How the Dime Dime Baby Prompt and Reference Video Work

The site's template combines a photo with a motion reference and a fixed instruction. When submissions are available, you supply the photo through the generator. The application selects the text, reference and output settings. Check the configured model, exact instruction, reference source, audio policy and readiness in the current template details. The generator shows whether new tasks can be submitted.

The instruction below records an independently adapted Mini test from 10 November 2026. Aireel's original model and prompt remain unknown. Its public effect page supplied the visual reference reviewed during implementation. Aireel reference effect

Give each input a specific job

Input	Intended role in this recipe	What to inspect in the result
Identity photo, referenced as @image1	Guide the central adult's face, hair, skin tone and clothing	Whether those details stay recognizable as the subject moves
Motion video, referenced as @video1	Guide party setting, framing and the ordered action	Whether approach, bite, camera glance, turn and return form a continuous scene
Written instruction	Specify which subject changes and describe the sequence and constraints	Whether the model follows those directions across the complete clip

These are the intended roles of the inputs. A generated result still needs review. The person can remain recognizable while a hand, background detail or transition changes unexpectedly.

The party scene has one central protagonist and background guests. A two-person BbY WOW Dance requires its own choreography reference and evaluation. This recipe has only been tested for the single-protagonist party action.

The instruction in the 10 November Mini test

The text below is preserved from the recorded test recipe, party-bread-v1. The @image1 and @video1 labels refer to reference inputs passed with the request.

Use @video1 only for the party setting, camera framing and action sequence. Replace the central protagonist with the adult in @image1. Preserve the uploaded person's face, hair, skin tone and clothing; the image is an identity reference, not the first frame. Keep the background guests as background guests.

An eight-second continuous party scene in a decorated indoor reception hall with colored lights and people moving in the background. A single fully clothed adult approaches the camera holding a small bread roll in one hand.

0–2 seconds: walk forward from the group toward the camera in a relaxed rhythmic step.

2–4 seconds: raise the bread, take one small visible bite, lower the hand and look directly into the camera with a calm, amused expression.

4–6 seconds: pause briefly in close view, keeping the same face and clothing, then turn around naturally.

6–8 seconds: walk back toward the party guests, ending with the protagonist moving away from the camera.

Maintain the continuous spatial layout, natural hands, one bread roll and the same protagonist throughout. This is the single-person party bread-bite scene, not a two-person dance routine. No captions, logos, speech or music. Silent output.

The timing blocks express the requested sequence. They give the model an order to follow. The test review described below did not establish frame-accurate timing for each beat.

The settings in that recorded test

The recorded test used Kie's bytedance/seedance-2-mini model with one identity image and one reference video. The test request selected 8 seconds, the 480p preset, 4:3 framing and audio generation disabled. The reference-input contract was checked during implementation on 10 November 2026. Provider model documentation

Request field	Value in that test
model	bytedance/seedance-2-mini
prompt	The complete instruction quoted above
reference_image_urls	One identity-image URL, assigned to @image1
reference_video_urls	One motion-reference URL, assigned to @video1
duration	8
resolution	480p
aspect_ratio	4:3
generate_audio	false
nsfw_checker	true

The motion input was the complete eight-second visual track from the reviewed hero reference, resized to 960 × 720 at 30 fps, with its audio removed. The identity input was a 600 × 800 JPG crop from a related party example at 4.5 seconds. That choice matters when interpreting the result because the person was already pictured in a similar scene.

The photo is supplied as an identity reference. The request uses reference-image and reference-video fields. It leaves first-frame and last-frame fields unused. To adapt the instruction elsewhere, first check that the selected interface supports both reference types and its own reference-label syntax.

What the single recorded test showed

The implementation record dated 10 November 2026 contains one billable task tracked to success. The resulting MP4 decoded completely. It contained 193 frames at 24 fps, measured 752 × 560 pixels, ran for 8.041667 seconds, and had no audio stream. The measured dimensions describe that file. “480p” was the selected provider preset.

The recorded full-clip review found recognizable curly hair, a blue sweater, a white collar and the central face through the action. The subject brought bread toward the mouth, bit or chewed, looked toward the lens, turned and moved back toward the group. The same review noted bright foreground light flares and generated hand and background details.

The provider reported 38.4 provider credits consumed, matching the immediately observed balance change. This is the charge recorded for that technical test. Review the site's app-credit charge beside the submission controls before submitting a task.

Because the identity crop came from a related party example, ordinary portraits, other clothing, pets and multiple uploaded identities remain unevaluated by this test. A single completed clip also leaves the success rate and repeatability unmeasured. The test fixture and output are private research materials. The homepage describes the reviewed action sequence and links to the original reference page. Read the party sequence

Use the prompt guide to review a result

Read the sequence once before watching an output. Then watch the complete clip and check the face during approach, the hand and bread during the bite, and continuity through the turn and departure. Keep separate notes for identity, action and scene details. This makes an observation such as “recognizable face, distorted bread hand” precise enough to act on.

That recorded Mini test requested a silent file, and its result was silent. The current production template includes the original music from its reference video in the downloaded MP4. Including the audio does not grant a commercial music license. The song guide covers identification and use permissions. Find BbY WOW and check the sound

If your goal is to place your media into a ready-made design, inspect a particular CapCut listing and its preview. The template guide provides three official entries and the fields to compare. Choose a CapCut template

For the site's photo-based party workflow, check the current availability and upload requirements at the generator. Open the generator