Higgsfield AI Music Video: Clone Yourself and Make a Singing Video (Workflow Guide)

Last updated: February 2026
Author: Greg Preece — I test AI video tools hands-on and show creators how to get usable results fast (without turning your workflow into chaos).

If you want an AI “clone” of yourself that can sing any line of a song in basically any location, this is the simplest workflow I’ve found inside Higgsfield.
You’ll create a consistent character from a handful of selfies, generate a strong “master” image, then use Speak + a motion preset (like Rap Flex) to turn it into a performance shot.

TL;DR (Summary box)

• Create a character in Characters by uploading lots of clear selfies.
• Generate a strong location image first (pick the best of the 4 options).
• In Speak, upload your audio, choose a motion preset (Rap Flex worked well for rap), and generate on High quality.
• If plan/pricing access has changed since recording, verify on the official site.

Prefer to watch? Here’s the video. Prefer to skim? The full breakdown is below.

Quick links

• Higgsfield — Try Higgsfield →

Table of contents

• What this workflow lets you do
• What I tested and what I found
• Step 1: Create your AI clone in Characters
• Step 2: Generate a strong “master” image in your location
• Step 3: Make the image sing in Speak
• Troubleshooting: the fixes that matter
• FAQ

What this workflow lets you do

Higgsfield (Higgsfield AI) combines character creation, image generation, and speaking/lip-sync video generation in one place — so you can produce a performance shot without filming anything in real life.

This is ideal if you’re a creator who:

• wants music-video style visuals without travel, permits, or a shoot budget
• needs a consistent “you” character across multiple scenes
• wants quick iteration: generate 4 options, pick the best, move on

Try it here: Try Higgsfield →

What I tested and what I found

Based on what you see me do in the video:

• More (clear) selfies = better clone quality. I specifically look for the upload status to hit “excellent” before generating.
• The still image matters more than people think. When the still looked right, the final performance felt way more believable.
• Default motion looked stiff. Picking a motion preset made it feel like an actual performance.
• High quality was noticeably better than Medium (Medium didn’t hold up in my tests in the video).

Step 1: Create your AI clone in Characters

This is where you build a consistent version of your face that Higgsfield can reuse.

Caption: Upload multiple clear selfies (no obstructions) until the status indicates strong coverage, then generate your character. Screenshot description: The “Characters” area showing “Create Character” with an upload panel and a progress/status indicator for training quality. Alt text: Higgsfield Create Character photo upload screen

Do this

1. Open Higgsfield and click the Character / Characters area at the top.
2. Click Create character.
3. Upload lots of images of yourself where your face is clearly visible (multiple angles helps).
4. Keep adding photos until the quality/status indicator shows strong coverage (in the video, I’m looking for it to say “excellent”).
5. Name your character (I used something like “AI Greg”) and click Generate.
6. Wait for it to finish (in the video, it took around a few minutes), then your character appears in your character list.

Outcome you’re aiming for

A reusable clone you can drop into any new scene without your face changing every generation.

Step 2: Generate a strong “master” image in your location

Before you make anything “sing,” generate a single great image of your clone in the location you want. This becomes the foundation for the final video shot.

Caption: Prompt your clone into a specific location and outfit, choose the right aspect ratio, then generate and pick the best result. Screenshot description: The image generation screen showing a prompt field, aspect ratio controls (horizontal/vertical), quality selection, and style/preset options. Alt text: Higgsfield image generation prompt box for a location scene

Do this

1. Hover over your character and click Generate (or the equivalent “generate image” action on your character).
2. In the prompt box, describe:
   • the location (e.g., “standing on a tall skyscraper”)
   • what you want the character wearing
   • any vibe cues (lighting, mood) in plain English
3. Set the aspect ratio for your target video format (horizontal if you’re making a standard landscape video).
4. Keep quality high (as shown in the video).
5. For style/presets, I used:
   • Spotlight (a dramatic, direct-light look)
   • Fisheye (a wide, lensy look)
6. Click Generate and review the 4 outputs.
7. Pick the best image — the one that actually looks like you, with the strongest lighting and detail.

Outcome you’re aiming for

One “this is undeniably me” still image in the right scene — because the next step builds motion and lip-sync on top of it.

Step 3: Make the image sing in Speak

Now you’ll take that still and generate a performance clip.

Caption: In Speak, attach your audio, choose a motion preset (Rap Flex worked well for rap), set quality to High, then generate your singing shot. Screenshot description: The Speak interface showing a selected image thumbnail, an audio upload/select area, motion preset dropdown, and a Generate button. Alt text: Higgsfield Speak screen with audio upload and motion preset selection

Do this

1. Open your chosen image and click the Speak button to enter the Speak area.
2. Select the still image you want to animate (your “master” image).
3. Upload your audio:
   • In the video, I used an MP3 and I’d already split it into line-by-line chunks, so I could pick one line at a time.
4. Choose your motion/performance:
   • I used Rap Flex because it looked the most natural for a rap music video vibe.
   • If you want something else, open the presets list and test a few options.
5. Set quality to High (I tested Medium and it didn’t look nearly as good).
6. Hit Generate.

Outcome you’re aiming for

A short performance shot where your clone looks like it’s actually delivering the line — not just a static face flapping.

Troubleshooting: the fixes that matter

• Clone doesn’t look like you: Add more selfie angles, avoid obstructed faces (hands, sunglasses, harsh shadows), and regenerate.
• Result feels “off” even with a good clone: Your still image is probably the weak link. Generate more options and pick a stronger “master” frame.
• Performance looks stiff: Change the motion preset. The default movement can look generic.
• Quality looks mushy: Use High quality (as shown in the video), and keep your source image sharp.
• Audio feels awkward: Try shorter audio segments (line-by-line like I did) so each clip is a clean, intentional shot.

FAQ

Do I need a lot of selfies?

In the video, I got better results by uploading many clear photos from different angles. More coverage gave a better clone.

Can I use any song?

You can upload an MP3 in the workflow shown. For the cleanest results, I split the song into shorter lines and generated one clip at a time.

What plan do I need?

In the video, I mention character creation required a paid plan at the time. Plan access and pricing can change — verify on the official site before you commit.