Event Report: A Generative-AI Workshop on Short Videos for Corporate PR

Summary: Miyako de IT held a practical online course on using AI-generated short videos for corporate PR. About 20 participants worked through subject selection, prompt design, improving generated output, editing, and the production of a company-introduction video.

Start with the communication goal

The session began with the point that a generated clip is not automatically a PR video. Participants first identified what a viewer should understand: a product’s appeal, a place’s atmosphere, a seasonal message, or an image of a service in use. That decision gives prompts and editing a common direction.

Prompt design and iteration

Participants tested how a prompt changes a scene’s subject, composition, movement, and mood. The workshop treated generation as an iterative process: create a short result, identify what is not communicating, adjust the instruction, and compare again. The goal was not a single “correct” prompt but a repeatable way to improve output.

Explore upcoming Miyako de IT events →

Editing is part of the work

Corporate PR requires information that video generators often render unreliably. Captions, BGM, brand assets, company names, logos, and inquiry paths were added in editing. The workshop therefore positioned AI generation as one stage of a production workflow rather than a replacement for all production decisions.

Rights and practical caution

Commercial use is not a blanket yes/no question. Participants were reminded to verify each tool’s current terms and plan, the rights in materials used, and the requirements of the distribution channel. Accuracy-sensitive names and logos should be inserted after generation.

Event outline

The original event details are available on connpass.

Item	Detail
Event	Introduction to Creating Video with AI, Session 2
Date	June 30, 2026
Format	Online
Instructor	maki, freelance web designer in Kyoto
Participants	About 20
Theme	AI short video polished for corporate PR
Organizer	Miyako de IT / Netsujo Inc.

What to make with AI video

Product close-ups
A shop or facility atmosphere
Seasonal staging
An image of service use
A recruiting or workplace scene

Design a prompt for corporate PR video

Main subject

State what should remain visually central.

Background

Describe the location or setting.

Movement

Describe the action and pace.

Camera

Describe framing and camera movement.

Atmosphere

Describe light, mood, and tone.

Negative instructions

State what should not appear or change.

Do not treat a generated result as finished

Check that a product or subject does not change shape.
Check hands and faces for unnatural rendering.
Check text and logos for broken display.
Check that camera movement is not too fast.
Check that the first seconds communicate on social media.
Leave room for captions in editing.

Uses for product, venue, recruiting, and seasonal PR

Product PR

Use a focused visual of the product and its benefit.

Shop and facility introduction

Use the atmosphere of the place as the subject.

Recruiting PR

Show an approachable image of the work environment.

Seasonal campaigns

Adapt a communication theme to a season or timely message.

The in-session work with about 20 participants

Participants chose a subject, made prompts, reviewed outputs, and discussed the next editing decision:

1. The product or service to promote.

2. The intended viewer.

3. The one message the video should convey.

4. The visual subject and background.

5. Movement and camera work.

6. Elements the AI should not generate.

7. Text and music to add in editing.

8. The publishing destination and video length.

Sharing the process for a Netsujo company-introduction video

The session shared a production process from message design through generation, review, editing, and publishing:

1. Decide the video’s purpose and target audience.

2. Select visual elements from the company’s characteristics.

3. Design prompts for each scene.

4. Generate several video assets with AI.

5. Select usable assets and regenerate where needed.

6. Connect the assets in editing software.

7. Add captions, BGM, company information, and a CTA.

8. Make a final check of how the result appears on social media.

This article covers only what was shared in the session on 30 June 2026. Videos produced separately after that date are not counted as outcomes of this session. The process above used the organizer's own material; it is not a delivery record for an external client.

Keeping or discarding a generated asset

The table below is a practice example in which the editors arranged the review points covered in the session into decisions. It was not handed out at the session. The point of pairing each observation with the action it should trigger is to be able to reach the same decision again when you next face the same kind of asset.

What you observe in the generated result	Decision	What to do next
The subject changes shape partway through	Reject	Narrow to one subject, reduce the motion instructions, and regenerate
Hands or faces look wrong	Partial use	Cut only the seconds in which that person is not visible
Text or a logo is distorted	Do not use the distorted text	Replace the segment in which the distorted text appears. Instruct the model not to render text, and add verified text as a caption during editing
The camera moves too fast	Partial use	Choose the steadier segment and slow it down if needed
The first few seconds do not convey the subject	Rebuild the structure	Pick an asset in which the subject appears first
There is no room for captions	Partial use	Regenerate with a wider framing, or add margins above and below

Deciding your options other than "redo the whole thing" in advance keeps each judgement from drifting into an impression of "good" or "bad". Whether to keep a single asset depends on whether the information required by the destination platform comes across, not on how polished the footage looks.

A 30-minute practice exercise

The sequence below is a practice exercise written by the editors. It is neither what was carried out at the session on 30 June nor a handout from it. It is offered as a way to run one video end to end instead of starting with a company PV. The times are suggestions, and waiting time for generation varies with the service you use and how busy it is.

Time	What to do	Condition for continuing
0–5 min	Pick one subject: a product close-up, the feel of a space, or a seasonal treatment	Do not pick anything requiring multi-person dialogue, fine handwork, or accurate text this time
5–10 min	Write one line each for subject, background, motion, camera, mood, and exclusions	All six lines filled in, one line each. Do not write paragraphs
10–20 min	Generate, and judge each result with the keep-or-discard table above	Example stopping criterion: if three generations yield nothing usable, change the subject and go back
20–28 min	Join the segments you kept, then add captions, company details, and a call to action in editing	Put company name, prices, and logos in the edit, never in the generation
28–30 min	Check the first three seconds at the aspect ratio of the destination platform	If the first three seconds do not say what the video is, start again from asset selection

The figure of three generations is an example stopping criterion, not a record of three generations producing nothing usable at the session. What you take away is not a finished video but three things: the six-line prompt sheet, the keep-or-discard results, and the conditions that forced a redo. Keeping those lets you reach your own decisions again the next time you work on the same kind of subject. We have not measured how the number of generations or the production time changes.

What this article does not carry

This article does not carry the footage generated in the session, the original prompt text, or before-and-after comparison screens. Their absence here does not mean that they do not exist.

We have also not checked what participants produced afterwards or where they used it. We do not infer participants' business use or results from the fact that the session took place.

Cautions for companies using AI video

Check the commercial-use terms of the generation service.
Check whether generated output carries a watermark.
Check rights in input images and materials.
Avoid expressions resembling real people or celebrities.
Do not rely on other companies’ logos or characters.
Verify AI-generated text, prices, and company information.
Check the advertising rules of the social platform.

Do not stop at generating a video

Choose a subject that suits the purpose.
Choose a composition that is less likely to break.
Make prompts specific.
Review and improve the result.
Add accurate information and a route to act in editing.
Check response after posting to social media.

Find future learning opportunities through Miyako de IT on connpass.

Frequently asked questions

Can AI generation alone complete a corporate PR video?

In most cases, no. Captions, music, logos, company information, and a contact path still need to be added during editing.

What kinds of corporate PR videos are easier to create with AI?

Product close-ups, the mood of a store or facility, seasonal scenes, and an image of a service after use are comparatively easier to design. Multi-person conversations, exact text or logos, and intense motion are more likely to break down.

Can we put a company name or logo into an AI video?

Generated text and logos can be distorted. Add company names and logos that require accuracy during the post-generation editing stage.

Can AI videos be used commercially?

Commercial-use terms depend on the service and plan. Check the terms in force at production time, rights in source materials, and the rules of the advertising medium.

Next steps

The workshop’s emphasis was practical: use a small, communicable subject, test prompts, edit for accuracy, and review the result against the intended audience. Future Miyako de IT events are listed on connpass. Read the Claude Code and OpenClaw hands-on report and the Kyoto IT community overview.

AI short video workshop for corporate PR