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(agent): "Queue them" batch bar above composer
mrbizarro969f20d3560a
diff --git a/mlx_ltx_panel.py b/mlx_ltx_panel.pyindex cd8658f..6ebc06f 100644--- a/mlx_ltx_panel.py+++ b/mlx_ltx_panel.py@@ -8526,6 +8526,54 @@ HTML = r"""<!doctype html>max-width: fit-content;animation: agent-fade-in 0.2s ease;}++ /* Batch-now bar: appears above the composer when there are picked+ anchors not yet submitted. One-click trigger to queue them all. */+ .agent-batch-bar {+ display: flex;+ align-items: center;+ justify-content: space-between;+ gap: 12px;+ padding: 9px 14px 9px 16px;+ margin: 0 0 10px 0;+ background: linear-gradient(135deg, rgba(63,185,80,0.10) 0%, rgba(47,129,247,0.08) 100%);+ border: 1px solid rgba(63,185,80,0.35);+ border-radius: 12px;+ animation: agent-fade-in 0.22s ease;+ box-shadow: 0 4px 16px rgba(63,185,80,0.06);+ }+ .agent-batch-bar[hidden] { display: none; }+ .agent-batch-bar-text {+ display: inline-flex;+ align-items: center;+ gap: 10px;+ font-size: 12.5px;+ color: var(--text);+ }+ .agent-batch-bar-icon {+ color: var(--success, #3fb950);+ font-size: 14px;+ line-height: 1;+ }+ .agent-batch-bar-btn {+ display: inline-flex;+ align-items: center;+ gap: 6px;+ padding: 6px 12px;+ background: var(--success, #3fb950);+ color: #04130a;+ border: none;+ border-radius: 8px;+ font-size: 12px;+ font-weight: 600;+ cursor: pointer;+ transition: filter 0.15s, transform 0.15s;+ width: auto;+ flex: 0 0 auto;+ }+ .agent-batch-bar-btn:hover { filter: brightness(1.08); transform: translateY(-1px); }+ .agent-batch-bar-btn:active { transform: translateY(0); }+ .agent-batch-bar-btn svg { flex-shrink: 0; }.agent-ref-chip.visible { display: inline-flex; }.agent-ref-chip .ref-icon {font-size: 11px;@@ -9508,6 +9556,25 @@ HTML = r"""<!doctype html><div class="agent-chat" id="agentChat"></div><div class="agent-composer">+ <!-- Batch-now bar: surfaces when the user has picked anchors that+ haven't yet been queued. One click injects a structured message+ telling the agent to submit each pick as i2v with the durations+ it already proposed. The agent's existing prompt covers the+ phrasing; we just trigger it. -->+ <div class="agent-batch-bar" id="agentBatchBar" hidden>+ <div class="agent-batch-bar-text">+ <span class="agent-batch-bar-icon">✦</span>+ <span id="agentBatchBarSummary">…</span>+ </div>+ <button type="button" class="agent-batch-bar-btn"+ onclick="agentBatchQueue()">+ <span>Queue them</span>+ <svg viewBox="0 0 24 24" width="13" height="13" fill="none" stroke="currentColor" stroke-width="2.2" stroke-linecap="round" stroke-linejoin="round">+ <line x1="5" y1="12" x2="19" y2="12"/>+ <polyline points="12 5 19 12 12 19"/>+ </svg>+ </button>+ </div><!-- Reference chip: appears when the user clicks Refine on anexisting clip. Carries the clip's job_id; on Send, weprepend "Refine <job_id>: " to the user's message so the@@ -13875,8 +13942,10 @@ async function agentLoadSession(sid) {window.AGENT.sessionId = sid;const sess = j.session || {};window.AGENT.selectedAnchors = (sess.tool_state || {}).selected_anchors || {};+ window.AGENT.submittedShots = (sess.tool_state || {}).submitted_shots || [];agentSetSessionTitle(sess.title || 'Untitled', sid);agentRender(j.rendered_messages || []);+ agentRenderBatchBar();} catch (e) { console.error('agentLoadSession', e); }}@@ -14680,6 +14749,57 @@ async function agentPickAnchor(label, cand, gridEl) {} catch (e) {console.error('agentPickAnchor', e);}+ agentRenderBatchBar();+}++// Sticky bar above the composer that surfaces "you've picked anchors;+// click here to queue them all". Drives the user's promised "Ok now batch+// everything" workflow. Counts picked shots vs already-submitted shots+// (matched by `label`); shows when picks > submitted.+function agentRenderBatchBar() {+ const bar = document.getElementById('agentBatchBar');+ const text = document.getElementById('agentBatchBarSummary');+ if (!bar || !text) return;+ const picks = window.AGENT.selectedAnchors || {};+ const pickLabels = Object.keys(picks).filter(k => (picks[k] || {}).png_path);+ // submitted_shots lives on the loaded session; we cache it on AGENT.+ const submitted = (window.AGENT.submittedShots || []);+ const submittedLabels = new Set(submitted.map(s => (s.label || '').trim()).filter(Boolean));+ const pendingLabels = pickLabels.filter(l => !submittedLabels.has(l));+ if (pendingLabels.length === 0 || window.AGENT.busy) {+ bar.hidden = true;+ return;+ }+ const total = pickLabels.length;+ text.textContent = pendingLabels.length === total+ ? `${total} anchor${total === 1 ? '' : 's'} picked · ready to render`+ : `${pendingLabels.length} of ${total} picks not yet queued`;+ bar.hidden = false;+}++// User clicked "Queue them" — inject a structured message that prompts+// the agent to submit each picked anchor as i2v, using durations from the+// plan it already produced. Mirrors how the user would say it themselves+// ("ok let's batch everything") — we just save them the typing.+function agentBatchQueue() {+ const input = document.getElementById('agentInput');+ if (!input) return;+ const picks = window.AGENT.selectedAnchors || {};+ const pickLabels = Object.keys(picks).filter(k => (picks[k] || {}).png_path);+ const submitted = (window.AGENT.submittedShots || []);+ const submittedLabels = new Set(submitted.map(s => (s.label || '').trim()).filter(Boolean));+ const pending = pickLabels.filter(l => !submittedLabels.has(l));+ if (pending.length === 0) return;+ const list = pending.map(l => `- ${l}`).join('\n');+ input.value =+ `Queue every picked anchor as an i2v shot now using the durations and prompts from your plan. ` ++ `Use balanced quality + Sharp 720p unless I overrode that earlier. ` ++ `Pending picks:\n${list}\n` ++ `Submit them in plan order, then call finish. Don't ask for confirmation — go.`;+ agentAutoResize(input);+ agentUpdateSendState();+ // Auto-fire — the user clicked the button, so they consented.+ agentSend();}function summarizeToolCall(call) {@@ -15791,6 +15911,13 @@ async function agentStageTick() {]);agentStageRender(statusResp, sessResp);agentRenderRamChip(statusResp);+ // Keep submittedShots cache + batch bar in sync with the live session.+ if (sessResp && sessResp.session && sessResp.session.tool_state) {+ const ts = sessResp.session.tool_state;+ window.AGENT.submittedShots = ts.submitted_shots || [];+ window.AGENT.selectedAnchors = ts.selected_anchors || window.AGENT.selectedAnchors || {};+ agentRenderBatchBar();+ }} catch(e) {/* swallow — next tick retries */}