UGC Wingman Release Notes

Released September 8, 2026

What is UGC Wingman?

UGC Wingman is an AI-powered copilot for beginner UGC (User Generated Content) creators. You give it a product or brand, answer a few questions about your style, and it walks you through an 8-step workflow that produces a hook, a structured script, and a shot-by-shot filming plan.

The goal is not to write your video for you. It's to remove the paralysis that stops most beginners before they ever start.

What's new since v0.6

A complete visual redesign

• A step navigation strip across the top, so you can always see which of the 8 steps you're on and what's still ahead
• A card-based layout that groups each step's content instead of running everything together in one long column
• The "Dusk Blend" palette, a darker, calmer look that replaced the original default-blue theme
• Locked-but-visible steps, so upcoming steps are greyed out rather than hidden and you can see where the flow is going

A clearer starting point

• New landing page hero explaining who the tool is for and what you get, before you commit to anything
• One obvious way to start. The page previously offered a few competing "start here" prompts, which was more confusing than helpful. Now there's a single "Start with step one" button.

Less friction getting going

• Rewritten Creator Voice quiz, shorter and easier to answer honestly
• Content type and video length moved earlier and simplified, so the tool knows what it's writing for sooner

Under the hood

• Free-text input security hardening covering the fields where your typing reaches the AI
• A request queue so several people using the app at once don't knock each other's requests over
• Codebase decomposition, splitting one large server file into focused modules
• Text provider switched to Google Gemini, explained in Known limitations below

Known limitations

• The first load can be slow. The app runs on a free hosting tier that puts the server to sleep when nobody's using it. If you're the first person to open it in a while, expect up to 30 seconds of loading before anything appears. It's fast after that.
• Generation steps take about 20 seconds each. That's normal, not a hang. The brand dossier is the longest one.
• Requests are handled one at a time. If several people are using the app at the same moment, yours waits its turn. During a busy stretch a request can give up after 90 seconds. If that happens, just try again.
• Text generation is running on a temporary provider. The original provider ran out of free quota mid-testing, so the app was switched to Google Gemini's free tier as a stopgap. Output quality may differ slightly from earlier versions.
• Mobile works but isn't fully optimized. The app was built and tested mainly on desktop. See Known defects below for the two specific rough edges.
• AI scoring trends high. The built-in eval scoring uses the same AI model to grade its own output, a known bias called "self-grading." Scores near 100% should be read with that caveat. This is admin-only and doesn't affect your session.
• Session data resets on server restart. In-memory data clears when the server restarts or redeploys. Expected for the beta.
• Deep Dive research mode is admin-only. The grounded web-search version of the brand dossier is reserved for admin accounts during beta to manage API costs. Quick Research is what you'll see, and it's still solid.

Known defects

Both are on a phone, both are known, both are being worked on:

• The step counter repeats the step number. On a narrow screen the readout renders as "1 of 8 · 1-Check-in" instead of "1 of 8 · Check-in". Cosmetic only, nothing is broken.
• The shot list table is stretched vertically on mobile. Step 7's table is hard to read on a phone. Rotating to landscape is a workaround until it's properly fixed.

What's coming next

1. Mobile optimization, starting with the two defects above
2. Handling more people at once, so concurrent users don't queue behind each other
3. Portfolio page and case study, telling the story of how this was built

Build stats (for the curious)

Security audit completed before beta, 11 findings reviewed, all medium-severity issues resolved. Both AI providers run on free tiers during beta.

Released July 8, 2026

The first beta release

v0.6 established the core 8-step workflow the app still runs on today:

1. Entry confidence check, a quick emoji rating of how ready you feel before you start
2. Creator voice profile, questions about your storytelling style, tone, and energy so the tool can write more like you (fully skippable)
3. Product input, the product name, a brief description, and what your target audience is currently doing instead
4. Brand dossier, an AI-researched profile covering target customer, brand tone, gap analysis, and a content angle for you to choose from
5. Hooks, 3 opening lines tailored to your chosen angle and voice profile
6. Script, a full structured script in the Hook / Problem / Experience / Result / CTA format, tuned to your target video length
7. Shot list + PDF export, a shot-by-shot filming plan with shot type, camera direction, and timing
8. Confidence layer + exit survey, coaching on whether your script is ready to film

It also introduced target video duration (15s, 30s, or 60s), the live word count bar, the sources panel on every brand dossier, and the feedback form.