How to Make a Rumpelstiltskin AI Video

Making a Rumpelstiltskin AI video comes down to two jobs. The template supplies the straw room, the candlelight, the dance and the camera move. Your job is to supply two clear photos and pick an output frame. Here is the full process, including the mistakes that waste the most credits.

Step 1: Prepare two photos, not one

The clip has two performers, so it needs two uploads. One slot is for the seated woman, the other for the little man who dances in front of her. The two slots are separate, so a single group photo will not work even if both people are in it.

For each slot, use a recent photo with even lighting and one person in frame. Avoid sunglasses, hats, heavy beauty filters, or anything covering the face. A front-facing portrait with the head and shoulders clearly visible gives the model the most to work with.

You do not need to match the pose. Sitting, standing, smiling or looking serious in your photo is fine. The seated posture, the arm movements and the timing all come from the template, not from your uploads.

Step 2: Check what the template keeps

The template is one continuous shot, and it is deliberately fixed. Everything below comes from the source clip and survives the swap untouched:

• The straw-filled spinning room and the stacked bales
• The candle-lit colour grade and the wide two-shot framing
• The camera move and the timing of the dance
• The original audio

Only two things change: the identity, hair and wardrobe of each figure. That is why the result stays consistent between runs, and why you can repeat a batch of posts without the scene drifting.

Step 3: Pick the output frame

Four format options ship with the generator:

Widescreen is the closest match to the source clip. Vertical crops in on the two figures, which usually works well because the dance happens in the centre of the frame. Square is the safest option when you are not sure where the clip will end up.

Step 4: Choose a resolution

Three tiers are available, and they cost different amounts of credit:

• 480P uses 50 credits and is the cheapest way to test a pairing.
• 720P Standard uses 100 credits and is the practical default for finished social content.
• 1080P Pro uses 200 credits and gives the highest detail.

The exact credit cost is shown in the form before you submit, so you always know what a render will spend.

Step 5: Generate and review

Submit the task and the preview panel shows live progress. Most tasks finish within a few minutes, but video models queue under heavy load, so a task can take longer during peak hours. The finished clip appears in the same panel with a download button.

Finished renders stay in your history, so you can compare several pairings and download the one that works. Each task gets its own tile under the preview, and you can delete the ones you do not want to keep.

What to avoid

Most disappointing renders come from the uploads rather than the model. A few things to check before you spend credits:

1. Two different people in two slots. Uploading the same photo twice produces the same face in both roles and wastes the run.
2. Covered faces. Sunglasses, masks and heavy shadows all reduce how much identity the model can carry across.
3. Group photos. One person per upload. A crowd gives the model nothing to bind to.
4. Extremely low resolution. A small, blurry source photo cannot produce a detailed face at 1080P.

Posting the result

The clip works on any feed, and the audio already matches the motion, so it does not need extra editing. If you want to use it as a green screen source instead, export the render and composite your own background around the two figures.

When you want several variations, the 5-video and 15-video packs cover multiple Pro renders, and a monthly plan lowers the per-video cost further. Failed provider tasks refund unused credits automatically, so a bad queue day does not cost you the render.