https://portfolio.xiaruize.org/publication_types/article-journal/Article-JournalHugo Blox Builder (https://hugoblox.com)en-usThu, 01 Jan 2026 00:00:00 +0000https://portfolio.xiaruize.org/media/icon_hu_982c5d63a71b2961.pnghttps://portfolio.xiaruize.org/publication_types/article-journal/https://portfolio.xiaruize.org/publications/text2sign/Thu, 01 Jan 2026 00:00:00 +0000https://portfolio.xiaruize.org/publications/text2sign/<p><strong>Authors:</strong> Ruize Xia<br> <strong>Published in:</strong> IEEE Access<br> <strong>DOI:</strong> <br> <strong>Source:</strong> ORCID record </p> <h2 id="abstract">Abstract</h2> <p>Sign language is a primary communication channel for millions of Deaf and hard-of-hearing people, yet generating signer video directly from text remains difficult because video diffusion models are expensive to train and evaluate. This article presents <strong>Text2Sign</strong>, a single-GPU diffusion baseline for text-to-sign language video generation. The model combines a frozen vision-language text encoder with a three-dimensional encoder-decoder backbone and factorized spatial and temporal attention, reducing the cost of full spatio-temporal attention while preserving motion coherence. The results indicate that pretrained text conditioning improves generalization under limited data, although the system remains limited to low-resolution short clips and does not yet include expert linguistic evaluation. The contribution should therefore be read as an efficiency-oriented research baseline rather than a complete sign-language production system.</p> <h2 id="read-the-paper">Read the paper</h2> <div id="pdf-viewer-2930aebb388a0f52036426dfd07fb207" class="pdf-inline-viewer" data-pdf-url="https://portfolio.xiaruize.org/publications/text2sign/paper.pdf"> <div class="pdf-inline-viewer__toolbar"> <span class="pdf-inline-viewer__status">Loading preview…</span> <a class="pdf-inline-viewer__fallback" href="https://portfolio.xiaruize.org/publications/text2sign/paper.pdf" target="_blank" rel="noopener noreferrer"> Open the compiled PDF in a new tab </a> </div> <div class="pdf-inline-viewer__pages"></div> </div> <script> (() => { const root = document.currentScript && document.currentScript.previousElementSibling; if (!root || root.dataset.initialized) { return; } root.dataset.initialized = 'true'; const statusEl = root.querySelector('.pdf-inline-viewer__status'); const pagesEl = root.querySelector('.pdf-inline-viewer__pages'); const pdfUrl = root.dataset.pdfUrl; const fail = (error) => { console.error('PDF viewer failed to load', error); statusEl.textContent = 'Preview unavailable in-browser. Use the PDF button above or the link here.'; root.classList.add('pdf-inline-viewer--failed'); }; (async () => { try { const pdfjsLib = await import('https://cdn.jsdelivr.net/npm/pdfjs-dist@4.10.38/build/pdf.min.mjs'); pdfjsLib.GlobalWorkerOptions.workerSrc = 'https://cdn.jsdelivr.net/npm/pdfjs-dist@4.10.38/build/pdf.worker.min.mjs'; const loadingTask = pdfjsLib.getDocument({ url: pdfUrl, useWorkerFetch: true, isEvalSupported: false, }); const pdf = await loadingTask.promise; statusEl.textContent = `${pdf.numPages} page${pdf.numPages === 1 ? '' : 's'} loaded`; for (let pageNumber = 1; pageNumber <= pdf.numPages; pageNumber += 1) { const page = await pdf.getPage(pageNumber); const baseViewport = page.getViewport({ scale: 1 }); const availableWidth = Math.min(pagesEl.clientWidth || root.clientWidth || 920, 920); const scale = availableWidth / baseViewport.width; const viewport = page.getViewport({ scale }); const pixelRatio = Math.min(window.devicePixelRatio || 1, 2); const pageShell = document.createElement('section'); pageShell.className = 'pdf-inline-viewer__page'; const pageLabel = document.createElement('div'); pageLabel.className = 'pdf-inline-viewer__page-label'; pageLabel.textContent = `Page ${pageNumber}`; const canvas = document.createElement('canvas'); canvas.width = Math.floor(viewport.width * pixelRatio); canvas.height = Math.floor(viewport.height * pixelRatio); canvas.style.width = `${viewport.width}px`; canvas.style.height = `${viewport.height}px`; canvas.setAttribute('role', 'img'); canvas.setAttribute('aria-label', `PDF page ${pageNumber}`); const ctx = canvas.getContext('2d', { alpha: false }); ctx.scale(pixelRatio, pixelRatio); pageShell.appendChild(pageLabel); pageShell.appendChild(canvas); pagesEl.appendChild(pageShell); await page.render({ canvasContext: ctx, viewport, }).promise; } } catch (error) { fail(error); } })(); })(); </script>