Hugging Face's logo

Spaces:
Beexly
/
studio-hub
Running

App Files Community
studio-hub / index.html
Beexly's picture
Beexly
Upload folder using huggingface_hub
7aac957 verified
Raw History Blame Contribute Delete
10.5 kB
<!doctype html>
<html lang="en">
<head>
<meta charset="utf-8">
<meta name="viewport" content="width=device-width, initial-scale=1">
<title>Beexly Studio</title>
<style>
:root {
--bg: #0b0d12;
--card: #151822;
--border: #262b3a;
--text: #e8eaf0;
--muted: #9aa1b4;
--accent: #7c9cff;
}
@media (prefers-color-scheme: light) {
:root {
--bg: #f6f7fb;
--card: #ffffff;
--border: #e2e5ee;
--text: #16181f;
--muted: #5a6072;
--accent: #3f5fe0;
}
}
* { box-sizing: border-box; }
body {
margin: 0;
background: var(--bg);
color: var(--text);
font-family: -apple-system, BlinkMacSystemFont, "Segoe UI", Roboto, sans-serif;
padding: 48px 20px 80px;
}
.wrap { max-width: 960px; margin: 0 auto; }
h1 { font-size: 1.9rem; margin-bottom: 4px; }
.sub { color: var(--muted); margin-bottom: 8px; }
.notice {
font-size: 0.85rem;
color: var(--muted);
border: 1px solid var(--border);
border-radius: 10px;
padding: 10px 14px;
margin: 20px 0 36px;
}
.grid {
display: grid;
grid-template-columns: repeat(auto-fit, minmax(340px, 1fr));
gap: 16px;
margin-bottom: 40px;
}
.card {
background: var(--card);
border: 1px solid var(--border);
border-radius: 14px;
padding: 20px;
display: flex;
flex-direction: column;
gap: 10px;
}
.card h2 { margin: 0; font-size: 1.15rem; }
.card .tagline { color: var(--muted); font-size: 0.92rem; margin: 0; }
.limits {
font-size: 0.8rem;
color: var(--muted);
border-top: 1px solid var(--border);
padding-top: 10px;
line-height: 1.6;
}
.limits b { color: var(--text); }
.status {
display: inline-block;
font-size: 0.72rem;
font-weight: 600;
letter-spacing: 0.03em;
text-transform: uppercase;
padding: 2px 8px;
border-radius: 999px;
width: fit-content;
}
.status.live { background: rgba(90, 200, 130, 0.15); color: #4cbb7a; }
.status.blocked { background: rgba(230, 90, 90, 0.15); color: #e05a5a; }
a.open {
color: var(--accent);
text-decoration: none;
font-size: 0.9rem;
font-weight: 600;
}
a.open:hover { text-decoration: underline; }
table {
width: 100%;
border-collapse: collapse;
font-size: 0.88rem;
margin-bottom: 40px;
}
th, td {
text-align: left;
padding: 10px 12px;
border-bottom: 1px solid var(--border);
}
th { color: var(--muted); font-weight: 600; font-size: 0.78rem; text-transform: uppercase; }
.overflow { overflow-x: auto; }
footer {
margin-top: 48px;
color: var(--muted);
font-size: 0.8rem;
text-align: center;
}
</style>
</head>
<body>
<div class="wrap">
<h1>Beexly Studio</h1>
<p class="sub">Garrett's personal free-GPU studio suite, built on Hugging Face's free tier.</p>
<div class="notice">
Personal tooling only. Nothing here is connected to, or part of, any
Galaxy Sports Edge product, database, Stripe, or customer data — and
it never will be.
</div>
<div class="grid">
<div class="card">
<span class="status live">Live &amp; verified</span>
<h2>Studio Chat</h2>
<p class="tagline">General chat with presets for code, writing, and structured JSON output.</p>
<div class="limits">
<b>Model:</b> Qwen2.5-72B-Instruct, 4-bit (nf4) &middot; License: Qwen (custom)<br>
<b>Packed size:</b> 44.5GB<br>
<b>Measured:</b> ~13.4 tok/s, 200-token Code-preset reply in 40.8s<br>
<b>Limits:</b> max 1024 output tokens/reply, 120s ZeroGPU cap
</div>
<a class="open" href="https://huggingface.co/spaces/Beexly/studio-chat" target="_blank" rel="noopener">Open studio-chat &rarr;</a>
</div>
<div class="card">
<span class="status live">Live &amp; verified</span>
<h2>Studio Media</h2>
<p class="tagline">Image generation, speech transcription, text-to-speech, and voice cloning.</p>
<div class="limits">
<b>Image:</b> FLUX.1-dev (non-commercial license, personal use only) &mdash; 1024x1024 in <b>22.6&ndash;27.8s</b>, batch of 4 in <b>102.8s</b>; optional AntiBlur detail-boost LoRA and Detail Daemon sigma-schedule texture boost, both off by default and both real-call verified with visible before/after; prompt enhancer via studio-chat in <b>17.9s</b><br>
<b>Upscale:</b> Real-ESRGAN 2x/4x (BSD-3-Clause) &mdash; 2x in <b>9.4s</b>; optional GFPGAN face restore (for degraded photos only &mdash; regresses already-sharp FLUX output, see README)<br>
<b>Remove Background:</b> BiRefNet (MIT) &mdash; transparent-PNG cutout in <b>7.5s</b>, clean edges verified on a real generated image<br>
<b>Transcribe:</b> Whisper-large-v3-turbo (MIT) &mdash; SRT/VTT + word timestamps in <b>7.9s</b><br>
<b>Speak:</b> Kokoro-82M (Apache-2.0) &mdash; in <b>8.8s</b><br>
<b>Clone:</b> F5-TTS (CC-BY-NC-4.0, personal use) &mdash; own-voice only, consent enforced server-side, watermarked filename, in <b>13.1s</b><br>
<b>Music:</b> MusicGen-small (CC-BY-NC-4.0, personal use) &mdash; 10s clip in <b>12.5s</b>, real audio verified<br>
<b>Packed size:</b> ~36GB (Vision tab, FLUX.1-schnell "fast draft", and Chatterbox-Turbo TTS all tried and reverted &mdash; packing budget, VRAM capacity, and a hard diffusers version conflict respectively; see the Space's README)
</div>
<a class="open" href="https://huggingface.co/spaces/Beexly/studio-media" target="_blank" rel="noopener">Open studio-media &rarr;</a>
</div>
<div class="card">
<span class="status live">Live &amp; verified</span>
<h2>GSE Proof MCP</h2>
<p class="tagline">Independent, read-only MCP server over Galaxy Sports Edge's public proof/verification endpoints.</p>
<div class="limits">
<b>Deployed on HF PRO</b> (cpu-basic) after the free-tier 402 blocker was resolved<br>
<b>5 MCP tools, all real-call verified:</b> ledger, receipts list, receipt verify, calibration status, board coverage<br>
<b>No auth, no writes, rate-limited, 60s cache</b>
</div>
<a class="open" href="https://huggingface.co/spaces/Beexly/gse-proof-mcp" target="_blank" rel="noopener">Open gse-proof-mcp &rarr;</a>
</div>
<div class="card">
<span class="status live">Deployed</span>
<h2>Podcast Pipeline</h2>
<p class="tagline">Markdown/PDF/DOCX script + your voice clip &rarr; narrated episode. Plus Clip Studio: cut, reframe, and caption your own video.</p>
<div class="limits">
<b>Deployed on HF PRO</b> (cpu-basic), calls studio-chat and studio-media &mdash; no local models<br>
<b>Clip Studio, real-call verified:</b> scene-detect (PySceneDetect, BSD-3) in 7.5s, 9:16/1:1/16:9 reframe in 2.4s, caption burn (via studio-media Whisper) in 12.1s<br>
<b>Full episode production</b>: not yet run end-to-end at real script length
</div>
<a class="open" href="https://huggingface.co/spaces/Beexly/podcast-pipeline" target="_blank" rel="noopener">Open podcast-pipeline &rarr;</a>
</div>
<div class="card">
<span class="status blocked">Code-complete, hardware pending</span>
<h2>Studio Video</h2>
<p class="tagline">Text-to-video generation on ZeroGPU.</p>
<div class="limits">
<b>Model:</b> LTX-Video-0.9.7-distilled (Lightricks, LTX-Video Open Weights License, personal use) &mdash; diffusers-native, 7 recommended inference steps<br>
<b>Blocked:</b> this session's OAuth token can create/upload to the Space but the Hub's hardware-change endpoint (zero-a10g) 401s for OAuth tokens &mdash; a real platform restriction, not a code bug (confirmed: the Space's own code loads and fails cleanly with "no NVIDIA driver" on the cpu-basic fallback, proving the code path itself is sound)<br>
<b>One manual step needed:</b> Space Settings &rarr; change hardware to ZeroGPU
</div>
<a class="open" href="https://huggingface.co/spaces/Beexly/studio-video" target="_blank" rel="noopener">Open studio-video &rarr;</a>
</div>
</div>
<div class="overflow">
<table>
<thead>
<tr><th>Space</th><th>Tab</th><th>Measured limit</th><th>Best for</th></tr>
</thead>
<tbody>
<tr><td>studio-chat</td><td>Chat</td><td>120s / 1024 tokens, ~13.4 tok/s</td><td>General Q&amp;A, code, writing, JSON output</td></tr>
<tr><td>studio-media</td><td>Image (FLUX.1-dev)</td><td>22.6&ndash;27.8s / 1024x1024, 102.8s for a batch of 4</td><td>High-quality image generation</td></tr>
<tr><td>studio-media</td><td>Upscale</td><td>9.4s, 2x</td><td>Sharpen a generated image</td></tr>
<tr><td>studio-media</td><td>Remove Background</td><td>7.5s, clean edges</td><td>Cutouts for compositing/overlays</td></tr>
<tr><td>studio-media</td><td>Music</td><td>12.5s for a 10s clip</td><td>Background music for video/podcast</td></tr>
<tr><td>studio-media</td><td>Transcribe</td><td>7.9s, word-level timestamps</td><td>Captions, SRT/VTT export</td></tr>
<tr><td>studio-media</td><td>Speak</td><td>8.8s</td><td>Quick text-to-speech</td></tr>
<tr><td>studio-media</td><td>Clone</td><td>13.1s, own voice only</td><td>Narration in your own voice</td></tr>
<tr><td>gse-proof-mcp</td><td>&mdash;</td><td>Live, 5 tools verified</td><td>Verifying GSE receipts from Claude Code</td></tr>
<tr><td>podcast-pipeline</td><td>Episode Production</td><td>Deployed, full run untested</td><td>Full episode production</td></tr>
<tr><td>podcast-pipeline</td><td>Clip Studio</td><td>7.5s scene-detect, 2.4s reframe, 12.1s captions</td><td>Turning your own video into social clips</td></tr>
<tr><td>studio-video</td><td>Text to Video</td><td>Hardware pending (see card)</td><td>Short video generation, once live</td></tr>
</tbody>
</table>
</div>
<footer>
Free-tier ZeroGPU caps every account at 2 GPU Spaces (chat + media,
packed by modality). CPU Spaces running Python (Gradio or Docker)
require HF PRO ($9/mo) &mdash; only static pages like this one are free
on cpu-basic. Both gse-proof-mcp and podcast-pipeline are code-complete
and waiting on that decision.
</footer>
</div>
</body>
</html>