Seedream 4

This graph is a super easy example of how to use Seedream 4 model to generate a 21:9 large 2k icon for our node browser entries. Seedream 4 model takes 2 inputs, one for prompt, and another one for images. Both are multiple inputs. This example shows the making of the icon for Sora 2 model.

# Short Film From Story Idea ## Overview This workflow turns a raw story idea into a structured, AI-assisted short film package. It combines multi-step text generation and visual concept art so you can move from concept to shot-ready pre‑production in a single Nodespell graph. ## What You'll Build - A concise story breakdown with clear cinematic beats. - A 10‑shot film shot list derived from your short story. - Deeply refined descriptions for key shots (including shots #5 and #10). - High‑res character and scene reference images for visual development. ## How It Works 1. Your story idea or short story enters through multiple input nodes (23 inputs in total), feeding a complex graph of 52 nodes and 77 connections. 2. Fifteen **Gemini text** nodes, configured with the `gemini-2.5-flash` model (up to 3000 tokens for longer outputs), interpret the narrative, extract key beats, and transform the prose into a filmable structure. 3. Guided by embedded sticky note instructions, the workflow converts the narrative into a ~10‑shot list (e.g., the note that instructs: “Convert the short story into a shot list, roughly 10 shots in length”). 4. Dedicated refinement passes focus on pivotal moments: one set of Gemini text nodes expands shot #5, and another expands shot #10, adding detailed camera, performance, and environment notes suitable for directors and cinematographers. 5. Twelve **Seedream4** image nodes (configured at 2K, 2048×2048, 16:9, single-image output) use prompts such as the full‑body character portrait note to generate high-quality character and keyframe-style visuals. 6. A **hailuo23Fast** text node can provide additional fast drafting or alternate takes, giving you variation on dialogue, action, or pacing. ## Best For - Filmmakers and indie directors turning written ideas into ready-to-shoot plans. - Writers who want to visualize their short stories as films. - Storyboard artists and pre‑production teams needing structured shot lists and references. - Creators prototyping short films, teasers, or proof‑of‑concepts with AI. Try this snippet in Nodespell to turn your next story idea into a ready-to-shoot short film package.

# Fashion Print Design ## Overview This Fashion Print Design workflow turns your fabric motifs and reference photos into production-ready dress visuals and motion previews. It combines multiple image models to apply prints to cotton garments with realistic studio lighting and consistent color. ## What You'll Build - High‑resolution **1:1 garment mockups** with your print applied across the full dress. - 2K **fabric texture tiles** suitable for textile sampling or e‑commerce. - Short **5‑second fashion clips** that showcase the dress and print in motion. - Iterative concept boards that stay aligned to your fashion print moodboard. ## How It Works 1. A moodboard image input (e.g., `fashion_print_moodboard`) anchors the overall style, palette, and motif direction. 2. Multiple **Seedream 4** nodes (25 total, key ones like `seedream9`, `seedream11`, `seedream13`) generate 2048×2048, 2K print swatches and fabric renders at a **1:1 aspect ratio**, optimized for color consistency where Nano Banana struggles with flower tones. 3. **googleNanoBanana** nodes (7 total, JPG output, 1:1) support fast ideation passes, while sticky notes guide prompts such as changing the dress material to match the reference and ensuring the print pattern wraps cleanly across the garment. 4. A dedicated instruction note drives photorealism: applying the print texture to **cotton fabric** under clear, realistic studio lighting. 5. **qwenImageEditPlus** and **qwenImage** refine fit, fabric details, and print placement, while **reveCreate** and `hailuo23Fast` assist with stylistic variations and composition. 6. **kling25ImageToVideo** nodes transform key frames into **5‑second videos** (CFG scale 0.5, negative prompt to avoid blur, distortion, and low quality), giving you animated fashion previews. ## Best For - Fashion and textile designers developing new print collections. - Apparel brands needing fast dress and fabric mockups from reference art. - Surface pattern designers pitching prints to clothing labels. - E‑commerce teams creating on‑model visuals and motion previews without a full photoshoot. - Creative studios prototyping AI‑assisted fashion print design workflows. Try this Fashion Print Design snippet in Nodespell to turn flat print references into polished, motion‑ready fashion visuals in a few guided steps.

## Overview This Nodespell snippet is an **AI sunglasses mockup workflow** that turns multiple reference photos and style notes into polished, production-ready eyewear visuals. It generates new sunglasses designs that match frame shape, lens color, and viewing angle instructions. ## What You'll Build - High-resolution (2K) sunglasses product renders based on Etsy-style reference images. - Front and side-view variations, including a right-hand side profile of the same frame. - Customised frame silhouettes with more curved, rounded outer edges and refined corners. - Consistent lens colour and finish driven by a dedicated lens colour reference image. ## How It Works 1. **Reference intake via stickyImage nodes** – Seven `stickyImage` nodes load base inspiration images from Etsy URLs plus a dedicated `lens_colour_ref` node to lock in lens tint and finish. 2. **Design intent with stickyNote prompts** – Fourteen `stickyNote` nodes (e.g. `sticky_note4`, `sticky_note5`, `sticky_note18`) specify shape changes, side-view requirements, and which reference to follow for frames vs. lenses. 3. **Primary 2K generation with seedream4** – Four `seedream4` nodes generate core sunglasses renders at 2048×2048 resolution (`size: 2K`, `aspect_ratio: match_input_image`, `max_images: 1`) based on the combined textual and visual guidance. 4. **Variant and layout handling with googleNanoBanana** – Nine `googleNanoBanana` nodes create 16:9 PNG outputs for marketing-ready images, web product cards, and banner layouts (`aspect_ratio: 16:9`, `output_format: png`). 5. **Detail edits with qwenImageEditPlus** – A single `qwenImageEditPlus` node performs targeted refinements like softening frame edges, rounding corners, and aligning the side view to the front-view design. 6. **Complex graph orchestration** – The 35 nodes and 38 connections coordinate 21 inputs and 12 outputs, ensuring reference images, notes, and model calls stay in sync through the full design cycle. ## Best For - Eyewear brands and independent makers prototyping new sunglasses lines. - Etsy and DTC sellers needing fast, on-brand sunglasses mockups. - Product designers exploring frame variations without manual 3D work. - Marketers creating consistent hero images and ad creatives from a few reference photos. Try this snippet in Nodespell to rapidly turn your reference shots into polished AI-generated sunglasses visuals ready for product pages and campaigns.

## Overview This Nodespell snippet is a multi-dish **AI food image prompt and generation workflow**. It turns simple dish ideas into detailed visual prompts, then renders high‑resolution 4:3 food images using Google Nano Banana and Seedream models. ## What You'll Build - A reusable pipeline that expands dish concepts into rich, camera-ready image prompts. - 2K, 4:3 food photos for menus, blogs, or social media, exported as JPGs. - Parallel image variants for multiple dishes in a single run. - Optional text extras (like jokes or captions) powered by Gemini 2.5 Flash. ## How It Works 1. You describe dishes or ingredients through the 11 input nodes, guided by 10 **stickyNote** instructions (for example, “Generate a detailed prompt for image generation of the dish #1/#6, include all visible ingredients”). 2. Eight **geminiText** nodes call the **gemini-2.5-flash** model to expand each dish into a scene-level prompt: plating, lighting, background, camera style, and visible ingredients. 3. These enriched prompts fan out into eight **googleNanoBanana** nodes configured to a **4:3 aspect ratio** and **JPG** output, generating fast concept images and low-cost visual drafts. 4. Once satisfied with prompts, they feed into six **seedream** image nodes (seedream4/5/6/7/8) set to **2K resolution (2048×2048), 4:3, max_images: 1, sequential_image_generation: disabled** for sharp, production-ready renders. 5. A **stickyImage** node can serve as an optional visual reference, helping align AI output to a brand or photography style. 6. Twelve output nodes collect final images and text so you can review, compare dishes, and export for menus, posts, or recipe apps. ## Best For - Food bloggers and creators needing consistent, high-quality dish imagery. - Restaurant owners and menu designers prototyping layouts and specials. - Recipe platforms and cooking apps generating scalable visual libraries. - AI artists and prompt engineers exploring food photography styles. - Marketing teams producing rapid A/B-tested visuals for campaigns. Try this snippet in Nodespell to rapidly turn raw dish ideas into polished, high-resolution food visuals.