Job Description:
We're a U.S.-based AI video production company creating marketing videos for real estate agents — property tours, agent-branded educational content, listing promos, and market updates. We've built a working storyboard platform and have API-based image generation, voice cloning, and video generation already running. Now we're adding ComfyUI to our stack and need production-ready workflows.
What We've Already Built:
Storyboard app ) that takes scripts → generates shot lists → produces images → shot-by-shot review UI
Image generation via FLUX Kontext Pro with agent identity preservation from reference photos
Minimax TTS voice cloning for agent voiceover
Active accounts with Kling, Pika, Runway for video generation
342-cut training dataset from professionally analyzed real estate videos
RunPod account (provisioned, not yet deployed)
What We Need — Workflows:
We're looking to purchase, commission, or adapt ComfyUI workflows for the following production tasks. If you've built workflows for AI influencer creation, virtual models, or consistent character generation — that's the exact same technical pipeline we need, just applied to professional real estate content instead of social media. The identity preservation, LoRA training, video generation, and lip sync challenges are identical.
1. Agent Identity Preservation (Most Important)
Real estate agents must look consistent across 15-25 shots per video — different poses, scenes, lighting, outfits. We need:
LoRA training workflow from 15-30 reference photos of an agent
Consistent generation of that agent in diverse real estate scenes (kitchens, living rooms, front of houses, offices, neighborhoods)
Full-body consistency, not just face matching
We're evaluating LoRA, InfiniteYou, IP-Adapter, and InstantID — show us what works best in your experience
We'll be building a library of 50+ trained agent LoRAs over time
2. Image-to-Video with Camera Movement
Converting approved still images into 3-5 second video clips:
Controlled camera motion: slow dolly, pan, zoom with parallax, push-in
Subtle environmental dynamics: light shifts, curtain movement, foliage
Temporal consistency — no flicker or artifacts
We're targeting WAN 2.1/2.2 for self-hosted generation. If you have WAN workflows, we want to see them.
3. Lip Sync
Audio-driven lip movement on AI-generated talking-head shots:
Must sync naturally with voiceover audio
Phoneme-accurate for frontal and three-quarter angles
Natural jaw movement, blinking, head movement
LatentSync 1.6, MuseTalk, or your preferred solution
4. Real Estate Scene Enhancement
Virtual staging workflows (empty rooms → furnished)
Property photo enhancement with cinematic camera movement
Architectural visualization and interior design generation
Consistent lighting and style across a batch of 15-25 images per video
5. Brand Content Templates
Talking-head agent generation with caption/lower third placement
Neighborhood and lifestyle B-roll generation
Data visualization scenes for market update videos
Repeatable template workflows where we swap agent LoRA and script content
Deliverable Requirements:
All workflows must be delivered as:
Editable ComfyUI JSON (not compiled or locked)
Written documentation explaining node connections, key parameters, and how to modify
Model/checkpoint list with download sources
Example inputs and outputs demonstrating the workflow in action
VRAM requirements and recommended GPU specs
Two Ways to Work With Us:
Option 1 — Workflow Delivery:
Sell us existing workflows or build custom ones for the tasks above. Tell us what you already have, what you'd build, and your pricing. We'll evaluate based on output quality and documentation.
Option 2 — Consulting + Workflow Delivery:
If you're experienced enough to also help us get set up — walking us through the ComfyUI ecosystem, advising on model selection, doing screen-share setup sessions, helping us evaluate workflows from other sources — we're interested in that too. We're technical (we build with AI daily, use Claude Code and Cursor, manage our own servers) but we're new to ComfyUI specifically. A few hours of consulting on top of workflow delivery goes a long way.
Either way, show us what you've built.
Long-Term Opportunity:
We produce content daily and are scaling to videos per month. The right person becomes our go-to workflow creator:
New workflows as models improve (WAN 3, better lip sync, new identity techniques)
Adapting workflows for new content types
Updates when ComfyUI or models release breaking changes
Ongoing hourly work at a rate we agree on together
What We Want to See From You:
Portfolio of ComfyUI workflows — screenshots of node graphs, output samples, video demos, GitHub repos. No portfolio = no review.
Identity preservation results — show us a person (real or AI-generated) maintained consistently across 5+ scenes with different poses and environments. What technique did you use?
Video generation samples — image-to-video with controlled camera movement. Smooth motion, real-world scenes.
Lip sync demo if you have one.
Which of our 5 needs can you address with existing workflows? What would you build new?
Your proposed pricing and engagement structure.
If your background is in AI influencer workflows — character consistency, LoRA training, video generation for social media personas — we want to hear from you. The technical pipeline is the same. Tell us how your existing work maps to our real estate use case.
Contract duration of 1 to 3 months. with 30 hours per week.
Mandatory skills: comfyui, LoRa, AI Image Generation, Machine Learning, Computer Vision, runpod, Stable Diffusion, AI Video Generation, Generative AI, Python