1. Choose two clear photos
Use one photo for each person. Choose a well-lit picture with an unobstructed face, a natural expression, and enough room around the head and shoulders. A recent photo is easier to compare with the result.
If your only image is a photo of you together, make a separate crop for each person. Keep the full face and hair in each crop. A crop with a tiny or heavily blurred face is a poor reference, even if the original group picture looks good on your phone.
- Prefer a front-facing or gently angled face over a strong profile.
- Avoid sunglasses, hands over the face, beauty filters, and screenshots of a screenshot.
- Use photos only with each person's permission. A public profile photo is not permission.
- Follow the upload requirements shown in the generator; supported file formats and limits may vary by route.
2. Pick the singer and the partner
In a duet, the singer is the person featured in the performance shots. The partner appears alongside them in the shared scenes. Choose those roles before submitting, and check that your two photos are in the correct slots.
The roles describe the video, not the people. A couple or two friends can use the same workflow. Selecting a singer does not mean the service will clone their voice or guarantee exact lip synchronization.
3. Choose the route that is available
Choose between Original MV transfer and Original scene. Original MV transfer uses prepared reference footage for its movement and camera timing. Original scene makes a new scene from your references and the chosen template; it is not original music-video footage.
Compare the scene structure, duration, aspect ratio, audio, and credit cost displayed for your route. Those specifications apply to that route, not to every clip described as Raindance AI elsewhere.
A route that depends on reference footage or additional production setup may be marked Coming soon. That label means it cannot currently be submitted or purchased. The existence of a route card is not evidence that a finished music-video recreation is already available.
Check the live availability notice before uploading. If generation is unavailable, you can read the guide and prepare your photos, but there is no real render to start and no need to buy credits for it.
4. Review the cost before you submit
Video generation uses credits from a one-time pack when paid generation is enabled. Read the exact credit charge for your selected route before confirming. Different settings can have different costs, so a credit balance is not the same as a guaranteed number of every type of video.
Review your photos and roles one more time. A new creative take may use additional credits; a technical failure is handled according to the published refund policy. Do not repeatedly submit while a job is already waiting or processing.
5. Check the result before downloading
When a real generation finishes, review the full clip. Check both faces, the person assigned to each role, visible hands, camera framing, and any audio the selected route actually includes. AI output can vary, and a pleasing scene does not guarantee every detail is correct.
Use the history or result page to find the generation's current status and available downloads. If there is a problem, keep the job or order reference and contact support through the published channel. Describe the issue rather than uploading more people's photos unnecessarily.
- Keep your downloaded original before editing or posting.
- Use a silent export when you plan to add a platform's own sound.
- A new take is a new output; it is not a manual edit to every frame of the first one.
6. Add your sound and publish transparently
Use audio that is available for your account and intended use. A track appearing in a social app does not automatically grant permission to embed it in an exported file or use it commercially. This tool is independent of Dave, Tems, and their record labels.
For a photo-to-video reveal, place a brief still image before the clip in your editor and choose a cut that suits your audio. Review the final crop so neither person's face is cut off. You do not need to imitate a specific post to make a duet that feels personal.
TikTok requires labeling realistic AI-generated images, audio, and video. Use its AI disclosure controls and review the platform's current requirements before posting. Other platforms have their own rules.
Common questions
Do we need a photo of us together?
No. Prepare a separate clear image of each person. If you start from a shared photo, crop each person into an individual reference without cutting off their face or hair.
Can two friends make a Raindance AI video?
Yes, the two-person workflow can represent friends as well as couples. Both people must consent, and the available route determines the scene and performance.
Does selecting a singer clone their voice?
No. The singer selection assigns a visual role. Check the route's audio description; it does not promise voice cloning or exact synchronization with a song.
How long will my video take?
Use the estimate shown by the active route and your job's live status. Queue length and provider availability can affect delivery. This guide does not promise a fixed generation time.
What can I do while generation is unavailable?
Prepare consented photos, choose the roles you want, and read the route descriptions. A Coming soon route cannot create a real output yet, and you should not be charged to submit it.
Sources and review
Reviewed on . Sources cover the music and platform context. Check the generator for current availability, output settings, and credit costs.