โ† All Guides
Video AI 8 min read ยท Free guide

Create a Viral Selfie UGC AI Video

A 4-step workflow to generate a hyper-realistic mirror selfie UGC video โ€” base image, skin detail pass, face close-up, then the final video with Veo 3.1. All prompts included.

๐Ÿ› ๏ธ Tool: Google Flow  + NanoBanana2 + Veo 3.1

How It Works

Open Google Flow and select NanoBanana2. Run the 4 steps in order โ€” each step uses the previous image as a reference. The final step generates the video using Veo 3.1 Fast with both the skin-detail image and face close-up as references.

1 Create the Base Image (NanoBanana2)

Start a new generation with the full prompt below.

IMAGE TYPE: Editorial Portrait PROMPT: High-impact cinematic depiction of Nicole, featuring a youthful face with high cheekbones, full lips, and groomed brows, long sleek brunette hair parted in the middle, and a neutral, confident expression. Dressed in a cozy grey off-the-shoulder oversized sweatshirt and matching grey ribbed knit shorts. Positioned in a luxurious minimalist bathroom with white marble walls and a large mirror. The scene conveys a relaxed, chic aesthetic through soft indoor lighting, with natural highlights and gentle shadows reflecting off the marble surfaces. Captured using a 35mm lens, mirror selfie angle, waist-up framing, shallow depth of field. Ultra-detailed textures, realistic skin tones, professional cinematic color grading, sharp focus, HDR, ultra-high resolution, 4K, masterpiece quality. STYLE: cinematic, ultra-detailed, 4K, dramatic lighting, shallow depth of field, HDR, professional, film still NEGATIVE PROMPT: low quality, blurry, flat lighting, distorted face, bad anatomy, extra limbs, oversharpening, text, watermark, logo

2 Add Skin Detail Pass (NanoBanana2)

Add the first image as a reference, then use this prompt to enhance skin realism.

Primary edit: Enhance skin texture to reveal natural, human-grade surface detail โ€” visible pore structure across nose bridge and cheeks, subtle capillary flush around nostrils and under-eye area, natural sebaceous micro-sheen on the T-zone, and fine vellus hair that lies pressed completely flat against the skin surface, barely perceptible, only catching light as a faint luminous edge at grazing angles. Vellus hair length MUST NOT exceed 1โ€“2mm in appearance. NO protruding, elongated, or standing facial fuzz under any condition. Preservation clause: Keep all existing facial structure, bone landmarks, expression, eye catch-lights, composition, background, and color grading EXACTLY as they appear. Do not alter skin tone, undertone, or overall exposure. Realism anchors: Subsurface scattering visible beneath cheekbone skin, natural pore shadow depth, slight desaturation at pore edges, micro-texture variation between oily and dry zones. Camera technical / negative constraints: Rendered as Sony A7R IV, 85mm f/2.8 โ€” avoid: smoothed skin, plastic sheen, AI anomalies, exaggerated pores, overly long facial hair, symmetrical artifacting, overexposed highlights. Context anchor: Vanity Fair editorial portrait standard โ€” authentically human, never retouched, never digitally perfected.

3 Generate Face Close-Up (NanoBanana2)

Add the second image as a reference and generate a detailed face close-up.

4K close-up face portrait, natural skin texture, visible pores, subsurface scattering, individual hair strands, sharp eyes with catchlight, soft diffused lighting, photorealistic, ultra-detailed, face only, tighter crop, same camera angle, same pose, do not alter subject identity or composition.

4 Generate the Video (Veo 3.1 Fast)

Add both the Step 2 and Step 3 images as references, switch the model to Veo 3.1 Fast, then use this prompt.

Handheld shot pointed at a large mirror, 9:16 vertical, slight natural camera shake. She stands facing the mirror holding her smartphone aimed at her reflection, so the camera sees her through the mirror. Her face and phone are visible in the reflection. Soft natural light from the side, simple bedroom or bathroom setting, slightly blurred background. She looks directly at the camera lens visible in the mirror from the first frame. Her expression is composed and still โ€” almost daring the viewer to find something wrong. No smile, no warmup. Just steady eye contact through the reflection. At the end a slow quiet smile forms. Mirror frame visible on edges of frame. Audio: She says calmly: "This is probably gonna catch you off guard." Pause. Then: "but I'm not real." Pause. Then slow and deliberate: "here is how it was done." Confident female voice, unhurried pacing, clear pause between every sentence, slight room reverb, no music. Aesthetic: authentic UGC selfie shot in mirror, lo-fi vlog feel. Soft natural side lighting, no ring light. Warm neutral color grade, slightly desaturated. Slight lens reflection visible in mirror. No cinematic grading, no filters, no text overlays, realistic handheld shake.

Tips for Best Results

Use the Step 2 output (not Step 1) as the reference for Step 3. For the video in Step 4, add both Step 2 and Step 3 images as references for the most consistent face result.

Want more free AI guides?

Follow on Instagram for new tips every week.