Upload 6 files

LICENSE

@@ -0,0 +1,22 @@
+
+MIT License
+
+Copyright (c) 2025 Eric Maldon
+
+Permission is hereby granted, free of charge, to any person obtaining a copy
+of this software and associated documentation files (the "Software"), to deal
+in the Software without restriction, including without limitation the rights
+to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
+copies of the Software, and to permit persons to whom the Software is
+furnished to do so, subject to the following conditions:
+
+The above copyright notice and this permission notice shall be included in all
+copies or substantial portions of the Software.
+
+THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
+AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
+OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+SOFTWARE.

README.md

@@ -0,0 +1,34 @@
+
+---
+title: AI Storyboard Creator
+emoji: 🎬
+colorFrom: blue
+colorTo: purple
+sdk: gradio
+app_file: app.py
+pinned: false
+---
+
+<p align="center">
+  <img src="ai-storyboard-banner.png" alt="AI Storyboard Creator banner" width="100%">
+</p>
+
+# AI Storyboard Creator
+
+## Overview
+Turn a single text prompt into a **mini storyboard**: the app generates 3–6 images, adds a short caption to each panel, renders a preview grid, and lets you **download a PDF**. Works on **CPU basic** and supports **ZeroGPU** (on-demand GPU).
+
+## Features
+- 3–6 panels with **style presets** (Realistic, Anime, Watercolor, Comic, Sketch).
+- **Caption strip** automatically added under each panel.
+- **PDF export** (one page with all panels).
+- CPU-friendly; faster with GPU/ZeroGPU.
+
+## How to Run Locally
+```bash
+pip install -r requirements.txt
+python app.py
+```
+
+## Acceptable Use
+This project is intended for educational and creative purposes. Do **not** use it to generate harmful, illegal, or offensive content. Please respect model licenses and the Hugging Face Acceptable Use Policy.

app.py

@@ -0,0 +1,200 @@
+import gradio as gr
+import torch, os, io, math
+from diffusers import StableDiffusionPipeline
+from transformers import pipeline, set_seed
+from PIL import Image, ImageDraw, ImageFont
+from spaces import GPU  # ZeroGPU
+
+# ------------------ Globals (CPU-safe) ------------------
+_txtgen = None
+_t2i_cpu = None
+
+STYLE_PRESETS = {
+    "Realistic photo": "realistic photography, 35mm, finely detailed, natural lighting, high dynamic range",
+    "Anime": "anime, vibrant colors, cel shading, highly detailed, clean lineart",
+    "Watercolor": "watercolor painting, soft edges, light wash, textured paper, gentle colors",
+    "Comic": "comic book style, halftone, bold lines, dramatic shading, vibrant colors",
+    "Sketch": "pencil sketch, cross-hatching, grayscale, rough paper texture",
+}
+NEGATIVE = "nsfw, nudity, gore, deformed, extra limbs, low quality, blurry, worst quality, lowres, text artifacts, watermark, logo"
+
+# ------------------ Loaders ------------------
+def get_txtgen_cpu():
+    """Text generator stays on CPU to avoid CUDA init in main process (required by ZeroGPU)."""
+    global _txtgen
+    if _txtgen is None:
+        _txtgen = pipeline("text-generation", model="distilgpt2", device=-1)
+    return _txtgen
+
+def get_t2i_cpu():
+    """CPU pipeline; heavy but safe for CPU basic hardware."""
+    global _t2i_cpu
+    if _t2i_cpu is None:
+        _t2i_cpu = StableDiffusionPipeline.from_pretrained(
+            "stabilityai/sd-turbo",
+            torch_dtype=torch.float32,
+            safety_checker=None
+        )
+        _t2i_cpu.enable_attention_slicing()
+    return _t2i_cpu
+
+# ------------------ GPU batch (ZeroGPU) ------------------
+@GPU(duration=120)
+def t2i_generate_batch_gpu(prompts, width, height, steps, guidance, negative_prompt):
+    """Runs entirely inside a GPU-allocated context (ZeroGPU)."""
+    pipe = StableDiffusionPipeline.from_pretrained(
+        "stabilityai/sd-turbo",
+        torch_dtype=torch.float16,
+        safety_checker=None
+    ).to("cuda")
+    images = []
+    for p in prompts:
+        img = pipe(
+            prompt=p,
+            negative_prompt=negative_prompt,
+            num_inference_steps=steps,
+            guidance_scale=guidance,
+            width=width,
+            height=height,
+        ).images[0]
+        images.append(img)
+    return images
+
+# ------------------ Helpers ------------------
+def build_prompt(user_prompt: str, style: str, panel_idx: int, num_panels: int) -> str:
+    style_desc = STYLE_PRESETS.get(style, "")
+    beat = ["opening shot", "rising action", "key moment", "twist", "resolution"]
+    beat_text = beat[min(panel_idx, len(beat)-1)]
+    return f"{user_prompt}, {style_desc}, storyboard panel {panel_idx+1} of {num_panels}, {beat_text}, cinematic composition, wide shot"
+
+def generate_captions(user_prompt: str, n: int = 4):
+    gen = get_txtgen_cpu()
+    base = f"Write {n} very short scene captions (max 10 words each) for a storyboard about: {user_prompt}. Number them 1..{n}. Keep it safe and family-friendly."
+    out = gen(base, max_new_tokens=80, temperature=0.9, top_p=0.95, do_sample=True)[0]["generated_text"]
+    lines = [l.strip() for l in out.split("\n") if l.strip()]
+    caps = []
+    for l in lines:
+        # Accept "1. xxx", "1) xxx", "1 - xxx"
+        if l[0:1].isdigit():
+            parts = l.split(maxsplit=1)
+            if len(parts) > 1:
+                caps.append(parts[1].strip(" .)"))
+    while len(caps) < n:
+        caps.append(f"Scene {len(caps)+1}")
+    return caps[:n]
+
+def grid(images, cols=2, bg=(20,20,20), pad=8):
+    if not images:
+        return None
+    w, h = images[0].size
+    rows = math.ceil(len(images)/cols)
+    grid_w = cols*w + (cols+1)*pad
+    grid_h = rows*h + (rows+1)*pad + 40
+    canvas = Image.new("RGB", (grid_w, grid_h), bg)
+    draw = ImageDraw.Draw(canvas)
+    title = "Storyboard"
+    try:
+        font = ImageFont.truetype("DejaVuSans.ttf", 22)
+    except Exception:
+        font = ImageFont.load_default()
+    tw, th = draw.textsize(title, font=font)
+    draw.text(((grid_w-tw)//2, pad//2), title, fill=(230,230,230), font=font)
+    for i, img in enumerate(images):
+        r = i // cols
+        c = i % cols
+        x = pad + c*(w+pad)
+        y = pad + th + pad + r*(h+pad)
+        canvas.paste(img, (x,y))
+    return canvas
+
+def images_to_pdf(images):
+    if not images:
+        return None
+    pil_images = [im.convert("RGB") for im in images]
+    buf = io.BytesIO()
+    pil_images[0].save(buf, format="PDF", save_all=True, append_images=pil_images[1:])
+    buf.seek(0)
+    return buf
+
+# ------------------ Main ------------------
+def create_storyboard(user_prompt, style, num_panels, width, height, seed):
+    if not user_prompt or not user_prompt.strip():
+        return None, None, "Please enter a prompt."
+    if seed and str(seed).strip().isdigit():
+        set_seed(int(seed))
+
+    # Build prompts and captions
+    captions = generate_captions(user_prompt, n=num_panels)
+    prompts = [build_prompt(user_prompt, style, i, num_panels) for i in range(num_panels)]
+
+    # Try GPU batch (ZeroGPU). If fails (no GPU), fallback to CPU.
+    images = None
+    try:
+        images = t2i_generate_batch_gpu(prompts, width, height, steps=2, guidance=0.0, negative_prompt=NEGATIVE)
+    except Exception as e:
+        # GPU not available or failed → CPU path
+        pipe = get_t2i_cpu()
+        images = []
+        for p in prompts:
+            img = pipe(
+                prompt=p,
+                negative_prompt=NEGATIVE,
+                num_inference_steps=4,
+                guidance_scale=0.0,
+                width=width,
+                height=height,
+            ).images[0]
+            images.append(img)
+
+    # Add caption strips
+    out_panels = []
+    for img, cap in zip(images, captions):
+        draw = ImageDraw.Draw(img)
+        try:
+            font = ImageFont.truetype("DejaVuSans.ttf", size=max(16, width//28))
+        except Exception:
+            font = ImageFont.load_default()
+        text = cap[:80]
+        text_w, text_h = draw.textsize(text, font=font)
+        strip_h = text_h + 14
+        strip = Image.new("RGB", (img.width, strip_h), (0,0,0))
+        d2 = ImageDraw.Draw(strip)
+        d2.text(((img.width - text_w)//2, 7), text, font=font, fill=(255,255,255))
+        combined = Image.new("RGB", (img.width, img.height + strip_h), (0,0,0))
+        combined.paste(img, (0,0))
+        combined.paste(strip, (0,img.height))
+        out_panels.append(combined)
+
+    preview = grid(out_panels, cols=2)
+    pdf_buf = images_to_pdf(out_panels)
+    pdf_file = ("storyboard.pdf", pdf_buf) if pdf_buf else None
+    return preview, pdf_file, "Done."
+
+# ------------------ UI ------------------
+with gr.Blocks(title="AI Storyboard Creator") as demo:
+    gr.Markdown(
+        """
+        # AI Storyboard Creator
+        Turn a single prompt into a tiny storyboard: 3–6 panels, captions, and a downloadable PDF.  
+        Works on **CPU basic** and supports **ZeroGPU** (GPU on-demand).
+        """
+    )
+    with gr.Row():
+        prompt = gr.Textbox(label="Story prompt", lines=2, placeholder="A cyberpunk cat exploring a neon-lit Tokyo at night")
+    with gr.Row():
+        style = gr.Dropdown(label="Style", choices=list(STYLE_PRESETS.keys()), value="Comic")
+        num_panels = gr.Slider(3, 6, value=4, step=1, label="Number of panels")
+    with gr.Row():
+        width = gr.Slider(384, 768, value=448, step=64, label="Panel width (px)")
+        height = gr.Slider(384, 768, value=448, step=64, label="Panel height (px)")
+        seed = gr.Textbox(label="Seed (optional)", placeholder="e.g., 42")
+    btn = gr.Button("Create Storyboard")
+
+    info = gr.Markdown("> First run may download models (approx. 1–2 minutes).")
+    preview = gr.Image(label="Preview (grid)", show_download_button=False)
+    pdf = gr.File(label="Download PDF")
+    msg = gr.Markdown()
+    btn.click(create_storyboard, inputs=[prompt, style, num_panels, width, height, seed], outputs=[preview, pdf, msg])
+
+if __name__ == "__main__":
+    demo.launch()

requirements.txt

@@ -0,0 +1,8 @@
+gradio>=4.36.1
+diffusers>=0.29.0
+transformers>=4.41.0
+accelerate>=0.30.0
+torch
+safetensors
+pillow
+spaces>=0.19.0

runtime.txt

@@ -0,0 +1 @@
+python-3.10