Can I turn a video into a depth-map video?

Yes. Open the converter, drop an MP4/MOV/WebM, and export a grayscale control video. Inference uses Depth Anything V2 Small in the browser.

Does this upload my footage?

The conversion step does not. If you attach the result as a generation reference, that file is stored so the video model can fetch it — same as any other reference asset.

Which models use a depth reference?

Seedance (including Fast, Mini and 2.5), MiniMax Hailuo 03, and Wan 3. Other clip rows ignore it.

Why not attach the original clip?

A full-color reference leaks identity. A depth map keeps camera and blocking so you can recast the shot while matching motion. For the song itself, start at Suno to Video or Udio to Video.

How long can the clip be?

The first 15 seconds, 8–24 fps, up to 80MB. Chrome or Edge with WebGPU is fastest; Safari falls back to WASM.

Yes. Credits are only charged when you confirm a reshoot that uses the depth clip as a reference.

Video to Depth Map: Keep the Camera, Recast the Shot

A prompt describes the scene. A depth video tells Seedance where the camera sat and how bodies moved. Convert any clip in your browser, then attach it on a SunoMV clip before you reshoot.

Open the converter

Runs in the browser No upload to convert Seedance · Hailuo · Wan 3

Jadikan SunoMV sumber pilihan di Google Lihat lebih banyak SunoMV di Berita Utama dan jawaban AI.

Runs: In the browser (Depth Anything V2 Small)
Length: First 15 seconds
Format: Grayscale depth (near = bright)
Models: Seedance · Hailuo 03 · Wan 3

Use this when a remake has to hold the original camera.

If you only have lyrics, skip it. If you are cloning a viral shot and the prompt keeps drifting the blocking, convert the source clip to a depth map and attach it. Conversion is local and free. Generation still costs whatever that clip row costs, shown before you confirm.

Why remakes fail without a depth map

Uploading the original clip as a reference invites the model to copy faces and wardrobe. A grayscale depth map keeps motion and camera, so you can change the cast.

Near = bright, the control format

Depth Anything V2 Small runs in the tab and exports a grayscale video. That is the format Seedance and Hailuo expect as a motion-control reference.

Nothing leaves the device to convert

The converter does not upload. First run downloads a ~65MB model into the browser cache; the selected video stays local.

Fits the clip window

The first 15 seconds at 8–24 fps, up to 80MB. That matches Seedance's clip length, so you are not converting footage the model cannot take.

Where it sits in a SunoMV reshoot

Identity still comes from your character stills and lyrics. The depth clip is an extra reference on models that accept reference video.

Attach on Seedance, Hailuo or Wan 3

Those rows take a reference video. Other clip models ignore the attachment — switch the row if the field is disabled.

Convert first, then confirm

The clip editor can convert in place, or you convert here, download, and attach. Credits are charged only when you confirm the reshoot.

Pair with TikTok-to-MV clone

Clone the song and structure from a viral clip, then hang the depth video on the shots that have to match the original blocking.

From a viral clip to a recast take, three steps

Nothing to install. Chrome or Edge is fastest.

1

Convert the source

Open the converter, drop the clip, pick grayscale, download the depth video. The original never leaves the tab.
2

Open the clip in SunoMV

Pick Seedance, Hailuo or Wan 3. Attach the depth video on the clip editor (convert-in-place is also there).
3

Reshoot

Character stills still lock identity. The depth map locks camera and blocking. Confirm the price, then generate.

When a depth map is the right reference

Not every clip needs one. These are the jobs it is for.

Viral remakes and dance locks

The shot has to hit the same silhouettes and camera path. A depth map is the control video; the stills are the new cast.

Prompt-only takes that drift

You already like the lyrics and the stills, but the camera wanders. Attach depth instead of rewriting the prompt again.

Clone a TikTok, keep the blocking

Use TikTok-to-MV for the song, then hang the depth video on the shots that have to match the original cut.

Video to depth map, answered

What is a depth-map video?

A grayscale video where brightness is distance from the camera — near is bright, far is dark. Video models that accept a reference video can follow that structure as motion control.

Is this the same as depth.cards?

Same job: video in, depth video out, in the browser. This page is the SunoMV path — convert, then attach on a clip that supports reference video.

Do I need a GPU?

WebGPU on Chrome or Edge is fastest. Safari and Firefox fall back to WASM, which is slower but works. No Python, no CUDA install.

FAQ

Pertanyaan Umum

Tanya apa saja ke kami!

Panduan terkait

Lihat semua 34 artikel di Teknik & Penyutradaraan Video Musik →

SunoMV core tools

Convert a clip, then reshoot with the depth map attached

Open the in-browser converter, download the grayscale video, and pin it on a Seedance clip in SunoMV.

Convert a clip free