Upload one headshot. Get an AI avatar of yourself. Generate videos of that avatar speaking your scripts — in your voice (cloned via ElevenLabs). Realtors who freeze on camera become content machines. Two separate credit pools (creation + per-video) make this a serious upsell layer on top of subscription.
Unlock the segment of realtors who refuse to film themselves.
Roughly 40% of realtors say "I'd post more video, but I hate being on camera." That's a massive segment we lose to camera anxiety. AI avatars solve it — same realtor, same voice, no actual recording. Script Generator writes the words → ElevenLabs clones the voice → Avatar speaks them. Done in 90 seconds. Realtor doesn't even open the camera app.
+ 40% segment
The "won't be on camera" cohort becomes addressable. Big TAM unlock.
2 credit pools
Avatar create (one-time per avatar) + Avatar video (per generation). Separate buckets, separate top-ups.
+ Script funnel
Funnels straight from Script Generator → Avatar speaks it. Already-built pipeline.
+ Tripod kit alt
For realtors who don't even want to use the AI Tripod, this is the alternative content engine.
Tech recommendation · F-070
Heygen API for v1 (ship in weeks). Higgsfield + NanoBanana for v2 (own the margin).
Don't build avatars from scratch in v1 — ship fast on Heygen's API while we validate demand. Once we have signal (volume of avatar credits being purchased), evaluate moving to Higgsfield + NanoBanana for in-house generation. The math: Heygen takes ~$5–15 per avatar minute; in-house can do it for $1–3 once we own the pipeline. Phase the build to validate first, optimize after.
Heygen API
v1 · Ship in weeks
Pros: Production-ready API, great voice clone quality, well-documented, scales out of the box.
Cons: $5–15 per minute generated — kills margin at scale.
Use Heygen for the initial rollout. Get to market fast. Validate that realtors actually buy avatar credits before committing engineering to in-house.
Higgsfield + NanoBanana
v2 · Own the margin
Pros: Per-minute cost drops 5–10×. Differentiated quality if we tune the model. Long-term moat.
Cons: Months of engineering. Quality risk during ramp.
Migrate to in-house once Heygen volume justifies the engineering investment. Run them in parallel for 60 days; cut over once quality matches.
Voice cloning (both v1 + v2)
ElevenLabs · paid tier
Same voice clone whether the avatar is rendered by Heygen or in-house. Lock in voice provider; swap out the rendering provider underneath. ~$5/mo per cloned voice.
Avatar creation flow · F-070 · F-071
3 steps: upload → generate → ready
First-time setup. Realtor uploads 1 headshot + records a 30-second voice sample (so we can clone). System processes for ~10 minutes. Done — they have a working avatar attached to their account. Burns 1 Avatar Create credit.
1
📷 Upload headshot
1 photo from Brand Profile. We use the best one we have.
→
2
🎙 30-sec voice sample
Read our short paragraph. ElevenLabs clones the voice.
→
3
⏱ ~10 min processing
Heygen renders the avatar + ElevenLabs trains the voice.
→
✓
✅ Avatar ready
Live in your account. Generate videos anytime.
Avatar video generation · F-072 · F-073
Pick a script → click Generate → 60 seconds later, video
The everyday flow. Realtor opens AI Avatar, picks a script (or generates one with Script Generator), clicks Generate. 60 seconds later, the video is rendered. Burns 1 Avatar Video credit. Then 4 destinations: download, schedule directly, send to Creative Studio for AI editing (captions, B-roll), or send to us as a project.
app.socialrealtr.com/ai-avatar
▲ SOCIAL REALTR
▦ Dashboard
✨ Creative Studio
▷ Script Generator
🤖 AI Avatar
📅 Calendar
▢ My Projects
◷ Brand Profile
🤖 AI Avatar
🤖 1 avatar create🎥 5 avatar video
SK
YOUR AVATAR
SK
Sarah · default voice
VARIATION
SK
Sarah · enthusiastic
+
Create another
Burns 1 create credit
📜 Script for the avatar to speak
SCRIPT · 60 SEC · TOM FERRY VOICE
Most agents lead with the granite countertops. Wrong move. When I tour a buyer through a $1.4M listing in West Surrey, I don't start with the kitchen. I start with the school catchment, the commute, and the fact that this neighborhood appreciated 12% last year…
Background
Length
~60 seconds to render. We'll notify you when it's ready.
📺 Latest video preview
PREVIEW
0:58
SK
▶
"Most agents lead with the granite countertops. Wrong move..."
1
2
3
4
5
1Credit balances in top bar. Two pools visible — Avatar Create (rare, 1 per realtor usually) + Avatar Video (each gen burns 1). Same pattern as Studio's credit bar.
2Multiple avatar variations. Same person, different "moods" or wardrobes. Each variation is its own avatar create burn. Power users will want a "default" + "enthusiastic" + "casual" set.
3Script slot funnels from Script Generator. Three sources: open Script Generator (with the funnel handoff from F-066), paste own, pick from saved. Same UX as the recorder's script intake.
4Generate button shows credit cost upfront. Burn-confirmation modal fires on click (per the credits mockup pattern). Never burn silently.
5Inline preview after generation. Watermarked PREVIEW until they pick a destination. Regenerate is free if they hate it within 60 seconds (1 free retry per generation — Heygen's policy).
Output destinations · F-073
Four paths after the avatar speaks
Same destination model as Creative Studio + Recorder output. One mental model across the platform — what you do with a finished asset is always the same four options.
📅
Schedule directly
Most common · publishes via your platforms
▢
Send as project
Our editor adds B-roll · burns 1 project credit
✨
Send to Studio
AI editor adds captions · burns 1 studio video credit
↓
Download
Use it however you want
Brand Profile integration
Headshot pool feeds avatar variations (F-052)
If the realtor has 5 headshots in Brand Profile, they can spin up 5 different avatars from the same person — different moods, outfits, settings. F-052 (headshot expansion) generates additional expression variations from each base headshot. Result: a library of "Sarah" personas the realtor can pick from per video.
This is why Brand Profile depth (F-021–F-030) and AI Avatar are linked. The richer the Brand Profile, the better the avatars feel.
Open questions for Trent
Heygen exact API tier. Their pricing scales by minutes generated. Need to estimate volume during build to pick the right tier. Recommend starting on metered billing, switching to a commit tier once we hit ~500 video minutes/mo.
Avatar Create credit count per plan. Mockup shows 1 free with Rising Star, 3 free with Top Producer. Confirm. Most realtors only need 1; power users want multiple variations.
Voice clone re-record cadence. Voices drift slightly with ElevenLabs over time. Should we prompt realtors to re-record their voice sample every 6 months? Recommend opt-in nudge, not mandatory.
Watermark policy. Mockup shows "PREVIEW" watermark before destination is picked. Should the watermark stay on if they pick "Download" without burning a final-render credit? Recommend NO watermark on full Download — they paid the video credit; let them have the clean asset.
v2 migration triggers. What volume of avatar minutes per month justifies switching from Heygen to in-house? Recommend ~2,000 min/mo as the break-even.
Restricted use language. Avatar of a real person with a cloned voice = important to spell out in our terms what realtors can/can't do with their own avatar. (Hint: they own it; they can't deepfake other people.) Legal review before launch.