Raindance AI / The trend guide

THE TWO-PHOTO TUTORIAL

How to Make the Raindance AI Trend with Two Photos

Your portraits. A new sunset-pier duet. Here is the complete browser workflow, with the inputs, result, and limitations from our own test.

To make a Raindance AI video: upload two clear adult portraits, assign the seated and singing roles, confirm permission, and generate in the studio below. Preview the result, then download the MP4 with the Raindance song excerpt already added. A free preview is available when capacity allows; no app installation is needed.

What is the Raindance AI trend?

Here, “Raindance” refers to the Dave and Tems music-video trend: people use AI to place personalized characters in a seaside duet. It is different from the literal idea of a rain dance, where someone dances in falling rain. This tutorial focuses on a warm sunset, two people on a wooden pier, and a singing close-up.

Our Raindance AI video generator uses two photos and a fixed scene prompt to create a new fictional scene. It is an independent tool, with no affiliation with the artists. If you want an exact face swap of a particular performance, the method on this page does not provide that.

Prepare two separate portraits

  • Use one adult per photo, with permission from everyone shown. You must be 18 or older to use the studio.
  • Choose a clear face with visible hair and outfit. Even lighting, a simple background, and minimal filters make the reference easier to interpret.
  • Upload JPG, PNG, or WebP files up to 10 MB each. Each image must be at least 301 pixels wide and high; avoid very narrow strips or extreme panoramas.
  • You do not need a joint photo, a filmed dance, or a downloaded reference clip. Keep the two portraits separate.

Our published example uses a fictional adult man for the seated role and a fictional adult woman for the singing role. That is the pairing we have tested; other pairings may produce different results.

FROM PHOTOS TO MP4

How to make a Raindance AI video in 3 steps

  1. Upload your photos and choose who sings

    In the studio below, put the first portrait in 01 · Seated role and the second in 02 · Singing role. The first person is intended to sit on the pier; the second is intended to stand behind them and appear in the close-up.

    Check both thumbnail previews. If the roles are reversed, press Swap roles · choose who sings before generating. Changing these references guides appearance and role assignment; it cannot guarantee perfect likeness in every frame.

  2. Confirm permission and generate your scene

    Read the consent statement and confirm that you have permission to use both photos. When the studio is available, press Generate my free preview. Standard settings request a 15-second, 720p, vertical 9:16 video using ByteDance Seedance 2.0 Mini through KIE.

    The studio reports the real task state, including queueing, generation, and adding music. Processing can take several minutes. Keep the page open or return in the same browser to check progress. Do not submit a second take while the first is processing.

    The free preview is limited to one per browser every 24 hours, with network-based limits and daily capacity checks. If the button reports that generation is unavailable, check again later. For additional takes, see the current credit packs and purchase recovery page.

  3. Review, download, and check the music

    Once the result is ready, play it before saving. Look at both faces, hands, feet, clothing, and each scene change. Check that the people stay distinct and that the singer appears where you expected.

    Press Download MP4 to save the finished clip. The original Raindance song excerpt is added automatically after the AI video is generated; you do not need to add a sound just to hear music in the download. Since the model receives no audio reference, singing motions may not match the lyrics precisely.

    Download while the link is available: it expires seven days after task creation. Label the scene as AI-generated if you share it. The download does not grant rights to the song; check the permitted use of music on your chosen platform before publishing.

TRY THE WORKFLOW HERE

Make your Raindance AI video.

No app to install

Your Raindance studio

15s · 720p
Two portraits · new scene

Two photos and a scene prompt generate a new fictional sunset-pier video.

Accepted example · 14s · 480p · with music

Video model: ByteDance Seedance 2.0 Mini · API access via KIE.

One authorized adult per photo. Show the face, hairstyle and outfit. Photos guide appearance in a newly generated fictional scene; no source-video face or character replacement is performed.

15 seconds · 720p · MP4 with musicVertical · 9:16

Terms · Privacy · AIGC use policy

The studio is temporarily unavailable. You can still watch the generated example.

One free preview per browser / 24 hours. Daily capacity is limited.
No card, subscription, or app download.

YOUR VIDEO15s / 720p
AI-generated sunset pier with two fictional adults
Create your seaside moment

Assign a photo to each role, then generate your clip.

Original AI scene · two portrait references
Your generated video will appear here.

♫

Raindance music added automatically.
After AI generation, we add the original Raindance song excerpt to the MP4. Preview and download include music. Exact lyric lip-sync is not guaranteed; downloads do not grant music rights.

THE RAINDANCE RESULT

See the generated sunset-pier scene.

AI-generated characters
First input photo: a fictional adult manSecond input photo: a fictional adult woman
01 / Two input photosFictional AI portraits
02 / Generated example14s · 480p · with music

This accepted 14-second, 480p sample was generated by Seedance 2.0 Mini using these two fictional adult portraits and our scene prompt. No source video or audio reference was sent to the model. The original Raindance song excerpt was added afterwards. The girl stands on the pier boards in the opening shot. AI can still alter details and cut timing. The standard generation configuration is 15 seconds at 720p.

FIRSTHAND RESULTS

What our two-photo test actually produced

The playable example above is our accepted output, shown beside the two fictional portraits used to make it. We generated the scene with Seedance 2.0 Mini using those images and a written prompt. No source performance video or audio reference was sent to the model. We added the song afterwards.

Published sample versus current standard settings
DetailPublished sampleCurrent standard request
LengthApproximately 14 seconds15 seconds
Video size496 × 864 pixels, provider’s 480p tier720p, vertical 9:16
InputsTwo fictional adult portraits + scene promptTwo authorized adult portraits + built-in scene prompt
MusicAAC audio added to the H.264 MP4 after generationRaindance song excerpt added after generation

What worked: the output begins with both people on the pier, moves to a close-up of the singer, and finishes with the pair looking toward the sunset. The singer’s feet rest on the wooden boards in the opening shot rather than in the water.

What varied: the close-up starts at about three seconds and the rear view at about eight seconds, earlier than the intended four- and ten-second cuts. The singer also stands closer to the pier edge than the requested one-meter clearance. Treat a timed prompt as direction, not frame-accurate editing.

The sample demonstrates this method at its tested settings. It is not evidence that every new take will match the same faces, proportions, timing, or resolution, and it is not a measured 15-second/720p delivery.

How our Raindance scene prompt works

The studio supplies its scene prompt automatically. There is no prompt box to fill in. The following is an English explanation of the built-in storyboard, rather than a separate prompt you must paste:

  1. 0–4 seconds: establish both people. Begin directly with a full-body pier scene. Seat the first character on the left and place the singer on dry wooden boards behind them, away from the edge. Show the boards under the singer’s feet and keep the sea below or beside the pier.
  2. 4–10 seconds: show the singer. Cut to a chest-up singing close-up with the sunset sea in the background. Preserve the singer’s face, hairstyle, and outfit from the second portrait.
  3. 10–15 seconds: finish at sunset. Show both characters from behind, seated at the end of the pier and looking toward the evening light.

Additional instructions keep the cast to two people, preserve each person’s appearance, and discourage floating feet, standing in water, body distortion, or swapped identities. These requests guide the model; the sample above shows why they should not be treated as guaranteed controls.

Music is a separate finishing step. This distinction matters: adding the song to a rendered clip makes it audible, but does not retroactively synchronize the generated mouth movements to every word.

Common problems and what to check

The face or outfit changes.
Try a sharper single-person portrait with visible hair and clothing. Avoid sunglasses, heavy filters, and faces partly hidden by hands. Compare the output with the references before sharing; a new photo may help, but consistent likeness is not guaranteed.
The wrong person sings.
Check the two role labels and use the swap button before starting another take. If the references were correct but the result still mixed roles, the model did not follow the instruction consistently.
The body proportions or feet look wrong.
Review the full-body opening, not just the close-up. Unnatural limbs or poor contact with the pier are generation defects. A clearer portrait may help a later take; the prompt cannot promise to eliminate them.
The cuts or singing do not match another video.
This method creates a new scene without a motion-reference clip. Exact choreography, original camera timing, and word-by-word lip-sync are outside this workflow.
The preview seems silent.
Use the player’s audio control and check your device volume. Play the downloaded MP4 too. The published example contains an audio track; if your completed file has none, report the issue through the support link in the footer.
The studio is unavailable or still processing.
Follow its live status. Limited provider capacity can prevent new takes, and an existing take can remain queued for several minutes. Return in the same browser instead of repeatedly submitting. If it fails, read the error before retrying.

Raindance AI tutorial FAQ

What is the Raindance AI trend?

In this guide, it means making a personalized AI music-video scene inspired by the seaside mood associated with Dave and Tems’ Raindance. It is different from generating a person dancing in the rain. Our version creates a new fictional sunset-pier duet from two adult portraits.

Can I make a Raindance AI video for free?

The studio offers one free preview per browser every 24 hours, subject to daily capacity and network-based limits. No card or subscription is required for that preview. Check the studio’s live availability before starting; free access is not unlimited.

Do I need an app, an account, or a reference video?

No app installation or account sign-up is required for the browser preview. You need two separate authorized adult portraits. This workflow does not accept a source performance video or require filming yourself dancing.

Can I use one photo, a group photo, or two photos of the same person?

This tutorial uses two separate portraits, one adult per image, to guide two different people. A single-photo workflow is not supported here. Our published test used two distinct fictional adults; we have not verified a group-photo or same-person duet.

Which photo becomes the singer?

Photo 1 is assigned to the seated role and photo 2 to the singing role. Use “Swap roles · choose who sings” to exchange the references before generating. These are instructions to the AI; appearance and role consistency can still vary.

Does the downloaded video include music?

Yes. The site adds the original Raindance song excerpt after AI generation. The preview and downloaded MP4 include it. The AI does not receive the song as an audio reference, so exact lyric lip-sync is not guaranteed. Downloads do not grant music rights.

Is this a face swap of the original Raindance video?

No. Two portrait references and a written scene prompt generate a new video. This does not replace faces in an existing clip, copy its exact choreography, or preserve every camera movement from the original.

Why is the example 14 seconds at 480p when the studio says 15 seconds at 720p?

The published example is an earlier accepted test: approximately 14 seconds, with a 496 × 864 pixel video in the provider’s 480p tier. The current standard request is 15 seconds, 720p, and vertical 9:16. The example demonstrates the method, not a measured 15-second/720p delivery.

How long does generation take, and what if it fails?

The provider queue can take several minutes. Follow the actual task state and return in the same browser to check progress. Failed attempts do not consume the free preview allowance. Read the displayed error before trying another photo; do not start a second take while the first is still processing.

What happens to my photos and download?

Photos are submitted when you generate. This site does not save original photos in its own file storage. KIE’s upload documentation states a 24-hour upload lifetime; that does not establish retention for all provider records. Video downloads expire seven days after task creation. See our Privacy Policy for storage, cookies, and deletion details.

Sources and further reading

The results and sample measurements on this page come from our own published test. The model reference is KIE’s ByteDance Seedance 2.0 Mini documentation. For how this site handles your uploads, downloads, and permissions, read our Privacy Policy, Terms, and AIGC use policy.

Ready to try your portraits? Return to the generator or watch the sample again.