Design & Creative
← Home

Design & Creative

AI tools for Canva and CapCut: templates, copy, palettes and visual ideas.

4 apps

TwitchApp
Nolan ReyesNolan Reyes

Twitch — Channel kit: avatar, banner, emotes, badges and panels

Channel identity kit for Twitch. Describe your channel once —name, game or theme, aesthetic, tone and mascot— and the app returns the pieces at the exact size Twitch accepts: 256x256 avatar, 1200x480 profile banner, 1920x1080 offline screen, the three stream screens, webcam frame with a transparent center, 320 px panel headers, 1280x720 VOD thumbnail and 1080x1920 vertical pieces. Emotes come in 112, 56 and 28 px with clean alpha and a real-size preview in chat; subscriber and Bits badges come in 72, 36 and 18 px and always under 25,600 bytes. It also writes the channel texts: bio, chat rules, panels in valid markdown, stream titles, tags, schedule, bot commands with their add-command line, alert texts, moderation replies and AutoMod sheet. It does not connect to your Twitch account: no OAuth, no permissions and no touching your channel.

Google PhotosApp

Google Photos: compress photos, remove location and repair Takeout dates

Browser mini app that prepares your photos before uploading them to Google Photos and fixes them after downloading them. On the same batch: compression with a report of MB saved and a target file size per photo, GPS purge keeping the capture date (verified by re-reading the file), iPhone HEIC conversion, crops for social media and for printing in centimetres at 300 dpi, 35x45 mm ID photo, AI content-based renaming with its mapping CSV, watermark-free PDF album and the Google Takeout date repairer. It does not connect to your Google account: you upload the files by hand and they are deleted shortly after you download them.

CanvaApp
Nolan ReyesNolan Reyes

Canva — Palette with measured contrast, exact sizes and copy that fits

Workshop for assets and numbers for people who design their own Canva posts without being designers. It is not an editor, not an alternative, and does not connect to your account: it makes what the editor does not give you and you have to add yourself. The accessible Brand Kit starts from a hex color and builds the entire palette in OKLCH with 5 harmonies (analogous, complementary, triadic, split-complementary and monochromatic), 10-step ramps from 50 to 900 for the 4 color roles —40 swatches with always-descending lightness— and then cross-references every text with every background in a 30-pair matrix with its WCAG 2.1 contrast ratio and verdict by use: AAA, AA, headlines and UI only, or do not use, with thresholds of 4.5:1 for normal text and 3:1 for large text (from 24 px) and UI elements. The one-tap fix is calculated on the server by moving only the color's lightness, never the hue, and translates it to the nearest step in the ramp that does comply. It includes simulation of the 4 types of color blindness and 10 font pairings with 20 open-license families, each with its cited license (SIL OFL 1.1 or Apache 2.0) and official link, plus 4 modular scales. Upload Prep Studio prepares the image for any of the 15 presets across 9 destinations, with safe zones for TikTok, Reels, Shorts, the Instagram grid and the LinkedIn cover, gesture-based framing, 4 ways to fill the missing canvas (crop, blurred background, mirrored edges or flat color) and target file size via bisection in up to 8 passes re-reading the file binary: the reported weight is the actual weight, and if the target is impossible it delivers the lightest result achieved and says so. Output in JPG or PNG, alpha preserved, batch from one image up to 8 sizes and ZIP with the sizes noted. There is never a watermark. The copy meter is the only thing of its kind: you enter your box width in pixels and the font size, and the server rasterizes the text and measures the actual ink to tell you how many characters fit, counted by graphemes, so a composite emoji counts as one character; it is free and unlimited. The headline generator is the only thing that uses artificial intelligence and is tightly constrained: it asks for 6 options, measures all of them and discards the ones that overflow before they reach your screen. The deterministic features —palette, contrast, color blindness, fonts, scale, size templates and meter— work without registration or cost, and guests can also prepare 3 files per day and do 1 copy run. Size catalogs are born marked UNVERIFIED with their version and date, and typographic measurement uses an equivalent metric installed on the server, not the exact family: the app says so on screen instead of hiding it. Unofficial tool: not affiliated with or sponsored by Canva Pty Ltd; Canva is a trademark of Canva Pty Ltd and is named here only to describe which editor these assets are used with.

CapCutApp
Nolan ReyesNolan Reyes

CapCut — .srt subtitles without typing, shot-by-shot edit plan, PNG titles and export dimensions

AI tools for people who already edit vertical video in CapCut and don't know what to cut. Paste your script and download the .srt subtitle file with timings distributed by reading speed: 5 segmentation presets (vertical, one line, karaoke, balanced and comfortable) and 6 manual parameters — characters per line, lines per title, characters per second, minimum and maximum title duration and gap between titles — with instant preview and download in .srt, .vtt or .txt. If you already have the audio or video recorded, it transcribes it with word-by-word timestamps (up to 15 minutes and 20 MB per file, checked before upload) and you can re-segment as many times as you want without paying for transcription again. The edit plan returns the hook, the shots with their duration and in/out seconds, what text appears on screen, at which second and for how long, the transition and effect of each cut, where the b-roll goes and where silence goes, the call to action and the totals — seconds, cuts and cuts per minute — and downloads as printable text. For the screen: 5 pieces (cover, lower third, quote, closing and subtitle plate) in 6 styles, in vertical PNG 1080×1920 and horizontal 1920×1080, with transparent background only when the crop is clean and all text drawn by the server, never by the image model. And the data people forget when exporting: 5 canvases, 5 export setting sets with fps, codec and bitrate, 5 safe zones per destination and a PNG margin template with rule-of-thirds grid to place text where the interface doesn't cover it. We do not create editable project files or pre-built templates, do not render video, do not access your account and do not download content from third-party platforms: we deliver files you import manually. Independent tool, not related to or affiliated with ByteDance or CapCut.