Team Ai
Modelpublic

aixk/BAAR5-IMG-335M

sourceHugging Faceotherupdated 7d agoView on Hugging Face
0likes4downloads
Model Card

<p align="center"> <a href="https://baar.uk/model/"> <img src="https://cdn.jsdelivr.net/npm/minicon1@latest/logo/baar.png" alt="BAAR Logo" width="180"> </a> </p>

<h1 align="center">BAAR4-IMG-335M (Multilingual)</h1>

<p align="center"> ๐ŸŒ <b>Official Website:</b> <a href="https://baar.uk/model/">https://baar.uk/model/</a></br> ๐Ÿ“„ <b>License:</b> Non-Commercial Only (Contact for Commercial Pricing) </p>

<p align="center"> <img src="https://cdn-uploads.huggingface.co/production/uploads/657c6756fa474954546351ef/eiB_YciHsLamLMOaMjL-l.png" alt="BAAR4-IMG Sample Output" width="600"> </p>


๐Ÿ–ผ๏ธ Gallery

<Gallery />


๐Ÿ“Œ Model Overview

BAAR4-IMG-335M-Realistic4 is an ultra-lightweight, native multilingual text-to-image generation model engineered with approximately 335M parameters. Built from the ground up for extreme efficiency, this model understands direct prompts across 20 major languages (Korean, English, German, Spanish, French, Japanese, Chinese, Italian, Dutch, Polish, Portuguese, Russian, Arabic, Bulgarian, Greek, Hindi, Swahili, Thai, Turkish, and Vietnamese) without requiring external translation pipelines.

Specifically architected to run directly inside client-side web browsers (WebGPU / ONNX Runtime Web / WebAssembly) and edge hardware without relying on expensive server-side GPU instances, it delivers crisp, high-fidelity photorealistic visual generation while maintaining a compact footprint and INT8 quantization friendliness.

BAAR4-IMG-335M-Realistic4๋Š” ์•ฝ 3.35์–ต(335M) ํŒŒ๋ผ๋ฏธํ„ฐ ๊ทœ๋ชจ๋กœ ์„ค๊ณ„๋œ ์ดˆ๊ฒฝ๋Ÿ‰ ๋‹ค๊ตญ์–ด ์ง€์› ํฌํ† ๋ฆฌ์–ผ๋ฆฌ์Šคํ‹ฑ ํ…์ŠคํŠธ-ํˆฌ-์ด๋ฏธ์ง€(Text-to-Image) ์ƒ์„ฑ ๋ชจ๋ธ์ž…๋‹ˆ๋‹ค. ๋ณ„๋„์˜ ๋ฒˆ์—ญ ํŒŒ์ดํ”„๋ผ์ธ ์—†์ด ํ•œ๊ตญ์–ด, ์˜์–ด, ๋…์ผ์–ด, ์ŠคํŽ˜์ธ์–ด, ํ”„๋ž‘์Šค์–ด, ์ผ๋ณธ์–ด, ์ค‘๊ตญ์–ด, ์ดํƒˆ๋ฆฌ์•„์–ด, ๋„ค๋œ๋ž€๋“œ์–ด, ํด๋ž€๋“œ์–ด, ํฌ๋ฅดํˆฌ๊ฐˆ์–ด, ๋Ÿฌ์‹œ์•„์–ด, ์•„๋ž์–ด, ๋ถˆ๊ฐ€๋ฆฌ์•„์–ด, ๊ทธ๋ฆฌ์Šค์–ด, ํžŒ๋””์–ด, ์Šค์™€ํž๋ฆฌ์–ด, ํƒœ๊ตญ์–ด, ํŠ€๋ฅดํ‚ค์˜ˆ์–ด, ๋ฒ ํŠธ๋‚จ์–ด ๋“ฑ 20๊ฐœ ์ฃผ์š” ์–ธ์–ด์˜ ์ž์—ฐ์–ด ํ”„๋กฌํ”„ํŠธ๋ฅผ ์ง์ ‘ ์ดํ•ดํ•˜๊ณ  ์ด๋ฏธ์ง€๋ฅผ ์ƒ์„ฑํ•ฉ๋‹ˆ๋‹ค.

๊ณ ๋น„์šฉ ํด๋ผ์šฐ๋“œ GPU ์ธํ”„๋ผ ์˜์กด์„ฑ์„ ์ตœ์†Œํ™”ํ•˜๊ณ , ํด๋ผ์ด์–ธํŠธ ์›น ๋ธŒ๋ผ์šฐ์ €(WebGPU, ONNX Runtime Web, Wasm) ๋ฐ ๋ชจ๋ฐ”์ผยท์—ฃ์ง€(Edge) ์˜จ๋””๋ฐ”์ด์Šค ํ™˜๊ฒฝ์—์„œ ์‹ค์‹œ๊ฐ„ ๋ Œ๋”๋ง์ด ๊ฐ€๋Šฅํ•˜๋„๋ก ์ตœ์ ํ™”๋˜์—ˆ์Šต๋‹ˆ๋‹ค. ๊ทน๋„๋กœ ๋‚ฎ์€ ๋ฉ”๋ชจ๋ฆฌ(VRAM) ์ ์œ ์œจ๊ณผ INT8 ์–‘์žํ™”๋ฅผ ์™„๋ฒฝํ•˜๊ฒŒ ์ง€์›ํ•˜๋ฉด์„œ๋„ ์‚ฌ์‹ค์ ์ด๊ณ  ์„ฌ์„ธํ•œ ์‹ค์‚ฌํ’ ์ด๋ฏธ์ง€ ์ƒ์„ฑ์„ ์‹คํ˜„ํ•ฉ๋‹ˆ๋‹ค.


๐Ÿ“„ License & Commercial Terms (๋ผ์ด์„ ์Šค ๋ฐ ์ƒ์—…์  ์ด์šฉ ์•ˆ๋‚ด)

  • โ€”Non-Commercial Use Only (๋น„์ƒ์—…์  ์—ฐ๊ตฌยท๊ฐœ์ธ ์ด์šฉ ํ•œ์ •): This model is strictly available for non-commercial, educational, and research purposes only. Any form of monetization (direct or indirect) without an explicit commercial license is strictly prohibited. ๋ณธ ๋ชจ๋ธ์€ ์ˆœ์ˆ˜ ์—ฐ๊ตฌ, ๊ต์œก ๋ฐ ๊ฐœ์ธ ๋น„์˜๋ฆฌ ๋ชฉ์ ์— ํ•œํ•ด ๋ฌด๋ฃŒ๋กœ ์‚ฌ์šฉํ•  ์ˆ˜ ์žˆ์Šต๋‹ˆ๋‹ค. ์‚ฌ์ „ ํ—ˆ๊ฐ€ ์—†๋Š” ์ผ์ฒด์˜ ์ˆ˜์ต ์ฐฝ์ถœ ํ–‰์œ„(์งยท๊ฐ„์ ‘์  ์ƒ์—…์  ์ด์šฉ)๋Š” ์—„๊ฒฉํžˆ ๊ธˆ์ง€๋ฉ๋‹ˆ๋‹ค.
  • โ€”Monetization & Commercial Licensing Inquiries (์ˆ˜์ต ์ฐฝ์ถœ ๋ฐ ์ƒ์šฉ ๋ผ์ด์„ ์Šค ๋ฌธ์˜): If you wish to monetize, integrate this model into commercial products, provide paid web apps/APIs, or require customized enterprise deployments, you must obtain a separate commercial license. Please reach out via email for licensing terms and enterprise adoption. ๋ณธ ๋ชจ๋ธ์„ ํ™œ์šฉํ•˜์—ฌ ์ˆ˜์ต์„ ์ฐฝ์ถœํ•˜๊ณ ์ž ํ•˜๊ฑฐ๋‚˜, ์ƒ์šฉ ์›น ์„œ๋น„์Šค/์œ ๋ฃŒ ์•ฑ์— ํƒ‘์žฌ, ๋˜๋Š” ๊ธฐ์—…์šฉ ์˜จ๋””๋ฐ”์ด์Šค ์†”๋ฃจ์…˜ ๋„์ž… ๋ฐ ๊ธฐ์ˆ  ์ง€์›์ด ํ•„์š”ํ•œ ๊ฒฝ์šฐ ๋ฐ˜๋“œ์‹œ ๋ณ„๋„์˜ ์ƒ์šฉ ๋ผ์ด์„ ์Šค๋ฅผ ์ทจ๋“ํ•˜์…”์•ผ ํ•ฉ๋‹ˆ๋‹ค. ๋„์ž… ๋ฐ ๋ผ์ด์„ ์Šค ๋ฐœ๊ธ‰์€ ์•„๋ž˜ ์ด๋ฉ”์ผ๋กœ ๋ฌธ์˜ํ•ด ์ฃผ์‹œ๊ธฐ ๋ฐ”๋ž๋‹ˆ๋‹ค.
  • โ€”Commercial Inquiries (์ƒ์šฉ ๋ผ์ด์„ ์Šค ๋ฌธ์˜): admin@099.kr
  • โ€”Official Website (๊ณต์‹ ์›น์‚ฌ์ดํŠธ): https://baar.uk/model/

โš ๏ธ Notes (์•ˆ๋‚ด ์‚ฌํ•ญ)

  • โ€”Multilingual Native Prompting: Generates images directly from native language prompts (ko, en, de, es, fr, ja, zh, it, nl, pl, pt, ru, ar, bg, el, hi, sw, th, tr, vi) without translation latency.
  • โ€”Client-Side Optimization: Designed specifically for WebGPU, WebAssembly (Wasm), and INT8 quantization pipelines (diffusers, ONNX Runtime).
  • โ€”Recommended Inference Setup: For optimal image fidelity and prompt comprehension, Euler A / DPM++ 2M Karras schedulers with 15โ€“25 inference steps at 512ร—512 resolution are recommended.
  • โ€”๋‹ค๊ตญ์–ด ๋„ค์ดํ‹ฐ๋ธŒ ํ”„๋กฌํ”„ํŠธ ์ง€์›: ๋ฒˆ์—ญ API ํ˜ธ์ถœ ์—†์ด 20๊ฐœ ์–ธ์–ด๋กœ ์ž‘์„ฑ๋œ ํ”„๋กฌํ”„ํŠธ๋ฅผ ์ง๊ด€์ ์œผ๋กœ ์ดํ•ดํ•˜๊ณ  ๊ณ ์œ  ๋‰˜์•™์Šค๋ฅผ ๋ฐ˜์˜ํ•œ ์ด๋ฏธ์ง€๋ฅผ ์ƒ์„ฑํ•ฉ๋‹ˆ๋‹ค.
  • โ€”ํด๋ผ์ด์–ธํŠธ ์ตœ์ ํ™” ์„ค๊ณ„: WebGPU, WebAssembly, INT8 ์–‘์žํ™” ๋ฐฐํฌ ํŒŒ์ดํ”„๋ผ์ธ์— ์ตœ์ ํ™”๋˜์–ด ์žˆ์Šต๋‹ˆ๋‹ค.
  • โ€”๊ถŒ์žฅ ์ถ”๋ก  ํŒŒ๋ผ๋ฏธํ„ฐ: ๊ณ ํ’ˆ์งˆ ์‹ค์‚ฌ ๋ Œ๋”๋ง์„ ์œ„ํ•ด 512ร—512 ํ•ด์ƒ๋„ ๊ธฐ์ค€ 15~25 ์Šคํ…์˜ Euler A ๋˜๋Š” DPM++ 2M Karras ์Šค์ผ€์ค„๋Ÿฌ ์‚ฌ์šฉ์„ ๊ถŒ์žฅํ•ฉ๋‹ˆ๋‹ค.

๐Ÿ’ผ Opportunities & Contact (์ฑ„์šฉ ์ œ์•ˆ ๋ฐ ํˆฌ์ž ๋ฌธ์˜)

This project demonstrates practical competency in custom generative model architecture design, multilingual cross-attention integration, model compression/quantization (INT8/Edge), and zero-server-cost in-browser inference deployment under compute constraints.

I am actively seeking AI Engineering / Research opportunities, team recruitment offers, and project investment/partnerships.

  • โ€”Official Model Site (๋ชจ๋ธ ๊ณต์‹ ์›น์‚ฌ์ดํŠธ): https://baar.uk/model/
  • โ€”Model Adoption & Licensing (๋„์ž… ๋ฐ ๋ผ์ด์„ ์Šค ๋ฌธ์˜): admin@099.kr
  • โ€”Recruitment & Hiring Inquiries (์ฑ„์šฉ ๋ฐ ํฌ์ง€์…˜ ์ œ์•ˆ): admin@099.kr
  • โ€”Investment & Technical Collaboration (ํˆฌ์ž, ํŒŒํŠธ๋„ˆ์‹ญ ๋ฐ ๊ธฐ์ˆ  ํ˜‘์—…): admin@099.kr

์ œํ•œ๋œ ํ•˜๋“œ์›จ์–ด ๋ฐ ํด๋ผ์ด์–ธํŠธ ํ™˜๊ฒฝ(์›น, ๋ชจ๋ฐ”์ผ, ์—ฃ์ง€)์—์„œ ๊ตฌ๋™ ๊ฐ€๋Šฅํ•œ ์ดˆ๊ฒฝ๋Ÿ‰ ๋‹ค๊ตญ์–ด ์ƒ์„ฑํ˜• AI ์•„ํ‚คํ…์ฒ˜๋ฅผ ์ง์ ‘ ์„ค๊ณ„ํ•˜๊ณ  ์ตœ์ ํ™” ํŒŒ์ดํ”„๋ผ์ธ์„ ๊ตฌ์ถ•ํ•  ์ˆ˜ ์žˆ๋Š” AI ์—”์ง€๋‹ˆ์–ด์ž…๋‹ˆ๋‹ค. ์ €์˜ ๊ธฐ์ˆ ์  ์—ญ๋Ÿ‰ ์˜์ž…(์ฑ„์šฉ)์ด๋‚˜ ํ”„๋กœ์ ํŠธ ํ˜‘์—…/ํˆฌ์ž์— ๊ด€์‹ฌ์ด ์žˆ์œผ์‹  ๊ธฐ์—… ๋ฐ ํŒ€์˜ ์—ฐ๋ฝ์„ ๊ธฐ๋‹ค๋ฆฝ๋‹ˆ๋‹ค.