A realtor we have been working with for over a year once said something that struck us: “I have 40 properties in the pipeline at the same time, and I only make videos once a quarter when I finally carve out a day for a cameraman.” The agency is small—not a federal network, but a team of six agents in one city—and hiring a videographer for each listing was both expensive and physically impossible due to timelines: by the time you agree on a shoot, the property might already be taken off the market.
The task sounded simple but was not solved on the first try: to ensure that each listing—about 30-40 properties a month—had its own short video, without the involvement of a cameraman or editor for each individual clip.
First Attempt: Photos of the Property and the Agent's Voice
The material for the video consisted of photos that agents already took on their phones during property viewings, plus the description text they wrote for the listing. We didn’t have to gather anything new—this was the main argument in favor of trying it at all. The description text served as the source, which was then processed by AI into a script for voiceover.
For the voice, we chose not a library option, but the cloning of the leading agent's voice from the agency—this decision was made immediately because, in real estate, personal trust in a specific person means more than a beautiful faceless voiceover. Clients had recognized this agent by voice during viewings for years, and it made sense to preserve that in the video.
The first batch of five clips was technically workable, but there was a problem we hadn’t anticipated: the property photos were taken with phones in a mix of horizontal and vertical shots, and some of them, when edited into a vertical format for social media, turned into poorly cropped panoramas of the room—exactly the wall with the window that sold the apartment was lost.
What We Fixed After the First Month
We sorted out the image format separately from the video format: we adjusted the aspect ratio so that horizontal photos of rooms were not awkwardly cropped for vertical clips but were composed with respect to the frame—without stretching or cutting out important parts of the interior. This was a project-level setting, not a manual adjustment for each individual photo, which was critical given the volume.
The second point was the voiceover. The first clips sounded too upbeat for an audience looking for family housing and wanting a calm, trustworthy tone rather than an advertising push. We switched the narrative pace from standard to calm—and this immediately changed the feel of the clips, even though the text changed minimally.
Thirdly—before the voiceover of numbers and addresses in the descriptions (area, floor, price, house number), they sometimes sounded awkward when the voice engine read them as they were. We enabled text normalization before voiceover, which correctly converts numbers into words for spoken language—this was a minor detail, but it was precisely because of this that the first drafts sounded a bit mechanical, and the agents immediately noticed it during playback.
Editing and Subtitles According to Audience Habits
We also had to spend time on subtitles. Part of the agency's audience watches the clips on Telegram with the sound off while on the go, so subtitles are not an option but a mandatory part. By default, the subtitles on a couple of clips covered the address of the property, which the agent had specifically shown in close-up in the photo. We decided to manually move the safe zone of the subtitles—this draggable area, where the text is placed, does not overlap with important parts of the frame. We set it up once for the vertical clip format and have since simply reused it.
The clips became short and uniform in structure: a few seconds on the property from the outside, a walkthrough of the rooms with key characteristics, and at the end—a call to schedule a viewing. Here, the binding of the call to action at the project level was useful: all agency clips have a unified and consistent call to action leading to scheduling a viewing, without each agent having to come up with their own wording from scratch.
Publication and Initial Results
The clips are published in the agency's Telegram channel and in the VK group, with a separate description text for each platform—in Telegram, it’s a short caption with the address and price, while in VK, it’s a slightly more detailed description with specifics. Automatic publication works for each property immediately after the clip is ready, without manual file transfers between agents and the SMM specialist, who is simply not a separate staff member in a team of six.
In the first three full months after the adjustments, the agency managed to publish videos for about 30-40 properties monthly—compared to four or five clips per quarter before. The view count is modest; it’s not federal reach, but the number of requests for viewings has noticeably increased specifically for those properties that have videos compared to those without—agents themselves began to notice this difference and asked for videos even for properties they previously considered “too ordinary to bother with.”
What We Would Do Differently from the Start
If we were to do it again, we would immediately set the image aspect and the safe zone for subtitles before the first batch, rather than after complaints from agents about cropped frames. The test batch of five clips on real, rather than showcase properties, proved its worth—it was precisely this that revealed issues with frame format and pacing that were not visible on paper.
The main takeaway for us is not in the technology, but in the fact that for such volume—40 properties a month—the only working approach is when video is integrated into the existing process of the agent (they are already taking photos and writing descriptions), rather than adding a separate step. As soon as video became a byproduct of regular work, rather than an additional task, the volume increased on its own, without pressure from above.