SilentCanoe
SilentCanoe /
🌿
Anne Story Generator
☕ Buy me a coffee 100% Local

"I'm so glad I live in a world where there are Octobers." — Anne Shirley

Free AI Photo Story Generator — Works Offline

🔒

Your photos never leave your device. Both models — ViT-GPT2 (image understanding) and Llama 3.2 (story generation) — run entirely in your browser using WebAssembly and WebGPU.

📸 Your Photos

Add up to 12 photos · drag to reorder

🌸

Drop photos here

JPG, PNG, WebP, HEIC

✍️ Story Settings

🤖 AI Engines

ViT-GPT2 Vision (~246 MB)
Image understanding
Llama 3.2 3B (~2 GB)
Story generation

Models download once and cache locally

🌿

Every photo holds a story

Upload your photos and let Anne Shirley's spirit weave them into a heartwarming Avonlea-style tale

📸 Add up to 12 photos
🤖 AI understands each photo
🌿 Anne crafts your story
💌 Export as PDF postcard or video

Frequently Asked Questions

How does the AI understand my photos?

The tool uses ViT-GPT2, a vision-language model that runs locally in your browser via WebAssembly. It generates a descriptive caption for each photo, which is then passed to Llama 3.2 to write the Anne Shirley-style story.

Do the AI models need to download every time?

No. ViT-GPT2 (~246 MB) and Llama 3.2 3B (~2 GB) are cached in your browser after the first download. On subsequent visits the models load from your local cache instantly, with no internet required.

What export formats are available?

You can export a PDF postcard (with photo + caption, in Avonlea Green, Ivory Parchment, or Moonlit Night styles at 4×6", 5.5×8.5", or 6×6" sizes), a Ken Burns-style video slideshow (WebM), or social media images optimised for Facebook (1200×630), Instagram (1080×1080), or Stories (1080×1920).

Can I edit the generated captions?

Yes — click the pencil icon (✏️) on any story chapter to open an editable text area. Your changes are saved automatically and used in all exports. You can also regenerate individual captions or the entire story with one click.

What browsers support the AI features?

Llama 3.2 story generation requires WebGPU (Chrome 113+ or Edge 113+ on desktop). ViT-GPT2 image captioning uses WebAssembly and works on all modern browsers, including Firefox and Safari. Template-based fallback captions are provided if Llama 3.2 cannot load.