DEV Community: WEI LIU

CADCraft: an open-source Rust 2D CAD workflow for automation

WEI LIU — Sat, 24 Oct 2026 18:55:53 +0000

When a project needs precise technical drawings and a workflow that can be scripted, CADCraft is a free, open-source 2D CAD application built in Rust. It is designed for architects, engineers, students, educators, makers, and technical designers who want a familiar drafting workflow without a subscription or account requirement.

A practical drafting core

CADCraft includes command-line tools for common drafting operations such as LINE, PLINE, CIRCLE, ARC, RECTANG, OFFSET, TRIM, FILLET, HATCH, and DIMLINEAR. Users can work with absolute or relative coordinates, distances, and angles. Object snaps, polar tracking, orthographic drawing, autocomplete, command history, aliases, and repeated commands help keep repetitive drafting work precise.

Drawings can be organized with layers and extended with associative dimensions, multileaders, tables, hatches, paper-space layouts, and plotting features. The application supports DWG and DXF workflows and can export designs to PDF, SVG, and PNG.

Built for automation

The interesting part for developer workflows is the built-in automation surface. CADCraft exposes MCP, JSON control, and a headless CLI so users and agents can create drawings, inspect entities, render screenshots, convert files, and verify results. That makes it possible to keep technical drawing steps closer to scripts, repeatable build jobs, and agent-assisted workflows instead of relying only on manual UI actions.

CADCraft runs on Windows, macOS, Linux, FreeBSD, and the web through WebAssembly. It is released under the MIT or Apache-2.0 licenses and is still in active development, so it is a project to evaluate, inspect, and contribute to as it evolves.

For anyone exploring open-source CAD with a programmable edge, the CADCraft project is worth a look.

GridCraft: An Open-Source Spreadsheet for Practical Data Work

WEI LIU — Thu, 22 Oct 2026 13:07:42 +0000

GridCraft is a free, open-source spreadsheet application created by ArtCraft for practical data work. It combines an Excel-style ribbon, worksheet tabs, a formula bar, formatting tools, charts, and PivotTables in one workspace.

With GridCraft, users can open, inspect, edit, and save XLSX files while keeping an original copy for safety. The application supports more than 500 worksheet functions, including SUM, IF, dynamic arrays, LET, and LAMBDA. Its calculation engine is written in Rust and runs locally through WebAssembly in the web version.

GridCraft works in browsers and as a desktop application for Windows, macOS, and Linux. It is distributed at no cost under the MIT and Apache 2.0 licenses. As alpha software, it has compatibility limitations with complex workbooks, specialist Excel functions, external data, and VBA macros, so important files should be backed up and checked after import or export.

Learn more at GridCraft or explore the source code on GitHub.

Ideogram AI 4.5: Sharp Typography, Native 2K Output, and Drift-Free Localized Editing

WEI LIU — Tue, 13 Oct 2026 11:19:32 +0000

Ideogram AI 4.5 is an advanced image generation and editing tool built for precise, high-quality visual creation. Its tagline sums up the workflow well: create boldly, edit precisely.

What it does

Sharp typography. Text rendered inside generated images stays crisp and legible, which matters for posters, labels, ads, and product shots.
Native 2K output. Results come out at 2K resolution without a separate upscaling step.
Localized, drift-free edits. Select an area, change it, and the rest of the image stays intact with minimal visual drift - so you can iterate on one detail without re-rolling the whole composition.

Who it is for

Anyone doing iterative visual work: marketers producing ad variants, designers exploring label or packaging concepts, and developers generating assets where consistent, readable text is part of the image.

Access

It follows a freemium model and launched on October 8, 2026. You can try it at ideogram-45.com.

If you work with AI image tools, the localized edit that preserves untouched pixels is the part worth testing first - it changes how many iterations a single concept can survive.

From Two Photos to Fairytale Motion: A Small AI Video Experiment

WEI LIU — Tue, 13 Oct 2026 08:21:59 +0000

Rumpelstiltskin AI is a creative video studio built around a small, simple prompt: start with two photos, choose a setting, add your direction, and generate an original AI-powered dance video.

Launched on October 8, 2026, Rumpelstiltskin AI is designed for playful, short-form storytelling. The product turns a pair of images into a whimsical fairytale scene, with the user guiding the setting and the creative direction rather than starting from a blank timeline.

The idea behind the workflow is to make the first step of visual creation feel approachable. Instead of assembling a full production, users can begin with images they already have, add a little context, and see that concept become motion. The result is meant to be a short scene—something you can share, remix, or use as a starting point for a bigger story.

Rumpelstiltskin AI uses a freemium model and is available at https://rumpelstiltskinai.org/.

If you are experimenting with playful AI video, what kind of setting would you choose for your own two-photo fairytale?

A Practical Guide to Better AI Image Prompts in the Browser

WEI LIU — Sun, 11 Oct 2026 19:08:10 +0000

AI image generation is most useful when the prompt is treated as a small, repeatable workflow rather than a one-shot command. This guide covers a simple loop for creating and refining images in the browser.

1. Start with the subject

Lead with the main subject, then add the setting, lighting, and mood. For example:

A ceramic mug on a linen tablecloth, soft window light

This gives the model a clear visual hierarchy before you add style details.

2. Quote text that must appear

If the image includes a sign, menu, poster, map, or chart, put the exact wording in quotation marks and say where it should appear. That makes the text requirement explicit instead of leaving it as an implied detail.

3. Add references with clear roles

For edits or multi-image compositions, name the role of each reference: the jacket from image 2, the person from image 1, and so on. Keeping the references explicit helps preserve the parts you care about.

4. Change one thing at a time

After the first generation, make small edits instead of rewriting the whole prompt. A focused request—recolor the wall, remove one object, or change the sky—makes it easier to keep the face, product, or layout stable.

A browser-based workflow

nanobanana 21 is a fast, simple image studio powered by Nano Banana 2.1. It supports high-quality image creation and editing, readable text, consistent characters, and up to 14 reference images. The workflow is intentionally lightweight: write a prompt, optionally add references, choose an aspect ratio and resolution, then generate and refine without code or a GPU.

The broader lesson is portable across image tools: describe the subject first, make constraints explicit, and iterate in small steps. That makes results easier to compare and easier to reproduce.

From Two Photos to a Cinematic Zombie Love Story: Riding the Zombie AI Trend

WEI LIU — Thu, 08 Oct 2026 03:06:37 +0000

The "zombie transformation" trend is all over short-video feeds right now: two photos in, one person slowly turns into a zombie, and the story ends with an unexpected emotional twist instead of a jump scare.

What ZombieTrend does

ZombieTrend is an AI video generator built around this single format. You upload two photos, and it produces a cinematic 9:16 vertical MP4:

One person gradually transforms into a zombie.
The other character chooses love over fear and holds them.
A flashback to their warmest shared memory closes the story.

Why the format works

The transformation gives you a hook in the first second.
The embrace flips the horror expectation into romance.
The flashback ending drives rewatches and shares.

Practical details

Output is 9:16 MP4, ready for TikTok, Reels, and Shorts.
There's a free preview before you commit to a full render.
Pricing is freemium, so you can test the format with zero risk.

If you have been looking for a fresh spin on couple content, the zombie love story format is worth a try — check it out at zombietrend.org.

From a Scene Idea to Short-Form Video with Genjutsu AI

WEI LIU — Wed, 07 Oct 2026 10:15:08 +0000

When a creative idea is clear but the production pipeline is not, a focused generation workflow can help turn that idea into something shareable.

Genjutsu AI is built for that first step: turning ideas into cinematic videos, product ads, social media shorts, and high-quality images in minutes.

The workflow is simple:

Choose a model.
Describe the scene you want to create.
Generate content in a format made for TikTok, Reels, YouTube Shorts, and more.

That makes it useful for exploring visual concepts, preparing product-ad variations, or moving from a rough scene description to a short-form asset without starting from a blank timeline.

If you are experimenting with AI-generated video and images, you can learn more at genjutsu-ai.org.

Qwen Image 2.1: More Control for AI Image Creation and Editing

WEI LIU — Fri, 02 Oct 2026 01:18:38 +0000

Developers working with image generation often need more control than a single prompt can provide.

Qwen Image 2.1 is an AI model for text-to-image creation and image editing. Its documented capabilities include detailed prompts, multiple reference images, portrait and typography refinement, and native transparent RGBA images.

That combination can be useful when prototyping visual interfaces, iterating on product concepts, or preparing assets that need transparent backgrounds.

A practical workflow

Start with a detailed prompt that describes the visual goal.
Add reference images when preserving a visual direction matters.
Refine portraits or typography in a focused iteration instead of rebuilding the whole concept.
Export transparent RGBA images when the asset needs to move cleanly between tools or layers.

The important idea is controllability: image generation becomes more useful when the workflow supports precise revisions, not only first-pass exploration.

Learn more about Qwen Image 2.1.

From Written Ideas to Cinematic Motion with Opus55video

WEI LIU — Thu, 01 Oct 2026 06:17:34 +0000

A written idea can already contain the beginning of a shot: a subject, a setting, an action, and a mood. Opus55video turns that starting point into a short video workflow.

Start with the scene

Opus55video is an online workspace for creating videos from text. Describe the scene you want to see, then add the details that matter to the direction: who or what is in the frame, where the action happens, and how the moment should feel.

Direct the format and timing

Once the idea is clear, choose the format and duration that fit the destination. Landscape, portrait, and square framing can serve different contexts, from a wide concept presentation to a mobile-first story. A focused duration also helps keep the shot purposeful.

Refine the direction

The creative direction stays with you. Try a different setting, camera movement, lighting choice, or pace, then compare the result with the original idea. The goal is a simple path from written prompt to a visual reference that can be watched, discussed, and refined.

Opus55video is built for creative exploration, storyboard references, social content ideas, and conversations about visual direction. Start with one clear scene and let the next frame follow from the words.

hotel-lobby: A Warm, Two-Person AI Video Concept

WEI LIU — Thu, 01 Oct 2026 05:29:45 +0000

hotel-lobby is a freemium AI video concept set in a warm orange recording booth. It features two performers sharing one microphone in a natural call-and-response scene.

Its tagline is “Two Faces. One Mic. Endless Chemistry.” Learn more at hotel-lobby.

I Built an AI Video Tool. The API Was the Easy Part

WEI LIU — Fri, 18 Sep 2026 12:28:18 +0000

A few weeks ago, I started building a small AI video project.

My assumption was pretty simple: if video generation becomes fast enough, there should be interesting products you can build around it.

So I connected a video generation API, built a basic interface, and got the first version working.

Prompt in. Video out.

At first, that felt like progress.

Then I actually tried using it as a real workflow.

And I realized the API was probably the easiest part of the whole product.

Generating one video is easy

Most AI video demos follow the same pattern:

Prompt → Generate → Wait → Video

That's fine when you're testing a model.

But imagine you want to create short-form content regularly.

Now the workflow becomes something like:

Idea → Prompt → Generate → Wait → Review → Download → Organize → Upload → Publish

Then you do it again.

And again.

If you want to create 10 videos, you don't really have a video generation problem anymore.

You have a workflow problem.

That distinction wasn't obvious to me when I started building.

Waiting becomes a UX problem

Another thing I underestimated was generation time.

With many normal web applications, a user clicks something and expects an almost immediate response.

AI video doesn't always work like that.

A generation can take long enough that keeping someone on a loading screen feels wrong.

My first version basically did this:

click → wait → result

It worked technically, but it didn't feel like a good product.

I started thinking about each generation as a job instead.

submit → queue → process → result

That small change affects a lot of things.

Now multiple generations can be submitted without forcing the user to wait for each one individually.

It also makes batch generation much easier to think about.

The next problem appears after the video is generated

Once I had generation working, another question appeared:

What happens after the video exists?

If the user still needs to download every video, rename files, open another platform, upload them manually, and repeat the process, I've only automated one part of the job.

This is where my original idea started changing.

Instead of building another page with a prompt box and a Generate button, I became more interested in the entire pipeline:

Create → Queue → Review → Publish

I'm currently testing this idea through a small project called ReelsLaunch.

I'm using it to experiment with short AI video generation, queues, batch workflows, and reducing the manual steps between creating a video and publishing it.

It's still early, and I'm not sure yet which part of that workflow will turn out to be the most important.

But building it has already changed how I think about AI products.

Three things I've learned so far

Don't build around the model

It's tempting to see a new model and immediately think:

I should build a website for this model.

I did something close to that at the beginning.

But models change very quickly.

Today's impressive model can become a standard feature somewhere else a few months later.

The model is infrastructure.

The more interesting question is what workflow becomes possible because that infrastructure now exists.

Long-running AI tasks need different UX

A traditional request/response interface isn't always a good fit for video generation.

Users shouldn't have to babysit a generation.

For longer AI tasks, queues, background processing, status tracking, retries, and notifications can become just as important as the generation itself.

This sounds obvious in hindsight.

It wasn't obvious to me when I started.

Look at what happens before and after the AI

This is probably the biggest lesson for me.

When we build AI products, it's easy to focus almost entirely on the AI step.

But users usually have a larger job they are trying to complete.

For short-form video, generation is only one piece.

There is planning before it.

There is reviewing, organizing, and publishing after it.

The opportunity may not be in making the Generate button better.

It may be in removing the need to click five other buttons afterward.

I'm still figuring this out

I don't know yet whether the best version of this product is primarily a video generator, a publishing tool, or something closer to an automated content pipeline.

That's part of why I'm building it.

But one question has become much more useful to me than asking which AI model I should use:

What repetitive work can disappear because this model exists?

If you're building something on top of an AI API, I'd be interested to hear how you think about this.

Are you building around the model itself, or around the workflow that happens before and after it?

The True Cost of Cinematic AI Video: Navigating MiniMax H3 Max API Pricing

WEI LIU — Mon, 07 Sep 2026 03:08:14 +0000

ntroduction
The AI video landscape is shifting rapidly. With the recent release of MiniMax H3 Max on platforms like fal.ai, we are finally seeing "cinematic" quality that rivals Hollywood storyboards. However, as these models move from research labs to production environments, a new challenge emerges: Cost Transparency.
Unlike flat-rate subscriptions, high-end models use a "per-second" billing model that can lead to significant "sticker shock" if not planned correctly. I’ve been tracking these shifts and built h3price.com to help creators navigate this new economy.
Why the "Per-Second" Model Changes Everything
Most creators are used to monthly credits, but the H3 Max API is billed based on output duration and resolution. Here’s the current breakdown for regular rates:

480p Resolution: $0.05 per second.
768p Resolution: $0.08 per second.

At first glance, $0.08 sounds small. But for a professional 60-second sequence at 768p, you are looking at $4.80 per minute. When you factor in the "trial and error" inherent in AI generation, costs can escalate quickly.
Optimization is the New Skill
To help the community manage these workflows, I developed a few tools on h3price.com:

H3 Cost Calculator: Essential for budgeting production runs. You can estimate total spend by resolution and clip volume before hitting "Generate."
Cinematic Prompt Generator: Since every second costs money, your first prompt needs to be your best. I’ve built a generator specifically tuned for H3 Max’s native understanding of camera movement and lighting.
Promo Tracking: We monitor launch offers (like the current 50% discount for the first 14 days) to ensure you're timing your heavy production runs during discount windows.

The Verdict
MiniMax H3 Max is a game-changer for visual storytelling, but it requires a "Developer Mindset" toward budgeting. Whether you are a solo creator or a creative agency, understanding the granular cost of your frames is the only way to scale.
Check out the tools and pricing guide here: https://www.h3price.com
I’d love to hear from other devs—how are you handling the shift to per-second API billing in your video workflows?