Cinematic Avatar Reel: Full Workflow

What this workflow produces

A real estate hook of your avatar:

• You pick the outfit — a real photo or a generated image of yourself
• You pick the house — any screenshot or image of any property
• Nano Banana blends the two into one starting frame
• HeyGen Cinematic animates that frame with your avatar's voice

This is an image-first pipeline. You lock the exact starting frame before generating a single second of video — the model animates your frame instead of inventing its own scene.

What you'll need

• An outfit shot — a photo of yourself, or a generated image of you, in the outfit you want to wear on camera.
• A house photo — a screenshot or image of the house you want to stand in front of. Listing photos work great.
• Nano Banana — to blend the two images into a single starting frame.
• HeyGen with Cinematic avatar access and your trained voice avatar.
• Your generation prompt — pasted in at generate time (see The Prompt below).

💡 Reference quality is everything. The blended image becomes the literal first frame of your video — a crisp, well-lit house screenshot pays off in every frame that follows. The better the photo, the better the result.

1. Get the outfit shot

Take a photo of yourself — or generate an image of yourself — wearing the outfit you want on camera. Framing matters less than your likeness and the outfit reading clearly.

2. Screenshot or image of the house

Screenshot a photo or take an image of the house you want to be in front of. Listing photos are perfect. The better the photo, the better the result.One thing to watch: most listing photos are landscape, but your starting frame needs to be 9:16 vertical. You don't need to pre-crop — the blend prompt asks for a 9:16 vertical composition, so Nano Banana handles the reframe. Just make sure the part of the property you want behind you (the entrance, the staircase) sits near the center of the screenshot, so it survives the vertical crop instead of getting cut away at the sides.

3. Blend both in Nano Banana

Give Nano Banana the outfit shot and the house screenshot and have it blend your two images into one: you, in the outfit, standing in front of the house. This blended image is the starting frame for your video.The blend prompt from the Utah reel (the bracketed names reference the two uploaded images):

place the character [Generated image 1 (5)] standing at the top of the stone staircase of the property in [Generated image 1 (4)]. full-body composition, 9:16 vertical, camera several feet in front at eye level. keep the character's face and outfit exactly as in the reference. match the property photo's lighting and time of day, with realistic scale, natural contact shadows, and feet planted on the steps.

Plug-and-play version:

place the character [YOUR OUTFIT SHOT] standing at [SPOT, e.g. the top of the stone staircase] of the property in [YOUR HOUSE PHOTO]. full-body composition, 9:16 vertical, camera several feet in front at eye level. keep the character's face and outfit exactly as in the reference. match the property photo's lighting and time of day, with realistic scale, natural contact shadows, and feet planted on the steps.

💡Tip: The blend prompt deliberately describes the same shot the video prompt opens on — full-body, 9:16, camera a few feet in front at eye level. That way frame one of the video is your blend, and the model animates it instead of fighting it. The lighting-match and contact-shadow lines are what stop the result from looking pasted on.

4. Generate in HeyGen Cinematic

Take the blended image into HeyGen and follow this exact click path:Click Avatar → click CinematicDrop your blended image in as the referenceUpload your avatar as the voiceUncheck enhance promptMatch the aspect ratio to your reference — for this example it's 9:16Paste your promptClick generate

5. The result

You'll get a result like the reel at the bottom of this page — you, in the outfit, in front of the house, delivering the voiceover.

The Prompt

The exact prompt behind the Utah reel — pasted in at step 4:

Use @dylan for the voice and @character-sheet as the appearance and outfit for my avatar in the video.

15-second ultra-photorealistic cinematic real estate sequence. Shot on a Sony FX3 with a 35mm Sony G Master lens. Warm golden-hour lighting, natural contrast, shallow depth of field, subtle film grain, premium luxury real estate commercial quality.

The scene opens with the character standing at the top of the exterior stone staircase of a luxury Lake Tahoe estate. The camera is positioned several feet in front of him on a stabilized gimbal.

As the shot begins, he immediately starts walking naturally down the staircase while the camera smoothly tracks backward at a perfectly matched pace, maintaining a consistent full-body composition throughout the descent. His movement is confident, relaxed, and conversational, with natural eye contact directed toward the camera.

As he walks, he delivers the following line naturally without stopping or breaking stride:

"This is what just under ten million buys you here in Lake Tahoe."

His speech is paced like a high-end real estate presentation, with subtle hand gestures that feel authentic and unscripted. His facial performance remains confident, approachable, and engaging.

The camera movement remains smooth and perfectly stabilized, matching each step as it glides backward. There are no zooms, cuts, or sudden framing changes. The luxury home's entrance, landscaping, and architectural details remain visible behind him, reinforcing the scale and elegance of the property.

He finishes the final word just as he reaches the bottom of the staircase. The camera continues drifting backward for a brief beat while maintaining the same framing before the shot ends.

Ultra-realistic human movement, natural foot placement on each stair, believable body mechanics, cinematic camera stabilization, realistic facial animation, premium luxury real estate cinematography, consistent character appearance, seamless environmental continuity, Hollywood-quality photorealism.

[No music. No sound effects. Only natural synchronized dialogue.]

💡 HeyGen @ syntax lets you reference your saved avatar voice and character sheet directly inside the prompt. Swap @dylan and @character-sheet with your own bindings.

Plug-and-play template

Copy the template below and fill in the bracketed parts for your own reel. The structure mirrors the Lake Tahoe example: avatar bindings, camera spec, opening position, one continuous movement with a matched camera track, the spoken line, delivery notes, camera rules, an ending beat, and quality keywords. Leave the unbracketed text as-is.

Use @[VOICE_NAME] for the voice and @[CHARACTER_SHEET_NAME] as the appearance and outfit for my avatar in the video.

[DURATION, e.g. 15-second] ultra-photorealistic cinematic real estate sequence. Shot on a [CAMERA, e.g. Sony FX3] with a [LENS, e.g. 35mm Sony G Master lens]. [LIGHTING, e.g. Warm golden-hour lighting], natural contrast, shallow depth of field, subtle film grain, premium luxury real estate commercial quality.

The scene opens with the character standing at [STARTING POSITION, e.g. the top of the exterior stone staircase] of [THE PROPERTY, e.g. a luxury Lake Tahoe estate]. The camera is positioned several feet in front of him on a stabilized gimbal.

As the shot begins, he immediately starts walking naturally [MOVEMENT PATH, e.g. down the staircase] while the camera smoothly tracks backward at a perfectly matched pace, maintaining a consistent full-body composition throughout. His movement is confident, relaxed, and conversational, with natural eye contact directed toward the camera.

As he walks, he delivers the following line naturally without stopping or breaking stride:

"[YOUR EXACT SPOKEN LINE HERE]"

His speech is paced like a high-end real estate presentation, with subtle hand gestures that feel authentic and unscripted. His facial performance remains confident, approachable, and engaging.

The camera movement remains smooth and perfectly stabilized, matching each step as it glides backward. There are no zooms, cuts, or sudden framing changes. [ENVIRONMENT DETAILS, e.g. The luxury home's entrance, landscaping, and architectural details] remain visible behind him, reinforcing the scale and elegance of the property.

He finishes the final word just as he reaches [ENDING POSITION, e.g. the bottom of the staircase]. The camera continues drifting backward for a brief beat while maintaining the same framing before the shot ends.

Ultra-realistic human movement, [MOVEMENT KEYWORDS, e.g. natural foot placement on each stair, believable body mechanics], cinematic camera stabilization, realistic facial animation, premium luxury real estate cinematography, consistent character appearance, seamless environmental continuity, Hollywood-quality photorealism.

[No music. No sound effects. Only natural synchronized dialogue.]

💡 Tip: Match the movement to what's actually visible in your blended starting frame — if your frame shows a staircase, walk the staircase. Time the spoken line to the movement so the final word lands as the move completes, and swap the pronouns to match your avatar. Vague brackets give vague videos — mirror the detail level of the Lake Tahoe example.

Assets used (outfit shot + house reference)

Left: the outfit shot (step 1). Right: the house screenshot (step 2). The Nano Banana blended starting frame they produce (step 3) is shown below.

Why this workflow works

Five details that make this workflow deliver:

• The starting frame is locked before any video exists. Blending in Nano Banana settles your likeness, the outfit, and the location in one still — the video model animates your frame instead of inventing a scene.
• The house screenshot drives the whole look. Lighting, framing, and detail in the reference carry straight into the generated footage.
• Your voice avatar handles delivery. Uploading your avatar as the voice keeps the read consistent with every other video on the account.
• Enhance prompt stays unchecked. HeyGen won't rewrite your prompt behind your back — what you paste is exactly what drives the generation.
• Aspect ratio matches the reference. A 9:16 reference generated at 9:16 means no cropping, stretching, or letterbox surprises.

Customize this for your scene

To adapt this for your own reel:

1. Swap the outfit shot. New photo or newly generated image of yourself.
2. Swap the house. Any screenshot works — mansion, cabin, penthouse, wherever you want to stand.
3. Re-blend in Nano Banana. Every new outfit + location pair gets its own starting frame.
4. Rewrite the spoken line in your prompt. Keep the structure, change the hook to match the property.
5. Keep enhance prompt unchecked and the aspect ratio matched unless the destination format changes.

✅ Ready to ship when: the generated video shows you — face and outfit matching your blended frame — standing in front of the right house, the voiceover plays in your avatar's voice, and the 9:16 frame is clean with nothing cropped or warped.

The reel

The finished cut that drives traffic to this guide:

Optional: The post pass

The raw HeyGen output is postable as-is — everything here is polish, not rescue. The Utah reel got a light finishing pass before publishing:

1. Trim to the action. Cut any dead air before the first step and after the final word so the hook lands immediately.
2. Lay a music bed. Keep it low — the dialogue is the reel; the music just adds production feel.
3. Add captions. Standard short-form treatment — most viewers hit the reel muted.
4. Export at 9:16, high bitrate. No re-cropping needed; the frame was locked back at the blend step.

💡 Tip: If the generation is clean, this whole pass is 15 minutes. If you find yourself fighting the footage in post, regenerate instead — a better starting frame beats a heavier edit.