The latest articles on DEV Community by jacex (@jacex). https://dev.to/jacex https://media2.dev.to/dynamic/image/width=90,height=90,fit=cover,gravity=auto,format=auto/https:%2F%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Fuser%2Fprofile_image%2F1151181%2F707617ba-079a-458d-ac79-2ec229a8ca40.jpeg https://dev.to/jacex en jacex Thu, 24 Sep 2026 05:33:49 +0000 https://dev.to/jacex/building-an-ai-creative-toolkit-image-video-upscale-and-motion-in-one-place-41b6 https://dev.to/jacex/building-an-ai-creative-toolkit-image-video-upscale-and-motion-in-one-place-41b6 <p>If you ship product marketing, social creatives, or short-form video, you probably bounce between five tabs: one model for text-to-image, another for video, another for upscaling, maybe a separate tool for watermark cleanup.</p> <p>That context-switching is the real cost — not the credits.</p> <h2> What a creative toolkit should cover </h2> <p>A practical AI creative stack usually needs:</p> <ul> <li> <strong>Text-to-image</strong> for campaign stills and concept art</li> <li> <strong>Text/image-to-video</strong> for short clips</li> <li> <strong>Motion control</strong> when you want a character to follow a reference move</li> <li> <strong>Upscaling</strong> so outputs survive social crop and retina screens</li> <li> <strong>Watermark removal</strong> for clean deliverables (when you own the rights)</li> </ul> <p>Models move fast (Nano Banana, Seedream, GPT Image, Kling 3.0, and friends). The useful product layer is less "one magic model" and more "one workflow that keeps swapping models underneath."</p> <h2> How I use it day to day </h2> <ol> <li>Draft a still with a strong prompt + reference style</li> <li>Push the still into video when the concept lands</li> <li>Upscale the frame or clip before publishing</li> <li>Keep experiments cheap — free daily credits help here</li> </ol> <p>I built <strong><a href="https://www.tyndallai.com" rel="noopener noreferrer">Tyndall AI</a></strong> around that pattern: an AI-native toolkit for image and video generation in one place (image gen, video gen, motion control, upscaler, watermark remover), with 10 free credits every day.</p> <h2> Honest caveats </h2> <ul> <li>Prompt quality still matters more than the brand logo on the button</li> <li>Video is slower and more expensive than stills — iterate on images first</li> <li>Always respect copyright and platform rules on watermarks and likeness</li> </ul> <p>If you are assembling your own creative pipeline, start by listing the five jobs you actually do weekly, then pick a toolkit that covers those jobs without five separate accounts.</p> <p>Curious what your current stack looks like — Midjourney + Runway? Local ComfyUI? Something else?</p> ai webdev jacex Mon, 10 Aug 2026 08:04:06 +0000 https://dev.to/jacex/i-started-tyndall-ai-4hfk https://dev.to/jacex/i-started-tyndall-ai-4hfk <p>I'm building <a href="https://tyndallai.com" rel="noopener noreferrer">Tyndall AI</a> on my own.</p> <p>The name comes from the Tyndall effect. My chemistry teacher explained it once, and it stuck with me, so I used it.</p> <p>I haven't decided yet whether to focus on image or video. Both are interesting, so for now I'm shipping both.</p> <p>There's one thing I really want to build: a tool where you can line up different models and compare them in one click. I couldn't find a good one. A lot of comparisons are just numbers. You barely see the actual images or videos. So you still don't know how your own prompt will look.</p> <p>That's what I'm experimenting with first. If you have thoughts, tell me.</p> <p><a href="https://tyndallai.com" rel="noopener noreferrer">https://tyndallai.com</a></p> <p>PS. I'm not great at writing, but I'll try to explain things clearly!</p> ai nanobanana seedream seedance