---
language:
- en
- hi
license: apache-2.0
tags:
- multimodal
- text-to-image
- image-to-image
- image-to-text
- vision
- ocr
- diffusers
- sashi
- sashi-1.0-vision
- 8k
- india
- varanasi
model_name: SaShi-1.0-Vision
pipeline_tag: visual-question-answering
inference: true
---

<div align="center">

# 👁️🎨 SaShi 1.0 Vision — Sovereign Indian Multimodal AI Ecosystem

**SaShi 1.0 Vision** is an all-in-one unified Visual Intelligence & Art engine developed by **[ftmdeveloperz006](https://huggingface.co/ftmdeveloperz006)** in **Varanasi (Banaras), Uttar Pradesh, India** (Two Brothers).

[![Model](https://img.shields.io/badge/Model-SaShi--1.0--Vision-indigo?style=for-the-badge)](https://huggingface.co/ftmdeveloperz006/SaShi-1.0-Vision)
[![Architecture](https://img.shields.io/badge/Architecture-T2I%20%7C%20I2I%20%7C%20I2T-blue?style=for-the-badge)]()
[![Resolution](https://img.shields.io/badge/Resolution-8K%20UHD%20Photorealistic-green?style=for-the-badge)]()
[![Origin](https://img.shields.io/badge/Origin-Varanasi%2C%20Bharat%20🇮🇳-orange?style=for-the-badge)]()

</div>

---

## 🌟 The 3-in-1 Unified Multimodal Engine

SaShi 1.0 Vision consolidates three essential visual AI capabilities into one unified framework:

```
                                ┌─────────────────────────┐
                                │   👁️ SaShi 1.0 Vision   │
                                └────────────┬────────────┘
                 ┌───────────────────────────┼───────────────────────────┐
                 ▼                           ▼                           ▼
        ┌──────────────────┐        ┌──────────────────┐        ┌──────────────────┐
        │ 🎨 Text-to-Image │        │ 🖼️ Image-to-Image│        │ 🔍 Image-to-Text │
        │      (T2I)       │        │      (I2I)       │        │   (I2T/Vision)   │
        │ 8K Ultra-HD Art  │        │ Style & Upscale  │        │ OCR & Visual QA  │
        └──────────────────┘        └──────────────────┘        └──────────────────┘
```

1. **Text-to-Image (T2I):** Generate photorealistic 8K images, Indian cultural monuments, temples, futuristic concepts, and DSLR portraits from natural language prompts.
2. **Image-to-Image (I2I):** Transform existing photos, modify styles, convert sketches to photorealistic art, and enhance lighting.
3. **Image-to-Text (I2T / Vision):** Visual Question Answering (VQA), reading Devanagari/English text from receipts and signboards (OCR), and understanding complex diagrams.

---

## 🛠️ Unified Python Code (T2I, I2I, and I2T)

```python
import torch
from PIL import Image
from diffusers import StableDiffusionPipeline, StableDiffusionImg2ImgPipeline, EulerAncestralDiscreteScheduler
from transformers import pipeline

device = "cuda" if torch.cuda.is_available() else "cpu"

# 1. Text-to-Image (T2I)
t2i_pipe = StableDiffusionPipeline.from_pretrained(
    "ftmdeveloperz006/SaShi-1.0-Vision",
    torch_dtype=torch.float16
).to(device)

def sashi_t2i(prompt):
    return t2i_pipe(prompt + ", 8k uhd photorealistic", num_inference_steps=30).images[0]

# 2. Image-to-Image (I2I)
i2i_pipe = StableDiffusionImg2ImgPipeline.from_pretrained(
    "ftmdeveloperz006/SaShi-1.0-Vision",
    torch_dtype=torch.float16
).to(device)

def sashi_i2i(init_image, prompt, strength=0.75):
    return i2i_pipe(prompt=prompt, image=init_image, strength=strength).images[0]

# 3. Image-to-Text (I2T / Vision Analysis)
vqa_pipe = pipeline("image-to-text", model="Salesforce/blip-image-captioning-large", device=0 if device=="cuda" else -1)

def sashi_i2t(image):
    return vqa_pipe(image)[0]["generated_text"]
```

---

## 🇮🇳 Origin & Creators
- **Name Origin:** **Sa-Shi** (Named after the Two Brothers who conceived and trained this AI model).
- **Location:** **Varanasi (Kashi / Banaras), Uttar Pradesh, India.**