How to Make a Gemini Omni UGC Video in Claude
Artem2 min readI made this in Claude, with Everygen connected. With all my coffee mistakes, this ad cost me about 250 credits. If you follow the workflow properly, it comes out closer to 165.
01Before the chat
Connect Everygen
The connector, then the first message.
You need a reference video, a script, and photos of the product. If you don't have the product, references are enough. Claude generates the avatar, the props, and one starting frame per shot, then the clips.
Everygen has to be connected before that. On the Everygen connect page I copy the connector URL. In Claude that's Customize, then Connectors, then Add custom connector. I name it Everygen and paste the URL.
02The frame
The reference
One frame from the TikTok, at 3.3 seconds.
I sent a TikTok and asked for a Gemini Omni UGC video. Claude loads the realistic UGC Omni workflow and picks one frame. The frame is the room, the light, and the camera. The person changes.
Let's make a gemini omni UGC video using Everygen https://www.tiktok.com/@heyimtran/video/7565944458433318174
What I sent.

I'll load the Everygen realistic UGC Omni workflow and start by analyzing the TikTok reference. I found a reference frame at 3.3 s. I picked it because: Face: she's looking straight at the camera, mid-word, with a relaxed expression. Lighting: soft window light with no blown-out patches on her skin. Hands: they're out of frame, apart from a fingertip at the bottom edge, so there's no motion blur to fix. The avatar edit will draw clean hands. Background: the lived-in living room has depth (plant, sofa, archway) without being cluttered. There's room around her chest for products. Text: the only caption is "+7.5M impressions" above her head, and the avatar edit will remove it. It's from the opening talking-to-camera part of the video. Most of the rest is B-roll of the pans or has a picture-in-picture or text panel over her chest. Is this frame OK, or do you want to upload your own? Once you approve it, I'll make two clearly different avatars from it for you to choose between. That costs about 6 credits.
Claude's reply.
The TikTok is this video.
03The person
The avatar
Two different people in the same room. I kept one.
I approved that frame.
approve
What I sent.
Here are both avatars. Each is clearly a different person from the original creator: face shape, nose, eyes, skin tone, hair colour and texture, and apparent age all changed. Avatar A: late twenties, copper-red waves, freckles Personal details: tortoiseshell glasses, a thin gold chain with a tiny pendant, and a ribbed oatmeal knit top. Reason for clean audio: one white AirPod in her left ear. Avatar B: mid thirties, black curly bob, deep brown skin Personal details: large gold hoops, and a sage linen shirt open over a white tank. Reason for clean audio: a black clip-on mic on her collar. Both kept the living room, the camera angle and the soft window light. All on-screen text is gone, and so is the stray fingertip at the bottom of the frame. One thing isn't right yet: both have a neutral, closed-mouth expression, while the original was mid-word. I'll fix that in the character sheet and the starting frames by making her smile or talk, so it costs nothing extra now. If you'd rather fix it on the avatar itself, I can regenerate the one you pick for about 3 credits. Which one do you want, A or B? Once you pick, I'll make her character sheet: front, three-quarter, profile, smiling and mid-sentence views.
Claude's reply.
Avatar A is only in that reply. The file I kept is Avatar B.
let's go with this one
What I sent.

Going with Avatar B. Making her character sheet now. Her character sheet is ready: front, three-quarter, profile, smiling and mid-sentence views. Her face, curls, sage linen shirt, gold hoops and clip-on mic all stay the same in every panel. The smiling and mid-sentence panels also show she animates well, so the flat expression on the avatar won't carry into the video. Do you have a script ready?
Claude's reply.
The character sheet from that reply is not in the files from this chat.
04The read
The script
About 9 shots, and roughly 30 seconds.
Then I pasted the script.
yep here it is: I just made a peanut butter cup protein coffee with 15 grams of protein. You guys know I'm a religious drinker of these Everygen protein coffees, and I freaking love that they keep coming out with these new delicious flavors. Peanut butter chocolate, are you kidding me? Getting 20 to 30 grams of protein right in the morning keeps you satiated, so you're not snacking. I like to mix it with a protein milk, which makes it super creamy. This has 15 grams of protein, zero sugar, and real coffee in it. So peanut buttery. If you like peanut butter cups, you will absolutely love this.
What I sent.
I've added a few natural fillers ("okay, so", "like", "Mmm", "I mean") and written the numbers as words so she says them cleanly. The words in CAPS are the ones she should stress; CAPS is how she knows where to put the emphasis. Okay, so I just made a peanut butter cup protein coffee with FIFTEEN grams of protein. You guys know I'm a RELIGIOUS drinker of these Everygen protein coffees, and I freaking LOVE that they keep coming out with these new delicious flavors. Peanut butter chocolate… are you KIDDING me? Getting like twenty to THIRTY grams of protein right in the morning keeps you satiated, so you're not snacking. I like to mix it with a protein milk, which makes it SUPER creamy. This has fifteen grams of protein, ZERO sugar, and REAL coffee in it. Mmm, so peanut buttery. I mean, if you like peanut butter cups, you will ABSOLUTELY love this. That comes to about 9 shots and roughly 30 seconds. Props the script needs: The Everygen peanut butter cup protein coffee packaging. This is the main product (pouch, tub or bottle, whatever it really comes in). The finished drink: a glass of the peanut butter cup protein coffee. A protein milk carton. If you don't have one, I'll make it Everygen-branded or unbranded. Optional: a few peanut butter cups for the last line. Do you have real photos of any of these? I've opened the uploader below: Props you have: upload front, back and label shots. Props you don't have: upload references instead, like a packaging design you like for the product and a delicious-looking photo of the drink. Then tell me which ones to generate. I've opened the uploader. Once your photos and references are in, I'll make the product sheets and one starting frame per shot. Then I'll show you all the starting frames and the cost to animate them before anything gets generated.
Claude's reply.
05The product
The props
A pouch, a drink, and a milk carton.
I uploaded three references and told Claude to generate the pouch, the drink, and a milk carton. No peanut butter cups.



I've uploaded 3 files: CleanShot 2026-11-06 at 10.32.08@2x.png peanut butter coffee.webp pouch reference.jpg I uploaded the pouch reference, what Everygen branding looks like and the finished coffee cup. Milk carton should be a generic one but with Everygen branding. No need for a few peanut butter cups. Generate the three props now
What I sent.
All three props are generated. The text on every label is clean, with no gibberish. Pouch: pick one A, peanut-butter tan: a tan pouch with a chocolate wave at the bottom and a blue "15g PROTEIN" badge. Its icon matches your real Everygen logo (play symbol plus bar). B, Everygen blue: a blue and white pouch with a peanut-butter-and-chocolate swirl at the bottom and a white "15g PROTEIN" badge. It's more on-brand in colour, but the icon came out as a plain play triangle without the bar. If you pick B, I'll fix the icon on the product sheet. Drink: a tall glass jar of creamy iced coffee with peanut butter dripping over the rim, chocolate flecks and a metal straw. It follows your reference closely. Milk carton: a white gable-top carton with the Everygen logo, "PROTEIN MILK" in big blue letters and a blue splash. Both pouches and the carton keep the front to a few big elements, so the text should hold up once she moves them around. Which pouch, A or B? Once you pick, I'll make product sheets for the pouch and the drink, then one starting frame per shot. After that I'll show you the full grid of starting frames with the cost to animate them.
Claude's reply.
06The package
The pouch
The blue pouch, with our logo.
I picked the blue pouch and asked for our real logo on it.
let's go for this pouch, but pls change the logo to our actual one
What I sent.

Going with pouch B. I'll make the product sheets first, and fix the icon to your real play-plus-bar logo on the pouch. The starting frames for all 9 shots are in starting-frames-grid.jpg. In every frame the labels are readable, nothing covers the "15g" badge, and her face and framing stay the same. Shot 1, 5 s. Okay, so I just made a peanut butter cup protein coffee with FIFTEEN grams of protein. Glass beside her face; pushes it toward the lens, then pulls it back. Shot 2, 4 s. You guys know I'm a RELIGIOUS drinker of these Everygen protein coffees. Pouch beside her face; taps the label. Shot 3, 5 s. and I freaking LOVE that they keep coming out with these new delicious flavors. Pouch and glass; bounces them side by side. Shot 4, 4 s. Peanut butter chocolate… are you KIDDING me? Glass in both hands, wide eyes; pulls it in close. Shot 5, 6 s. Getting like twenty to THIRTY grams of protein right in the morning keeps you satiated, so you're not snacking. Pouch at her chest; taps the 15g badge. Shot 6, 5 s. I like to mix it with a protein milk, which makes it SUPER creamy. Milk carton; gives it a little shake. Shot 7, 4 s. This has fifteen grams of protein, ZERO sugar, and REAL coffee in it. Pouch beside her face (same frame as shot 2); holds it steady and taps once per claim. Shot 8, 4 s. Mmm, so peanut buttery. Straw at her lips; takes a sip, then reacts. Shot 9, 4 s. I mean, if you like peanut butter cups, you will ABSOLUTELY love this. Pouch and glass (same frame as shot 3); pushes both toward the lens. Cost to animate: about 110 credits for all 9 clips (Omni 1.1 Flash, 1080p, audio on), plus any rerolls. Shots 7 and 9 reuse the frames from shots 2 and 3. They're far enough apart in the cut that the repeat won't show, and it saved 6 credits. So far I've spent about 54 credits on avatars, props, product sheets and starting frames, including 4 redos to fix the pouch icon and the framing. Should I go ahead? I'll generate shot 1 first so you can hear her voice. Once you approve it, I'll start the other 8 together.
Claude's reply.
The first contact sheet, starting-frames-grid.jpg, is not in the files I kept. The later sheet is below, before I tell Claude to go ahead.
07The glass
The cup
A real iced latte, then go ahead.
I sent a real cup in place of the jar.
I've uploaded coffee cup.png. Let's use this cup instead
What I sent.

The new cup is in. I'm swapping it into clean versions of the starting frames for shots 1, 4 and 8 rather than editing the edits again. Shot 4 gets its wide-eyed expression back this way too. All 9 starting frames now use your iced latte glass. You can see them in starting-frames-grid-v5.jpg. Shots 1, 4 and 8: I swapped the glass into the clean original frames, so there are no edit-on-edit artefacts. Shot 4 has its wide-eyed "are you KIDDING me?" face back. In shot 8 she now holds the glass with both hands, with the glass straw at her lips. Shots 3 and 9: rebuilt from the sharp pouch frame, with the same glass and hand as shot 1. Both hands match her face. The glass: the layered espresso and milk, ice and glass straw all pick up the room's daylight and grain, so it reads as a real drink rather than a cutout. Shots 2, 5, 6 and 7: unchanged. These frames cost 12 credits. Both shot 1 takes are paid for (26 credits), so I've spent about 143 credits in total. Since you've already approved her voice, I can generate all 9 clips at once with the identical base prompt. That's about 110 credits plus any rerolls. Everygen runs about 4 at a time, so expect around 15 to 20 minutes for the full set. Should I go ahead?
Claude's reply.
The starting frames
This is the grid from that reply. The 9 clips are not in the chat.

I told Claude to generate all 9.
yes go ahead
What I sent.
What it cost
With all my coffee mistakes, this ad cost me about 250 credits. But if you follow the workflow properly, it comes out closer to 165.
I burned through about 50 credits trying to fix the cup.
So even our Basic annual plan covers it: you'll get 200 credits a month for $14.
You can get started with the Basic plan today and make your own UGC video.
The same connector is on the Everygen connect page.
Start from a reference video
You can get started with the Basic plan today and make your own UGC video.