AIタレントが写真の中のベンチに座るのは完璧に出来たと思う。
ところが自分の名前、Nanamiを正確に発音できていない。
以前七海と表記した時に、うまく発音できなかったので、あえてNanamiにしたのだが。
次回は七海（ななみ）にした方が良いのかも。
わざと長いセリフをAIタレントに喋らせてみたが、まあまあの結果だと思う。
正直言って、失敗すると予想していたのだ。
長いセリフは失敗する確率は高いからだ。
今回のプロンプトは以下です。
最初の動画生成用プロンプト
A beautiful young Japanese woman named Nanami, exact same face and appearance as reference images, shoulder-length brown hair with bangs, heart-shaped earrings, bright genuine smile, photorealistic, highly detailed skin texture, realistic eyes.
She is wearing the exact same floral summer dress, straw hat, beige sandals, and woven bag as in the provided reference image.
Scene: Exact same park scene as the attached background photo with blue benches, trees, paved area, buildings and parking lot in the background, bright daytime natural sunlight.
Camera starts at waist shot (waist up), facing Nanami who is standing in front of the camera.
Full sequence (10 seconds total):
- Nanami stands in waist shot, smiles at camera and quickly says: 「AIのNanamiだよ、これから写真の中のベンチに座ってみるね」 (natural lip sync, friendly expression, about 2.5 seconds).
- She turns and walks naturally but at a brisk pace toward the leftmost blue bench.
- Camera smoothly follows her while gently zooming in.
- She reaches the bench, sits down gracefully, turns to face the camera.
- Camera settles into a clean bust shot as she gives a warm smile and holds the gaze (final 2-3 seconds).
Cinematic, smooth natural motion, realistic physics, detailed fabric movement, subtle breathing, realistic lip sync, seamless camera follow and zoom, exactly 10 seconds, high quality, 16:9, photorealistic, no deformation, no blur, perfect character consistency.
Exact same woman as all reference images: identical face, eye shape, nose, mouth, smile, jawline, skin tone, hair style, body proportions, height, exact outfit details. Strong face and body consistency, no face drift, no body distortion, perfect match to provided character references.
Negative prompt: blurry, deformed, ugly, extra limbs, mutated hands, poor anatomy, bad proportions, text, watermark, logo, low quality, cartoon, anime, 3d render, plastic skin, overexposed, underexposed, motion blur on face, unnatural smile, distorted background, deformed face, extra fingers, bad hands, bad lip sync, abrupt camera movement
生成された10秒動画をさらに10秒延長するプロンプト
A beautiful young Japanese woman named Nanami, exact same face and appearance as in the attached video, shoulder-length brown hair with bangs, heart-shaped earrings, bright genuine smile, photorealistic, highly detailed skin texture, realistic eyes.
She is wearing the exact same floral summer dress, straw hat, and bag as in the attached video.
Scene: Exact same park with blue benches as in the attached video, bright daytime natural sunlight.
Camera continues seamlessly from the last frame of the attached video (Nanami sitting on the left blue bench, smiling at camera in bust shot).
Full sequence (exactly 10 seconds total):
- Starts exactly from the end of the attached video.
- Nanami maintains eye contact with the camera and speaks naturally with a friendly, gentle tone: 「この公園は、ふらの市のやまべ地区にあります。国道のすぐ近くです。国道をふらの側に行くと、メロンの直売店がたくさん並んでいるよ。」
- Natural and clear Japanese lip sync, soft head movements, warm smile, expressive eyes, charming and informative expression as if explaining to the viewer.
- After finishing the sentence, she ends with a gentle smile looking at the camera.
Cinematic, smooth natural motion, realistic physics, subtle breathing, perfect lip sync, seamless continuation from attached video, exactly 10 seconds, high quality, 16:9, photorealistic, no deformation, no blur, perfect character consistency.
Exact same woman as the attached video: identical face, eye shape, nose, mouth, smile, jawline, skin tone, hair style, body proportions, height, exact outfit details. Strong face and body consistency, no face drift, no body distortion, perfect match.
Negative prompt: blurry, deformed, ugly, extra limbs, mutated hands, poor anatomy, bad proportions, text, watermark, logo, low quality, cartoon, anime, 3d render, plastic skin, overexposed, underexposed, motion blur on face, unnatural smile, distorted background, deformed face, extra fingers, bad hands, bad lip sync, abrupt cut
