27. Claude + Seedance 2.5 AI UGC Ads
- Section: Code Assistant
- Video (NOT downloaded): https://www.youtube.com/watch?v=A10IZZDrpxs
- Video length: 14min50s
Description
See the prompt pack in the resources file attached below
Downloaded Resources
27-claude-seedance-25-ai-ugc-ads.pdf— The-Seedance-2.5-UGC-Ads-Pack.pdf (ok (358775 bytes))
FREE PROMPT PACK · SEEDANCE 2.5 + THE CODE ASSISTANT
The Seedance 2.5 UGC Ad Pack How to generate 30-second creator ads that look like they were filmed by a real person on a real phone — the three-step recipe, nine ready-to-copy-and-paste prompts (seven ready-to-use ad formats and two templates for your own product), and the realism rules that keep them from looking artificial.
Why Use Seedance 2.5 for UGC Ads
UGC-style creator videos are the best-performing ad format right now — authentic, benefit-focused, and native to the feed. The problem has always been production: creators, filming, reshoots. Seedance 2.5 in the Higgsfield MCP takes care of all that: one generation delivers a single continuous 30-second shot of someone speaking, in 1080p, with the voice generated in the same step and lip-synced to the lines you specify word for word. No splices, no separate voice-over, no cuts to hide edits.
PARAMETER | WHAT YOU USE IT FOR IN UGC
Model + mode | seedance_2_5 with mode: omni_reference — the mode is required whenever you provide reference images. Pure t2v mode rejects them.
References | Two images via image_references: your product photo and your creator portrait. Pass the generation job IDs, not uploads — platform-generated images skip the copyright check that blocks raw uploads.
Length | 4–30 seconds. The 30-second continuous shot is the centerpiece: an uninterrupted selfie video, exactly the length at which UGC converts (10–45s).
Resolution | 1080p (1080×1920 vertical) for the final ad. Draft in 720p at half the cost while testing a concept.
Aspect ratio | 9:16, always. TikTok, Reels, Shorts.
Audio | generate_audio: true — speech, ambient sound, and lip sync are rendered natively in the same job. Include each line of dialogue in the prompt with its timestamp.
Cost | A single 30s take in 1080p costs about 390 credits. A 720p draft of the same video costs about 195. Static product and creator images cost pennies.
Render time | 18–22 minutes for a 30s job in 1080p. Submit all your ads in a single parallel batch and review quality together.
Formats That Convert
The 2026 performance data is clear: UGC video accounts for most winning creatives across all formats, the product should appear in the first three seconds, and creative wears out in about 72 hours on TikTok — which is exactly why generating variations beats hiring creators. The nine prompts in this pack cover the seven UGC formats that keep winning — the review, the confession, the routine, the unboxing, the try-on, the product-only POV, and the taste test — plus two templates for the most common real-world case of all: you already have the product and only have a photo or screenshot of it.
The Three-Step Recipe
Every prompt in this package runs through the same pipeline. Give the full brief to your coding assistant with the Higgsfield MCP connected, and it will run the three steps — or run the steps yourself in the Higgsfield interface.
1 Generate the product photo first (GPT Image 2). A clean e-commerce photo: product, simple surface, natural light, short label text — one brand name plus two short lines render perfectly legible; a paragraph becomes unreadable. Set quality to high. This image is your product's identity for every ad you create for it.
2 Generate the creator (Soul 2). An "unretouched iPhone front-camera portrait" in the exact room where the ad takes place — bare skin texture, no beauty filter, amateur framing. The room, outfit, and lighting in the portrait should match the video prompt so the model locks into a coherent scene. Keep the creator job ID: the same face can star in an entire series.
3 Generate the ad (Seedance 2.5, one take). Pass both job IDs as image_references with mode: omni_reference, 9:16, 1080p, duration 30, audio enabled. The prompt is a scene list with timestamps: five moments, each with the exact line the creator says in quotes, plus a realism block and an audio specification. The model renders the performance, voice, and lip sync in one pass.
The Five-Beat Script Skeleton
Every 30-second ad in this package follows the same structure, and the timestamps matter — they set the performance rhythm. 0–4s HOOK: open mid-thought, with the product already rising into frame. 4–11s WHAT IT IS: name it, show the label. 11–18s DEMO: use it on camera — apply it, eat one, put it on. 18–25s RESULT: the proof moment, leaning toward the lens. 25–30s CLOSE: a low-pressure line, never a hard sell. About 70–80 spoken words fill 30 seconds at a natural pace — write more and the delivery feels rushed.