Reference Image: Enable strict facial consistency mode… | Deepfake AI Video

Text-to-video prompt used to generate this video

Reference Image: Enable strict facial consistency mode. Prioritize the facial features from the provided reference image for all subsequent generations. Maintain the subject's identity accurately while only adapting the pose, lighting, and background. Do not alter the core facial structure. Total Duration: 10 Seconds Perspective: 1st Person POV (Moving forward alongside the character) Vibe/Tone: Romantic, intimate, teasing (Photorealistic Cinematic FMV style, natural heavy/low-pitched vocal tone, mature Onee-san presence) [00:00 - 00:03] Phase 1: The Walking Conversation Visual: 1st-person POV. A bustling Japanese summer night festival pathway lined with glowing red lanterns casting warm amber light. The camera smoothly moves forward (simulating walking side-by-side). The mature woman is walking right beside or slightly ahead of the camera view, wearing an elegant black yukata with subtle floral motifs, holding a glossy red candied apple (ringo ame). Ultra-realistic cinematic lighting and lifelike skin textures. Action: She turns her head toward the camera while keeping a steady walking pace, locking her eyes directly onto the viewer with a soft, natural smile as they walk together through the festival. Audio/SFX: Ambient festival sounds, distant chatter, footsteps on gravel, and clean firework bursts in the distance. Intonation: Soft, captivated, and relaxed. A natural, mature female voice speaking casually while walking, with a slightly lower pitch. [00:04 - 00:06] Phase 2: The Halt & The Intimate Offer Visual: The walking movement smoothly slows down and comes to a gentle stop. The camera frames her closely as she pauses. The festival background softly blurs into a realistic bokeh effect, with warm lantern light reflecting off her dark yukata. Action: Having stopped walking, she turns fully to face the camera. She physically leans her upper body forward and thrusts the candied apple directly close to the camera lens, noticeably invading the viewer's personal space to make the offer highly immersive. Audio/SFX: Subtle rustle of the yukata fabric as she stops and moves her arm. The footsteps fade out, leaving a closer, more intimate audio space. Intonation: Cheerful, warm, and inviting, with a playful rising intonation at the end, while maintaining her grounded mature tone. [00:07 - 00:10] Phase 3: The Quick Pull-Back & Tease Visual: Extreme close-up framing. Sharp focus on her facial expression, capturing natural, photorealistic micro-expressions and highly accurate lip movements. Action: In a swift, sudden movement, she quickly yanks the candied apple back out of reach just as the viewer attempts to take a bite. She immediately takes a bite herself, tilts her head slightly, and gives a teasing, mischievous smile. Audio/SFX: Crisp, loud, and distinct crunch of the candied apple's sugar coating breaking, synchronized perfectly with her swift bite. intonation: A mix of sweet satisfaction on the first word, followed by a subtle, alluring challenge on the question. Delivered in a low, breathy, mature voice. (CRITICAL AUDIO DIRECTIVES: Force exact phonetic rendering of the Japanese text. The AI must process the dialogue strictly as "Kirei da ne, hanabi." -> Pause -> "Taberu?" -> Pause -> "Amai. Hontou ni tabenai no?". Hard cut the vocal track immediately after the question mark on "Taberu?" to prevent trailing particles like "wo", "wa", or breath noises. Generate clean, natural vocals with zero random speech artifacts, digital distortion, or over-mixed background noise. Strictly avoid high-pitched anime tropes; prioritize a deep, mature vocal profile).