How to Make a Hawak Mo Ang Beat AI Video

1. Choose a photo that gives the dancer an identity

Start with one person, a sharp face and even light. A waist-up or full-body image gives the model more information about proportions and clothing than a tightly cropped selfie. Keep the eyes, hairline and outline of the face visible. Avoid motion blur, a hand over the face or a distant person in a group photo.

Upload a JPG, PNG or WebP up to 10 MB in the template workspace. Preview the file before continuing. If you change your mind, replace the photo without rewriting the prompt. Use a picture that shows the outfit you want: asking to preserve an outfit while also replacing it gives conflicting directions.

2. Read the dance before describing it

The reference supplies a short, continuous performance at the water’s edge. Watch the sway, the gestures near the face and the overhead pose. The camera stays relatively steady while the person moves inside the frame. A complicated orbit, sudden zoom or new setting would make it a different shot.

A written prompt can describe these actions, but it cannot carry every frame of choreography. For close motion matching, choose a generation workflow that actually accepts reference video. Keep the source clip and your person photo as separate inputs. Do not assume a model supports motion transfer simply because it supports image-to-video.

3. Use the template and carry your photo forward

Open the Hawak Mo Ang Beat template, add your photo and expand the prompt only if you want to edit it. Start with the full-performance prompt above. Use the optional notes for specific details such as short hair, a blue shirt or keeping the face unobstructed; avoid adding a second dance sequence.

Select Continue with this photo to move into the video workspace. Your image, prompt, notes and motion-reference link travel in the same tab. Review the prompt and save or copy the brief. Keep the tab open while working: local file references may need to be selected again after a full reload.

Treat duration, aspect ratio and resolution in a creative brief as planning choices. Before sending a generation task, check the selected model’s accepted inputs, output options and price. The reference’s duration is not a universal model setting.

4. Keep the soundtrack and the full gesture in view

Click the example to play it with its original sound. Hover previews elsewhere may be muted, but explicit playback includes a volume control. The reference is approximately 14.4 seconds; its original 540 × 958 frame is close to, but not exactly, 9:16. The lighter page copy keeps the full performance and audio.

Source audio does not automatically transfer through a text prompt. Check whether your renderer keeps it, accepts a separate audio input or needs a later editing step. When preparing a finished clip for TikTok, use fit mode in the video resizer if a crop would remove raised hands. Listen to the exported file as well as watching it.

5. Review identity, hands and timing before sharing

Pause at the beginning, during the gesture nearest the face and at the highest arm position. Compare the face, hairstyle and clothing across those moments. Look for changing fingers, merged arms, fabric that changes shape or hands that disappear beyond the frame. These are better checks than judging only a flattering first frame.

If the face changes when a hand crosses it, try a clearer source photo and simplify appearance instructions. If the arms are cropped, increase framing room rather than demanding larger gestures. If the dance is unrelated to the reference, verify that the motion input was accepted instead of adding more adjectives to a text prompt.

For a timing mismatch, compare picture and sound together. Check that the correct source was used and that trimming or frame-rate conversion did not alter the sequence. Revisit the model’s motion and audio capabilities before assuming a different prompt will solve synchronization.