README / README.md
YarMalik's picture
Org card: Gulf Arabic, Saudi today and Qatari next
4b6057f verified
|
Raw History Blame Contribute Delete
9.21 kB
---
title: README
emoji: 🎙️
colorFrom: green
colorTo: gray
sdk: static
pinned: false
---
<div style="font: 16px/1.6 ui-sans-serif, system-ui, -apple-system, 'Segoe UI', Roboto, Inter, Arial, sans-serif; color: #e6ede9; margin: 0; padding: 0;">
<div style="max-width: 980px; margin: 0 auto; padding: 40px 20px 8px; text-align: center;">
<img src="https://huggingface.co/spaces/VohoAI/README/resolve/main/logo.png" alt="voho.ai" width="200" style="display: inline-block;">
<p style="margin: 18px 0 4px; font-size: 30px; font-weight: 700; letter-spacing: -0.02em; color: #3fcb8b;">Voice AI for Gulf Arabic: Saudi and Qatari.</p>
<p style="margin: 0; color: #9aaaa3; font-size: 16px;">Open models for Arabic as it is spoken on the phone, not as it is written in the news. Saudi dialects (Najdi, Hijazi, Khaleeji) are trained, measured and published today. Qatari is next.</p>
<p style="margin: 22px 0 0;">
<a href="https://huggingface.co/VohoAI/voho-saudi-stt-small" style="display: inline-block; margin: 4px; padding: 9px 16px; border: 1px solid #1f6b48; border-radius: 999px; background: #0f2a1f; color: #c9f2dc; text-decoration: none; font-weight: 600;">🎙 Saudi STT</a>
<a href="https://huggingface.co/VohoAI/voho-saudi-chat-4b" style="display: inline-block; margin: 4px; padding: 9px 16px; border: 1px solid #1f6b48; border-radius: 999px; background: #0f2a1f; color: #c9f2dc; text-decoration: none; font-weight: 600;">💬 Saudi Chat 4B</a>
<a href="https://huggingface.co/VohoAI/voho-saudi-speak-0.6b" style="display: inline-block; margin: 4px; padding: 9px 16px; border: 1px solid #1f6b48; border-radius: 999px; background: #0f2a1f; color: #c9f2dc; text-decoration: none; font-weight: 600;">🗣 Saudi Speak</a>
<a href="https://huggingface.co/datasets/VohoAI/voho-saudi-dialogues" style="display: inline-block; margin: 4px; padding: 9px 16px; border: 1px solid #1f6b48; border-radius: 999px; background: #0f2a1f; color: #c9f2dc; text-decoration: none; font-weight: 600;">📚 Saudi Dialogues</a>
<a href="https://voho.ai" style="display: inline-block; margin: 4px; padding: 9px 16px; border-radius: 999px; background: #0d7a4c; color: #ffffff; text-decoration: none; font-weight: 700;">voho.ai →</a>
</p>
</div>
<div style="max-width: 980px; margin: 22px auto 0; padding: 0 20px;">
<div style="background: #121b18; border: 1px solid #24322d; border-radius: 16px; padding: 26px 28px; margin-bottom: 16px;">
<p style="margin: 0 0 10px; font-family: ui-monospace, Menlo, monospace; font-size: 12px; letter-spacing: 0.1em; text-transform: uppercase; color: #3fcb8b;">Head to head · held-out tests</p>
<div style="border-top: 1px solid #24322d; padding: 16px 0;">
<div style="font-size: 18px; font-weight: 700; color: #ffffff; line-height: 1.35;">Beats OpenAI's speech model on Saudi Arabic.</div>
<div style="color: #9aaaa3; font-size: 14px; margin-top: 4px;">Word errors on Saudi speech: <b style="color: #3fcb8b;">39.2%</b> vs 103.9%. 62% fewer.</div>
<div style="color: #74847d; font-size: 11.5px; margin-top: 6px; font-family: ui-monospace, Menlo, monospace; letter-spacing: 0.04em; text-transform: uppercase;">vs OpenAI Whisper Small · 4,582 held-out Saudi clips</div>
</div>
<div style="border-top: 1px solid #24322d; padding: 16px 0;">
<div style="font-size: 18px; font-weight: 700; color: #ffffff; line-height: 1.35;">Beats Google's Gemini at spoken Saudi, at a fraction of the size.</div>
<div style="color: #9aaaa3; font-size: 14px; margin-top: 4px;">chrF++ to real Saudi speech: <b style="color: #3fcb8b;">75.7</b> vs 68.7, with a 0.6B model.</div>
<div style="color: #74847d; font-size: 11.5px; margin-top: 6px; font-family: ui-monospace, Menlo, monospace; letter-spacing: 0.04em; text-transform: uppercase;">vs Gemini 2.5 Flash · 1,000 held-out Saudi sentences</div>
</div>
<div style="border-top: 1px solid #24322d; padding: 16px 0;">
<div style="font-size: 18px; font-weight: 700; color: #ffffff; line-height: 1.35;">Sounds more Saudi than Alibaba's Qwen, in a fifth of the words.</div>
<div style="color: #9aaaa3; font-size: 14px; margin-top: 4px;">Replies that read as Saudi: <b style="color: #3fcb8b;">89.8%</b> vs 62.3%. 6.4 words a reply vs 30.2.</div>
<div style="color: #74847d; font-size: 11.5px; margin-top: 6px; font-family: ui-monospace, Menlo, monospace; letter-spacing: 0.04em; text-transform: uppercase;">vs Qwen3-4B-Instruct · 400 held-out questions</div>
</div>
</div>
<div style="background: #121b18; border: 1px solid #24322d; border-radius: 16px; padding: 26px 28px; margin-bottom: 16px;">
<p style="margin: 0 0 12px; font-size: 22px; font-weight: 700; color: #ffffff;">The stack: hear → decide → say</p>
<div style="border-top: 1px solid #24322d; padding: 13px 0;">
<a href="https://huggingface.co/VohoAI/voho-saudi-stt-small" style="color: #3fcb8b; font-weight: 700; text-decoration: none;">voho-saudi-stt-small</a>
<span style="float: right; color: #9aaaa3; font-size: 13px;">CC BY-NC-SA</span>
<div style="color: #c4d0ca; font-size: 14px; margin-top: 2px;">Saudi speech to text · 0.24B</div>
</div>
<div style="border-top: 1px solid #24322d; padding: 13px 0;">
<a href="https://huggingface.co/VohoAI/voho-saudi-chat-4b" style="color: #3fcb8b; font-weight: 700; text-decoration: none;">voho-saudi-chat-4b</a>
<span style="float: right; color: #9aaaa3; font-size: 13px;">Apache 2.0</span>
<div style="color: #c4d0ca; font-size: 14px; margin-top: 2px;">Answers in spoken Najdi · 4B · GGUF</div>
</div>
<div style="border-top: 1px solid #24322d; padding: 13px 0;">
<a href="https://huggingface.co/VohoAI/voho-saudi-speak-0.6b" style="color: #3fcb8b; font-weight: 700; text-decoration: none;">voho-saudi-speak-0.6b</a>
<span style="float: right; color: #9aaaa3; font-size: 13px;">CC BY-NC-SA</span>
<div style="color: #c4d0ca; font-size: 14px; margin-top: 2px;">Formal Arabic to spoken Saudi · 0.6B · GGUF · ONNX</div>
</div>
<div style="border-top: 1px solid #24322d; padding: 13px 0;">
<a href="https://huggingface.co/datasets/VohoAI/voho-saudi-dialogues" style="color: #3fcb8b; font-weight: 700; text-decoration: none;">voho-saudi-dialogues</a>
<span style="float: right; color: #9aaaa3; font-size: 13px;">Apache 2.0</span>
<div style="color: #c4d0ca; font-size: 14px; margin-top: 2px;">13,156 Najdi conversations across eight enterprise sectors</div>
</div>
</div>
<div style="background: #121b18; border: 1px solid #24322d; border-radius: 16px; padding: 26px 28px; margin-bottom: 16px;">
<p style="margin: 0 0 14px; font-size: 22px; font-weight: 700; color: #ffffff;">Quick facts</p>
<div style="text-align: center;">
<div style="display: inline-block; vertical-align: top; width: 210px; margin: 5px; background: #0c1311; border: 1px solid #24322d; border-radius: 12px; padding: 14px; text-align: center;"><div style="color: #9aaaa3; font-size: 13px;">Gulf Arabic</div><div style="font-size: 18px; font-weight: 700; color: #ffffff; margin-top: 2px;">Saudi today · Qatari next</div></div>
<div style="display: inline-block; vertical-align: top; width: 210px; margin: 5px; background: #0c1311; border: 1px solid #24322d; border-radius: 12px; padding: 14px; text-align: center;"><div style="color: #9aaaa3; font-size: 13px;">Open models</div><div style="font-size: 18px; font-weight: 700; color: #ffffff; margin-top: 2px;">3 + 1 dataset</div></div>
<div style="display: inline-block; vertical-align: top; width: 210px; margin: 5px; background: #0c1311; border: 1px solid #24322d; border-radius: 12px; padding: 14px; text-align: center;"><div style="color: #9aaaa3; font-size: 13px;">Sizes</div><div style="font-size: 18px; font-weight: 700; color: #ffffff; margin-top: 2px;">0.24B to 4B</div></div>
<div style="display: inline-block; vertical-align: top; width: 210px; margin: 5px; background: #0c1311; border: 1px solid #24322d; border-radius: 12px; padding: 14px; text-align: center;"><div style="color: #9aaaa3; font-size: 13px;">Delivering</div><div style="font-size: 18px; font-weight: 700; color: #ffffff; margin-top: 2px;">Aramco AI call centre</div></div>
</div>
</div>
<div style="background: #121b18; border: 1px solid #24322d; border-radius: 16px; padding: 26px 28px; margin-bottom: 16px;">
<p style="margin: 0 0 10px; font-size: 22px; font-weight: 700; color: #ffffff;">Run one now</p>
<pre style="margin: 0; background: #0c1311; border: 1px solid #24322d; border-radius: 10px; padding: 14px 16px; color: #c9f2dc; overflow-x: auto;"><code>ollama run hf.co/VohoAI/voho-saudi-chat-4b-GGUF:Q4_K_M</code></pre>
<p style="margin: 12px 0 0; color: #9aaaa3;">Then ask, in Najdi: <span dir="rtl">أبي أحجز موعد بكرة الصبح، فيه وقت فاضي؟</span></p>
</div>
<p style="text-align: center; color: #74847d; font-size: 13px; margin: 18px 0 30px;">The chat model and dataset are Apache 2.0, for commercial use. The speech models are CC BY-NC-SA 4.0; for production speech use the <a href="https://docs.voho.ai" style="color: #3fcb8b;">Voho API</a>. · <a href="https://voho.ai/models" style="color: #3fcb8b;">voho.ai/models</a></p>
</div>
</div>