04 · Creators
YouTubers. Streamers. Podcasters. Course creators. Newsletter writers. Indie game devs. Every "channel of one" who needs to ship a cover, banner, thumbnail, or post — daily — without flipping between five tools and three subscriptions.
The job
Tiramisu treats the creator workflow as the first-class case — not the "designer working on a hero brief" case Photoshop was built for in 1990. Every feature below exists because a creator we talked to said "I waste an hour on this every week."
Direct social integration
No more export → upload → re-crop → upload → re-tag. Connect once via OAuth, then publish, schedule, and pull analytics back in.
YouTube thumbnails, IG posts, TikTok covers, X images, LinkedIn posts, Discord embeds — direct OAuth uploads with the right metadata each platform wants.
One canvas → optimized variants for every platform. Not just resize — intelligent reflow: a 1920×1080 horizontal thumbnail relays out as a 1080×1920 TikTok cover, subject re-anchored, headline repositioned, safe-area respected. Different aspect ratios, same intent.
Pick a time, walk away. The app holds the asset, posts on schedule. No third-party Buffer / Hootsuite / Later subscription needed.
CTR, view count, retention, watch-time pulled back into Tiramisu. Click any past asset to see how it performed — learn from your own catalogue.
Run two YouTube channels and three IG accounts? Connect them all. Each project tags which account it's for; each export uses that account's brand kit.
Connect once, Tiramisu pulls your past posts, builds a "your style" profile, suggests templates calibrated to what's worked for you.
Templates & sharing
Templates aren't asset dumps — they're remixable, attributable, often profitable. We treat the template ecosystem as a creator economy, not a clipart bin.
Hundreds of pro templates organized by niche — gaming, vlog, tech review, beauty, food, fitness, business, education, podcast, dev. Every category, multiple aesthetics.
Anyone publishes templates. Community rates, remixes, forks. Trending feed surfaces what's actually working this week.
Fork any template, your version inherits a credit chain. The original creator can see their work's lineage and is paid (if priced).
Templates · LoRAs · brushes · brand kits · plugin actions. List free or paid — your call. Free, open-source app. No subscription, no Pro tier, ever.
What thumbnails are top YouTubers in your category posting? Daily curated feed, no scraping required.
One template, parameterized. Generate consistent thumbnails for every episode of a season; same frame, swap subject + title.
Guides & safe areas
No more discovering after upload that your title got chopped, your face got covered by a play button, or your CTA hit the swipe-up bar. Every platform has its own occlusion rules — we have them all.
A/B testing & test pages
Stop guessing. Generate variants, predict performance, compare side-by-side, push the winner. Then learn from real data.
"Generate 8 versions of this thumbnail with different hooks." Side-by-side gallery; pick winners; ship.
"This variant is 22% more likely to outperform your average" — based on your past data + niche-trained models. Calibrated, not vibes.
Where will eyes go first? Visual saliency model overlays heatmap on each variant. Catch "the title is invisible" before posting.
Render each variant at 1× / 0.5× / mobile-grid / dark-mode / on-feed-mockup. See it the way the audience will see it.
YouTube's official thumbnail-testing API: push 3 variants, let YT rotate, pull winner back. Auto-promote.
"Top 50 thumbnails in your category last 30 days, sorted by CTR." Learn what's working before you design.
AI review · rate · suggest
Most creators don't have a team. The AI plays the role of the colleague who looks over your shoulder and says "the title's too small, the face is in shadow, this won't pop in the feed."
"82/100 — strong contrast and emotion, but the title competes with the subject for attention. Try moving title up-left."
"For tech YouTube, the audience expects a hands-on shot. Yours has no product visible."
Paste your title; AI checks the thumbnail delivers on the promise. Curiosity gap analyzed without clickbait penalty.
Preview at thumbnail size: is the text legible? Is contrast sufficient? Any element so small it disappears? Auto-flagged.
Contrast ratios, color blindness simulator, screen-reader preview. Make sure your audience can read it.
One click → 5 alternative compositions, ranked. Each with a one-line rationale. Pick what works; remix what's close.
Style transfer · style stealing
Found a creator whose vibe you love? Drop a reference image. Tiramisu extracts the style fingerprint — palette, lighting, composition, typography — and applies it to your work. Original assets stay original.
Drop a reference, transfer its style. Color, light, mood — not pixels.
Analyze your last 50 thumbnails, save as a style preset. Now every new thumbnail can match your channel automatically.
Paste 5 of their thumbnails. Get a style preset that captures the formula without copying anything specific.
Pre-built presets: Tech-Review, Gaming-Highlight, Vlog-Lifestyle, Cooking-Closeup, Fitness-Bold, Edu-Authority. One-click vibe.
Live linting: "this design is drifting from your channel's style." Toggle off if you're intentionally shaking things up.
Reverse the flow: paste a title, get 5 style suggestions that historically perform with that kind of headline.
Brand packages
A brand kit isn't a Figma file you forget about. It's a live constraint on every export — fonts, colors, logos, voice, watermarks, signature framing.
Logo, fonts, colors, accent colors, signature framing, recurring elements. Run multiple channels? Multiple kits.
"Energetic but not cringe." "Authoritative, dry humor." Stored as a profile the AI uses when suggesting copy.
One-click watermark with your logo, channel handle, episode number — placement and opacity tuned per platform.
Intro stings, outro music, transition SFX. Exportable to FCP / DaVinci / OBS so video creators have a single brand system.
Wrong font? Color outside palette? Logo wrong size? Tiramisu warns before export. Fix or override per project.
One click → PDF brand guidelines doc. Hand to a freelancer; lock in consistency without writing it from scratch.
Workflow features
Drop a CSV of 10 video titles. Get 10 thumbnails out, all matching your channel style. Tweak the ones that need it; ship the rest.
Paste a title; get 5 thumbnail concepts back, each with a different hook. Pick one, refine.
Reverse: paste your thumbnail, get 5 title suggestions. Catches "the visual says X but the title says Y" mismatches.
Drop a webcam recording. Tiramisu cuts out your face on transparent BG, builds a "face library" you reuse across thumbnails.
Save your "shocked face," "confused face," "stoked face" once. Reuse forever. One click to swap expression on a thumbnail.
Lock the frame, vary only the subject + title. Twelve episodes of a season look like a series, not twelve random images.
Asset libraries
Discoverability & growth
Daily curated feed of top-performing posts in your category. Save, fork, remix.
Pin 5 competing channels. Tiramisu auto-pulls their new thumbnails as they post — see the meta shift in real time.
AI-curated based on what's working in your niche AND your historical wins. Not a generic Pinterest dump.
"Your last 20 thumbnails are 78% blue. Variety drives CTR — try a warm palette next."
Trending search queries in your niche → titled-and-thumbnailed concepts you could film this week.
Visual weekly grid: what's posted, what's scheduled, what's still untouched. Drag-drop to reschedule.
Collaboration
Editor working with you on a launch? Real-time presence, cursors, comments. Exit-and-resume.
Public view-only link with comments. No account required for the reviewer. Comments come back into the editor as pins.
"Final," "Final v2," "Final FOR REAL." Plus auto-saved every save. Revert any layer to any past state.
One file → all assets, fonts, plugins needed to open the project somewhere else. No "missing font" panic.
Cross-app integration
A day in the life
Tuesday upload day. Title pasted, 8 thumbnails generated, saliency map flags two winners. Native YT test pushes 3 variants live; in 48h the winner auto-promotes.
Monthly stream refresh. Brand kit drives every asset; export pushes panels to Twitch, alerts to OBS, banner across socials. No alt-tabbing.
New season. Drop CSV of 12 episode titles + guest names. Out: 12 covers ready for Apple Podcasts (3000²) and Spotify (1400²), all matching the season's visual identity.
Course launch. Sales page hero, 12 module covers, IG carousel, X post, LinkedIn. All from one brand kit; auto-resized; scheduled to go live as launch unfolds.
Friday morning. Title in Tiramisu, two header concepts in 30 seconds, picked one, pushed to Substack draft. Inbox-ready.
Pre-launch week. Steam capsule (1920×620 + 374×448 + 467×181), Twitter announcement, Discord embed, devlog header — one project, six exports, every dimension correct.
The bet
The art director ships one big brief a week. The creator ships five things a day across four platforms. That's not the same workflow with different sliders — it's a completely different product. Photoshop has spent 30 years optimizing for the wrong user; Canva spent 10 doing it on the web; Photoroom focused on e-commerce. Nobody is building the native, free, AI-first, social-deeply-integrated tool the modern creator economy needs.
So we are.