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
chore(panel/image-studio): drop redundant in-pane gallery; Generate now goes through queue
mrbizarro08cf1c0d4e1b
diff --git a/mlx_ltx_panel.py b/mlx_ltx_panel.pyindex de1ef67..c5b5e07 100644--- a/mlx_ltx_panel.py+++ b/mlx_ltx_panel.py@@ -12942,22 +12942,13 @@ HTML = r"""<!doctype html><div style="margin-top:8px;display:flex;align-items:center;gap:8px;flex-wrap:wrap"><button class="small primary" id="imgStudioGenBtn" onclick="imgStudioGenerate()">Generate</button><span id="imgStudioStatus" class="hint" style="flex:1"></span>- <button class="small" onclick="imgStudioRefreshLibrary()">Refresh library</button></div>- <!-- Generation results inline -->+ <!-- Submission status. Image jobs flow through the same queue as+ video jobs now, so the actual results show up in the right-rail+ Recent tab (with the Photos filter) along with everything else.+ This pane is just a thin "submitted" confirmation. --><div id="imgStudioResults" style="margin-top:14px"></div>-- <!-- Library inline (Phase 1). Phase 2 will move this into the right- rail so it sits next to the queue / current / history cards. -->- <div style="margin-top:18px;border-top:1px solid var(--border);padding-top:14px">- <h4 style="margin:0 0 6px 0">Library</h4>- <div class="hint" style="margin-bottom:8px">Recent stills from this panel. Click any image to copy its path for use as a reference.</div>- <input type="search" id="imgStudioLibrarySearch" placeholder="Filter by prompt…" oninput="imgStudioRefreshLibrary()" style="width:100%;margin-bottom:10px">- <div id="imgStudioLibrary" style="display:grid;grid-template-columns:repeat(3,1fr);gap:8px;max-height:50vh;overflow-y:auto;padding-right:4px">- <div class="hint">Loading…</div>- </div>- </div></div></aside>@@ -13908,80 +13899,54 @@ async function imgStudioGenerate() {engine_override: document.getElementById('imgStudioEngine').value,refs,};+ // Submit through the same /queue/add endpoint video jobs use, with+ // mode='image'. The worker thread routes mode==='image' to+ // run_image_job_inner (commit 1). Progress streams into the panel+ // log + the right-rail Now/Queue/Recent cards exactly the way video+ // jobs do — no need for the Studio's own elapsed-counter or in-pane+ // result grid. The Recent tab's Photos filter (commit 2) surfaces+ // the result with thumbnail + Animate.IMG_STUDIO.busy = true;document.getElementById('imgStudioGenBtn').disabled = true;- document.getElementById('imgStudioResults').innerHTML = '';- // Live elapsed-time counter so the user doesn't sit on a frozen- // "Generating…" string while mflux works. Stops on completion or- // failure (the finally block clears the interval).- const tStart = Date.now();- const fmtElapsed = (ms) => {- const s = Math.floor(ms / 1000);- const mm = String(Math.floor(s / 60)).padStart(2, '0');- const ss = String(s % 60).padStart(2, '0');- return mm + ':' + ss;- };- // Engine-specific ETA hint so the user knows what to expect- const eta = (() => {- const eng = body.engine_override || 'auto';- if (eng === 'qwen_edit_lightning_inline') return '~10-15 s/image, batched';- if (eng === 'qwen_edit_inline') return '~1 min/image at 8 steps Q4';- if (eng === 'qwen_edit_high_inline') return '~5 min/image at 30 steps Q8';- if (eng === 'flux2_inline') return '~15 s/image at 4 steps';- if (eng === 'z_image_turbo_inline') return '~30 s/image at 9 steps';- if (eng === 'mock_inline') return 'instant (testing only)';- return ''; // 'auto' — engine TBD on server- })();- document.getElementById('imgStudioStatus').textContent =- 'Generating… 00:00' + (eta ? ' · ' + eta : '');- IMG_STUDIO.elapsedTimer = setInterval(() => {- const s = document.getElementById('imgStudioStatus');- if (s && IMG_STUDIO.busy) {- s.textContent = 'Generating… ' + fmtElapsed(Date.now() - tStart) + (eta ? ' · ' + eta : '');- }- }, 1000);+ document.getElementById('imgStudioResults').innerHTML =+ '<div class="hint">Submitted to queue. Watch the Now / Recent tabs on the right ' ++ '(Recent → Photos for just images).</div>';+ document.getElementById('imgStudioStatus').textContent = 'Queueing…';try {- const r = await fetch('/image/generate', {- method: 'POST',- headers: { 'Content-Type': 'application/json' },- body: JSON.stringify(body),- });- const j = await r.json();- if (!r.ok || j.error) throw new Error(j.error || `HTTP ${r.status}`);- const cands = j.candidates || [];- // refs_ignored: surface a warning when the user passed refs but the- // selected engine silently dropped them (anything other than qwen_edit).- // Without this, the user sees plausible-looking T2I output and assumes- // their character ref worked.- const refsIgnored = (refs.length > 0) && cands.some(c => c.refs_ignored === true);- let warn = '';- if (refsIgnored) {- warn = ' · ⚠ refs IGNORED — switch to Qwen-Image-Edit-2509 for character composition';- }+ const fd = new URLSearchParams();+ fd.set('mode', 'image');+ fd.set('prompt', body.prompt);+ fd.set('engine_override', body.engine_override || 'auto');+ fd.set('aspect', body.aspect || '16:9');+ fd.set('n', String(body.n || 4));+ fd.set('seed', String(body.seed != null ? body.seed : -1));+ // refs is a list of paths — the existing /image/generate code+ // accepts a JSON-encoded list as a form field, and make_job()+ // mode='image' parses the same shape (commit 1).+ fd.set('refs', JSON.stringify(body.refs || []));+ const r = await fetch('/queue/add', { method: 'POST', body: fd });+ if (!r.ok) throw new Error(`HTTP ${r.status}`);document.getElementById('imgStudioStatus').textContent =- `${cands.length} image${cands.length === 1 ? '' : 's'} in ${j.elapsed_seconds || '?'} s · engine: ${j.engine || '?'}` + warn;- const html = cands.map((c, i) => `- <div class="lib-tile" onclick="imgStudioCopyPath('${(c.png_path || '').replace(/'/g, "\\'")}')" title="Click to copy path">- <img src="/image?path=${encodeURIComponent(c.png_path)}&t=${Date.now()}" alt="">- <div class="lib-meta">cand ${i + 1} · seed ${c.seed}</div>- </div>- `).join('');- document.getElementById('imgStudioResults').innerHTML =- `<div style="display:grid;grid-template-columns:repeat(${Math.min(4, cands.length || 1)},1fr);gap:8px">${html}</div>`;- imgStudioRefreshLibrary();+ 'Submitted. Watch Now / Recent → Photos.';} catch (e) {- document.getElementById('imgStudioStatus').textContent = 'Failed: ' + e.message;+ document.getElementById('imgStudioStatus').textContent = 'Submit failed: ' + e.message;} finally {IMG_STUDIO.busy = false;document.getElementById('imgStudioGenBtn').disabled = false;- if (IMG_STUDIO.elapsedTimer) {- clearInterval(IMG_STUDIO.elapsedTimer);- IMG_STUDIO.elapsedTimer = null;- }}}+// imgStudioRefreshLibrary used to populate the right-pane Studio+// gallery. The unified Recent tab (commit 2) now covers that need —+// the function is kept as a no-op so any setMode('image') call sites+// that still reference it (defensive setMode plumbing) don't error.async function imgStudioRefreshLibrary() {+ // intentionally empty: replaced by the unified Recent tab.+ return;+}++// Disabled stub so the old Library renderer path stays inert.+async function _imgStudioRefreshLibraryLegacy() {const grid = document.getElementById('imgStudioLibrary');if (!grid) return;const q = (document.getElementById('imgStudioLibrarySearch').value || '').trim();