Flux 3 vs. Seedance 2.5 · Selfie Vlog Comparison

Flux 3 + Seedance 2.5VlogVerticalWorkflowLip SyncSeedance 2.5Flux 3Drone AerialHandheld CameraOcean20s
The same Instagram Reels-style selfie prompt with native audio was fed into both Flux 3 and Seedance 2.5, with both finished cuts placed side by side on the same page. A lip-synced selfie is one of the toughest tests for comparing models.
PROMPT · 视频提示词
Create with AI
Create a photorealistic 20-second, 9:16 Instagram Reels-style selfie vlog with synchronized native audio. A beautiful 22-year-old adult woman with long honey-blonde hair wears a tasteful cream V-neck summer dress, delicate gold jewelry, sunglasses, and a small luxury handbag with no visible logo. Keep exactly the same face, hair, outfit and voice throughout. The entire video is filmed by her at arm's length on a front-facing smartphone. Authentic rich-girl influencer energy: relaxed, playful and casually privileged, never theatrical. Natural skin texture, slight hand shake, imperfect framing, brief autofocus and exposure changes, wind in her hair, occasional glances at her screen. It must feel like a spontaneous Instagram Reel, not a commercial. 0–6s: Walking from a lived-in Monaco penthouse onto a sunny terrace, harbor and yachts casually visible behind her. She adjusts the phone angle and says in natural American English: "I thought today was low-key, but apparently we're having lunch in Capri." Small amused smile. Her palm covers the lens. 6–13s: The hand-cover transition reveals a private marina. She walks toward a sleek white yacht while staff load one suitcase behind her. She brushes windblown hair from her face and says: "Dad sent the little boat. The other one's still in Saint-Tropez." She glances at the yacht, realizes how ridiculous that sounds, and gives one short genuine laugh. 13–20s: A quick whip-pan reveals her seated on the moving yacht, sparkling sea and coastline behind her. She lowers her sunglasses and says: "I packed nothing, but whatever—I'll shop when we land. Normal Tuesday." Hold a straight face, then a tiny knowing smirk; reach toward the lens to end recording. Real phone audio: coastal wind, footsteps, marina water, distant seabirds and yacht engine. Very soft chic house beat under the dialogue. Dialogue exactly as written, clear lip sync, no extra speech. No subtitles, text, logos, watermark, third-person shots, drone shots, beauty filter, plastic skin, distorted hands, outfit changes or face changes.0:200:20
✍️ Editor’s Notes

模型名「Flux 3 + Seedance 2.5」组合本身就是这条词的关键信息——大概率是先用Flux 3出人物一致性强的基础帧或角色锚点,再交给Seedance 2.5做动态和对口型,这种双模型接力的写法要求提示词里反复强调「exactly the same face, hair, outfit and voice throughout」,就是在给交接环节上保险。三段剧情靠一句对白加一个手掌遮镜头的硬转场推进,台词写得极口语化且带精确的情绪注释(「realizes how ridiculous that sounds」),这是在给对口型和微表情同时下指令。结尾负面清单点名「no third-person shots, drone shots, beauty filter, plastic skin, distorted hands」,全是自拍vlog类视频最容易露馅的AI感来源,说明作者对这类真实感UGC题材的常见翻车点很熟。

Create with AI
Alt. Take
Related Works
×