Make a listing video with your AI assistant

Create a finished video through the AgentOS Assistant or connected ChatGPT. The assistant can read a template, prepare its inputs and start production. Narrated storyboards add a specific review step before rendering.

1. State the listing, format and output

Give the exact address or MLS identifier, target format, language and whether you want narration. Mention an existing template if you have one. Otherwise ask the assistant to show suitable templates with their real requirements.

For ChatGPT, use @bsquare-agentos with an authorized connection. For the AgentOS Assistant, use the same request without the mention.

2. Start with a simple, supported recipe

The current bsquare.listing-fact-pills-reels template is a concrete option for a vertical photo-based video. It accepts 2–8 photos, supports English and Simplified Chinese text, runs for 17 seconds and does not require narration.

Copy this prompt for a first production:

@bsquare-agentos Make one finished vertical listing video for [EXACT ADDRESS OR MLS ID] using bsquare.listing-fact-pills-reels.

Read the current template and listing before proceeding. Use English display text, 2–8 authorized listing photos in a sensible viewing order, the verified price and address, and concise facts supported by the listing. Use my authorized public contact details only in the template's optional contact fields.

Keep this 17-second template without narration. Preserve the template itself and supply all of its current required inputs. Generate the final MP4, follow the returned operation to completion, and return the actual playback/download link and operation ID. Check the available output evidence and say clearly which aspects you could not inspect. Do not publish or send the video.

The assistant should use the current template definition, not rely on those example requirements forever. A missing template, insufficient usable photos or a required field needs to be resolved before production.

3. For narration, draft and review before rendering

Choose a template that declares narration, a language it supports and an available voice. The current bsquare.property-narration-tour recipe is a vertical five-scene tour with Simplified Chinese narration. It uses five scene photos plus an ending photo, and its narration should fit within 27 seconds to leave its three-second ending inside the standard 30-second production limit. It is not an English-narration template by default.

Use this staged request:

@bsquare-agentos Prepare a Simplified Chinese narrated tour for [EXACT ADDRESS OR MLS ID] with bsquare.property-narration-tour.

Read the current template, listing, usable photos and authorized voice choices. Use my available Profile voice if supported; otherwise ask me to choose an available voice. Draft five concise scene descriptions from verified facts and the actual photos. Keep the combined narration within this template's 27-second audio budget and retain its three-second ending.

Show the complete current text-and-photo storyboard in scene order, including the ending photo. Let me review or edit it before any speech synthesis or final video generation. Do not invent room features or reuse an approval for older text or photos. Do not publish or send anything.

For another language or structure, ask for a compatible template or an explicitly requested personal-template adaptation first. Do not assume that changing the prompt language changes a template's speech configuration.

4. Confirm the exact storyboard

Read each scene's words next to its photo. Check that details mentioned in the narration are present or supported by the listing. Review cropping and motion choices if the template exposes them.

When satisfied, reply:

I approve the current text-and-photo storyboard you just showed, including the ending image. Generate this version as the final MP4 using the selected voice. Return the video link and operation ID. Do not publish or send it.

Templates that declare scene-to-photo bindings require confirmation of the current review snapshot. If you change any words, photos, photo order or template details, review the new snapshot before generating. A technical review identifier alone is not your approval.

5. Wait for a real delivery and inspect it

The assistant should report a real operation state and return the finished artifact. “Template saved,” “inputs valid” and “job started” are not delivery.

Watch and listen to the actual MP4. Check pronunciation, caption timing, picture order and the contact ending. Photo pan and zoom are two-dimensional image motion, not newly filmed footage or automatically generated video scenes.

Revisions, failures and links

• For a content change, specify the exact change and keep the same template unless you want to redesign it. A new movie requires a new production.
• For an uncertain or failed job, ask the assistant to inspect the original operation and use its permitted recovery path rather than starting duplicate production.
• For an expired link, request a refreshed link to the existing result.
• Ask separately if you want the video sent or published. A creation request does not include distribution.

Current limits to plan around

The standard paid final-generation path accepts up to 30 seconds at 30 fps. Formats and resources depend on the selected template. GIF needs a GIF-capable template with its required link and alternative text; a normal MP4 request does not include an email GIF. Voice access and provider availability must be checked for the actual account.