions">- <button class="primary" id="imgStudioGenBtn" onclick="imgStudioGenerate()">- <span id="imgStudioGenBtnLabel">Generate</span>- </button>- <button type="button" class="danger" onclick="api('/stop', 'POST').then(poll)">Stop</button>- </div>- <!-- Same queue-action strip the video form has. Pause/Clear are- queue-wide so they work regardless of mode; Batch paste opens- the same modal. Mirrors the video form's row exactly so the- Studio composer doesn't lose chrome users built muscle memory- around. -->- <div class="row-actions" style="margin-top:8px">- <button type="button" class="small" onclick="openBatch()">Batch paste</button>- <button type="button" class="small" id="pauseBtnStudio" onclick="togglePause()">Pause queue</button>- <button type="button" class="small" onclick="api('/queue/clear','POST').then(poll)">Clear queue</button>- </div>-- <div class="studio-status-row">- <span id="imgStudioStatus" class="hint"></span>+ <!-- Sticky action footer — same chrome as the video form's. Generate+ stays anchored at the bottom of the form-pane no matter how far+ the user has scrolled. Pause/Clear/Batch live above it as compact+ pills, never the primary touch target. -->+ <div class="form-action-footer" id="formActionFooterStudio">+ <div class="queue-strip">+ <button type="button" class="qchip" onclick="openBatch()" title="Paste many prompts at once and queue them">⊞ Batch</button>+ <button type="button" class="qchip" id="pauseBtnStudio" onclick="togglePause()" title="Pause / resume the render queue">⏸ Pause queue</button>+ <button type="button" class="qchip" onclick="api('/queue/clear','POST').then(poll)" title="Clear all queued jobs (running job continues)">✕ Clear</button>+ <span class="qchip-spacer"></span>+ </div>+ <div class="actions">+ <button class="primary" id="imgStudioGenBtn" onclick="imgStudioGenerate()">+ <span id="imgStudioGenBtnLabel">Generate</span>+ </button>+ <button type="button" class="danger" onclick="api('/stop', 'POST').then(poll)">Stop</button>+ </div></div><!-- Submission status. Image jobs flow through the same queue as@@ -15112,6 +15520,37 @@ function _autoMainOutputsFilterForMode(mode) {document.getElementById('audio').value = BOOT.default_audio;// ====== Pill-button group helpers ======+// Avoid (negative prompt) toggle — keeps the textarea collapsed by+// default so the prompt-and-go flow stays tight. The button label flips+// to "Avoid −" when open so the affordance is clear. Auto-opens once+// the user types into it (so it stays visible) — shouldn't auto-close+// from typing-then-deleting because that would yank the field out from+// under their next keystroke.+function toggleAvoidRow(forceOpen) {+ const row = document.getElementById('avoidRow');+ const btn = document.getElementById('avoidToggleBtn');+ const lbl = document.getElementById('avoidToggleLabel');+ const ta = document.getElementById('negative_prompt');+ if (!row) return;+ const wantOpen = (forceOpen === true) ? true : !row.classList.contains('show');+ row.classList.toggle('show', wantOpen);+ if (btn) btn.classList.toggle('active', wantOpen);+ if (lbl) lbl.textContent = wantOpen ? 'Avoid −' : 'Avoid +';+ if (wantOpen && ta) {+ try { ta.focus(); } catch (e) {}+ }+}+// Auto-open the Avoid row if it has content (e.g. loaded from a sidecar+// via loadParams). Run once at boot AND after loadParams sets values.+function syncAvoidRowFromValue() {+ const ta = document.getElementById('negative_prompt');+ if (!ta) return;+ if ((ta.value || '').trim() !== '') {+ const row = document.getElementById('avoidRow');+ if (row && !row.classList.contains('show')) toggleAvoidRow(true);+ }+}+function setMode(mode) {currentMode = mode;// The Studio pill swaps the form-pane in place: hide the video form@@ -15994,6 +16433,23 @@ function updateDerived() {document.getElementById('derived').innerHTML = `Duration <strong>${dur}s</strong> @ ${FPS}fps${temporalText} · ${finalRes} · Steps ${document.getElementById('steps').value}${accelText}`;+ // Compact derived line in the sticky action footer — same info, tighter+ // typography. Lets the user see what they're about to render WITHOUT+ // scrolling back to the customize disclosure. Mirrors the Customize+ // summary style (just dimensions + duration; full details stay in the+ // expanded Customize body).+ const derivedFooter = document.getElementById('derivedFooter');+ if (derivedFooter) {+ derivedFooter.innerHTML = `<strong>${dur}s</strong> · ${finalRes}${temporalText}${accelText}`;+ }+ // Also update the Quality strip's right-side meta line (e.g. "5s · 1024×576")+ // so the Quality picker block reads as a self-contained summary.+ const qualityMeta = document.getElementById('qualityMeta');+ if (qualityMeta) {+ const qBare = `${w}×${h}`;+ qualityMeta.textContent = `${dur}s · ${qBare}`;+ }+const warns = [];if (w % 32 !== 0) warns.push(`Width ${w} isn't a multiple of 32 (closest ${Math.round(w/32)*32})`);if (h % 32 !== 0) warns.push(`Height ${h} isn't a multiple of 32 (closest ${Math.round(h/32)*32})`);@@ -16015,6 +16471,11 @@ function updateDerived() {document.getElementById('extendSection').classList.toggle('show', currentMode === 'extend');document.getElementById('keyframeSection').classList.toggle('show', currentMode === 'keyframe');document.getElementById('sizingSection').classList.toggle('show', currentMode !== 'extend');+ // quickMetricsRow (Duration / Frames / Seed) doesn't apply to Extend+ // (extend_seconds drives the new content; the source video provides+ // the rest). Hide it in extend, show otherwise.+ const qmr = document.getElementById('quickMetricsRow');+ if (qmr) qmr.classList.toggle('show', currentMode !== 'extend');document.getElementById('audioSection').classList.toggle('show', mode === 'i2v_clean_audio');// I2V audio source picker (Advanced) — only relevant in I2V flow.// In T2V/Extend/FFLF the model generates audio jointly; there's nothing@@ -16932,6 +17393,9 @@ async function loadParams() {if (p.upscale_method) setUpscaleMethod(p.upscale_method);document.getElementById('prompt').value = p.prompt || '';document.getElementById('negative_prompt').value = p.negative_prompt || '';+ // If the loaded sidecar carried an Avoid value, surface the row so the+ // user can see it without having to click the toggle.+ syncAvoidRowFromValue();if (p.frames) { document.getElementById('frames').value = p.frames; document.getElementById('duration').value = framesToDuration(p.frames); }if (p.steps) document.getElementById('steps').value = p.steps;if (p.seed != null) document.getElementById('seed').value = p.seed;