DEV Community: 佐藤光

Design a Blind Review for AI Video Before You Generate the First Clip

佐藤光 — Mon, 26 Oct 2026 12:21:16 +0000

A model name can influence an evaluation before the video starts. Hiding that name is useful, but it does not fix unequal prompts, retries or selection rules.

This is an untested evaluation protocol adapted from FablessBridge's production notes. We have not generated or scored clips for this proposed comparison, and there is no winning model or measured improvement to report.

1. Freeze the task and selection rule

Use a narrowly observable task. The source proposes one person opening a box on a desk and then removing their hand. Check instruction adherence, hand–box contact, character consistency and completion of the action separately from visual appeal.

Before generation, record duration, aspect ratio, audio, input image, model version and access route. Document settings that cannot be matched. Decide whether you will compare the best output from each model or every output from a fixed number of attempts. Do not quietly grant one model extra retries.

2. Separate the identity map from the review package

Label clips A, B and C. Keep the model-to-label mapping separate until scoring is complete. Inspect filenames, player labels and surrounding descriptions for identity clues. A visible service watermark makes blinding incomplete; record that limitation.

If an evaluation copy needs cropping or other processing, preserve the original and document the changed area. Processing can change the very quality or framing being judged. Record presentation order and rotate it so the same model is not always shown first.

3. Use an observation sheet, not a single preference score

This blank template turns the proposed checks into fields that another reviewer can reuse. It contains no test results.

Field	What to record
Clip label and timestamp	Which output and moment the observation concerns
Instruction adherence	Whether the requested action is visible
Contact and geometry	Hand–box contact and changes in object shape
Character consistency	Whether the person stays consistent across the shot
Action completion	Whether the hand actually leaves the box
Preference	Mood, pace and color, kept separate from defects
Review limitation	Identity clues or prior knowledge of production conditions

Save individual assessments before the group discussion. If a reviewer changes their mind after discussion, keep the original assessment as well. The source's examples of useful observations include a box retaining its shape and fingers appearing to increase when a hand moves away; these are illustrative observations, not findings from a completed test.

4. Reveal the model, then assess delivery constraints

After scoring, open the mapping and add attempt counts, wait time, cost and available features. Explain how many failed outputs were excluded. A clip preferred visually may still be unsuitable for the project's deadline or budget.

Report the result within its task and conditions. A comparison of one box-opening task cannot establish an absolute ranking of video models.

Before using this protocol

Check that the selection rule was fixed in advance, unmatched settings are visible, original files remain available and individual assessments have not been overwritten. These checks make the proposed review easier to inspect; they do not prove that it is unbiased or that a particular model is best.

If you are planning a business video, define the required actions and delivery constraints before choosing a model. You can explore FablessBridge's production examples, visit the official website, or discuss a production brief. Business contact: fabless-bridge@shian.space.

Source: FablessBridge's proposed blind-evaluation workflow. This English adaptation was written using AI from that source. No experiment was performed for this article.

Track AI Generation Costs Without Confusing Estimates, Jobs and Invoices

佐藤光 — Sun, 25 Oct 2026 05:52:07 +0000

A completed video tells you which take you used. It does not tell you how many attempts you paid for, whether an unfinished request is still running, or whether an estimated cost has been reconciled with an invoice.

FablessBridge's production-ledger guide separates those questions. Here is a reusable recording procedure based on that guide. This post describes a workflow; it does not report a newly executed integration or cost benchmark.

1. Give every request its own row

Record the project, attempt ID, model, access route, input asset, output conditions and submission time. Five seconds at one resolution and a longer clip at another resolution are different conditions, even if the model name matches.

For a retry, create a new row and relate it to the attempt being corrected. Do not overwrite the earlier request: that would hide how many attempts were made for the same purpose. Keep API keys and authentication headers out of the ledger.

2. Separate the estimate from the bill

An estimate needs its pricing formula, conditions, currency and the date you checked it. The billed amount needs a separate reconciliation against the provider's usage statement.

If that reconciliation has not happened, leave the billed value blank and mark it unconfirmed. A blank amount is not a zero-cost generation. Similarly, credits supplied through a web product are not evidence of a model's general API price.

The source links to a drip-video comparison record, where estimated Higgsfield costs and free Kling Web credits are distinguished, and an API prototyping record. Those are production references, not price promises for a new run.

3. Track the request and the asset separately

Use request states such as not submitted, accepted, processing, succeeded, failed and unknown. Separately record the output location and whether the asset was adopted or rejected.

A timeout is not proof that generation failed. Before sending an unknown request again, check whether the existing request can be queried. A failed status is also not sufficient evidence of no charge; billing reconciliation answers a different question.

4. Keep rejected attempts in the total

The source uses this fictional example. Assume identical model, route and five-second output conditions for all three rows:

Attempt	Estimate	Generation	Billed amount	Selection
01	JPY 20	Succeeded	Unconfirmed	Rejected
02	JPY 20	Succeeded	Unconfirmed	Rejected
03	JPY 20	Succeeded	Unconfirmed	Adopted

The estimated total is JPY 60. The actual bill is still unconfirmed. There is one adopted clip, but the two rejected attempts remain part of its production history. These figures illustrate bookkeeping only; they are not actual charges or a provider's unit price.

5. Review what your ledger can establish

Before using the record for the next project's budget, check that each retry has a distinct row, the output conditions are recorded, estimates and reconciled charges remain separate, and adoption has not erased rejected attempts. Keep the spending cap and retry rules alongside the ledger.

This makes it possible to explain what was tried, what was retained and which costs are still unknown. It does not, by itself, prove a saving or predict a new project's final cost.

From an experiment to a video brief

When commissioning a product video, bring the intended output and revision scope into the discussion. A record of the attempts helps explain which generation or editing decisions remain open.

Read the original Japanese guide, browse FablessBridge production examples, or discuss a video brief. Our business contact is fabless-bridge@shian.space.

Disclosure: This article was prepared with AI assistance from FablessBridge's published guide. The three-row ledger is fictional; no new API requests, invoice reconciliation or cost measurements were performed for this post.

Build a Japanese TTS Test Script Before Comparing Voices

佐藤光 — Thu, 22 Oct 2026 12:06:02 +0000

A pleasant greeting is a weak test for a narration system that will later have to read product names, dates and emotionally specific lines. Before comparing TTS outputs, define what the test sentence is meant to reveal—and what its correct reading should be.

This workflow adapts FablessBridge's published guide to building Japanese comparison scripts. It is a reusable review procedure, not a completed benchmark. The example sentences below have not been generated or scored for this article.

1. Build four small test cases

The source proposes four categories. Keep them relevant to the intended production rather than filling the script with unusually difficult words.

Case	Illustrative Japanese input	What to inspect
Everyday sentence	窓を閉めてから、こちらの席へお座りください。	Phrase boundaries and natural pauses
Proper noun	架空の商品「灯ノート」をご紹介します。	Whether the fictional product name is read as 「ひのーと」
Numbers	受付は10月21日、午前9時からです。	Date, time and missing numbers
Emotion	間に合ってよかった。ずっと待っていたんです。	Relief, and whether the acting becomes too strong

These are explanatory fixtures from the source, not quoted production audio. Substitute the words and situations your application actually needs. Put exceptional edge cases in separate tasks so they do not silently replace your real use case.

2. Freeze the intended reading before listening

A name can have more than one plausible reading. Write down the intended one before judging the output, rather than changing the answer to fit what a model produced. For numbers, establish whether the context calls for individual digits, a quantity or a date.

Keep the display text separate from the text passed to speech generation. A pronunciation aid should not accidentally become the subtitle. Link the two fields with the same line ID.

For example, a minimal record for the fictional product-name case could be represented as:

{
  "line_id": "proper_noun_01",
  "display_text": "架空の商品「灯ノート」をご紹介します。",
  "speech_text": "架空の商品「ひのーと」をご紹介します。",
  "intended_name_reading": "ひのーと"
}

This JSON is an illustrative representation of the guide's fields, not an executed API request or a file from a measured run. The guide links to a separate 14-line production record where display and speech text were kept apart.

3. Record the conditions, not just the model name

The same sentence spoken with different voices, languages, speeds or performance instructions is not the same comparison condition. Keep those settings and the number of generations with the outputs. If providers do not offer the same voice, state that difference instead of implying it was controlled.

A review record should make it possible to answer:

Which model and selected voice produced the file?
What language, speed and performance instructions were used?
Which input line and intended reading apply?
How many generation attempts were made?
Is this the original audio or a processed version?

4. Keep evaluation dimensions separate

Review textual accuracy, phrasing, voice quality and acting separately. A voice can be appealing while misreading the product name. A single overall score hides whether the next action is to correct the reading, adjust a pause or choose another voice.

The source does not prescribe a numerical scale or report a winning model. Use the categories to make the required correction identifiable, rather than presenting an unsupported ranking.

5. Compare originals first, then the finished mix

Retain the audio immediately after generation. If you adjust volume or edit pauses, record the processing and distinguish that version from the original during review.

Finally, check intelligibility with the actual subtitles and background music. An isolated voice test and the completed video answer different questions. These four sentence types are a starting point for your own application, not a comprehensive test of every speech capability.

A practical handoff for a narrated video

When commissioning a Japanese product video, provide official spellings and intended readings separately, especially for names and numbers. Agree on the required expression and retain the script and audio version being approved. This gives the editor a concrete basis for narration review.

Original Japanese guide · Related 14-line production record · FablessBridge production examples

For a video that combines narration, captions and music, discuss your brief or email fabless-bridge@shian.space.

Disclosure: This article was prepared with AI assistance from a published FablessBridge guide. No new TTS generation, scoring or model benchmark was performed for this post.

Measure the Event, Not Just the Clip: A Test Protocol for AI Video Timing

佐藤光 — Wed, 21 Oct 2026 12:03:18 +0000

A five-second output setting and a prompt that says “turn toward the camera at two seconds” describe different requirements. The first sets the clip duration. The second asks for an event inside it.

Before comparing video generators, define what counts as that event and how you will record it. This is a test-design workflow adapted from FablessBridge's published note. The example task has not been run, and the numbers below are arithmetic examples—not measured model results.

1. Define the event before writing the score

“Turns toward the camera” can mean the head starts moving, the face reaches a side view, or the gaze reaches the camera. Those moments produce different timestamps.

Pick the observable condition first. If completion is what matters, describe the completed state in the prompt and use that same condition during review. Record the requested output duration separately from the event timestamp. A dedicated timing control, where available, is a separate input from an ordinary sentence in a prompt.

2. Change one condition

An unexecuted example from the source uses a single action: a person turning their face toward the camera from the same starting image. Compare a prompt without a time instruction against one with it. Keep the starting image, framing and requested clip length consistent.

Do not change the camera move and the action while also adding timing instructions. Otherwise, the comparison no longer isolates the timing instruction. Even if a seed can be fixed, retain the actual settings and outputs rather than assuming identical results.

Before submitting any jobs, choose the number of trials, cost cap and stopping conditions. This article does not ask you to generate a batch or claim that one model will follow the timing reliably.

3. Keep a record that allows “unfinished”

The source proposes these fields; a CSV header is a convenient way to preserve the same structure:

trial_id,input_conditions,requested_time_s,observed_start_s,observed_completion_s,reason_not_judgeable

Keep the model, provider route, date and source version with the record. Use the same observation rule across clips. If the action never finishes, mark it unfinished instead of choosing a convenient intermediate pose as completion. An unjudgeable result needs a reason; it should not silently disappear from the comparison.

4. Inspect the unedited output

Observe the original clip before trimming or speed changes. Watch the motion at normal speed, then step through the relevant frames to locate its start and completion. If only an edited timeline survives, do not use its timestamps as direct evidence about the original prompt.

For a completed action, the difference is:

timing difference = observed completion time - requested completion time

A requested time of 2.0 seconds and an observed completion at 2.4 seconds would give +0.4 seconds. That is an illustrative calculation, not a measured clip.

5. Separate timing accuracy from whether the shot is usable

A timestamp alone cannot tell you whether the action is coherent or visually natural. Record whether the action completed and whether its appearance is usable, alongside its timing difference. One close result does not establish reliable behavior across a model, provider route or later version.

For a shot that must match a musical beat or a caption, editing can also be part of the solution. Keep “the generator met the requested time” separate from “the editor aligned the shot afterward.” Both can produce a usable sequence, but they describe different evidence.

Put the acceptance condition in the brief

When commissioning a product or service film, identify which action must coincide with a caption or soundtrack cue. That gives the production team something concrete to review and helps decide whether timing should be addressed during generation or editing.

Original Japanese procedure · FablessBridge production examples

To discuss coordinated motion, captions and sound for a video, send a production brief or email fabless-bridge@shian.space.

Disclosure: This article was prepared with AI assistance from a published FablessBridge procedure. No new generation experiment or performance benchmark was performed for this post.

Make Japanese TTS Revisions Traceable: Separate Captions, Readings and Audio Files

佐藤光 — Tue, 20 Oct 2026 12:00:16 +0000

A pronunciation correction should have a clear scope: which line changes, which audio file is replaced, and which subtitle must stay intact.

In a FablessBridge production run on October 16, 2026, we generated 14 Japanese narration lines as separate files with Gemini 3.8 Flash TTS. The saved files total approximately 44.72 seconds. That is evidence of the saved outputs and their durations, not a benchmark of voice quality or an end-to-end video runtime.

1. Keep two text fields linked by one line ID

Use one version for the subtitle and another for speech. The subtitle keeps the intended spelling, kanji and official product names. The speech version can express difficult readings in hiragana or katakana.

For example, a fictional announcement such as “新サービスの受付は10月21日からです” can keep its written date on screen while the speech text specifies its pronunciation. This is an explanatory example, not a quotation from the recorded narration.

The important invariant is the line ID: changing the reading must not silently replace the caption. Review both fields together when the meaning changes; review only the reading field when correcting pronunciation.

2. Make the line the regeneration unit

Generate and save each line separately, then assemble the files in the editor. A correction can target that line instead of regenerating the entire narration.

This does not guarantee a seamless result. Short files can differ at their boundaries in breath, volume or delivery. Review the corrected file with the preceding and following lines, not only in isolation.

3. Hold the voice instructions constant

The documented run used the model ID gemini-3.8-flash-tts and a shared direction: calm advertising narration, natural Japanese, restrained intonation and a slightly brisk pace. It received WAV output and converted that to MP3 for editing.

Keep the original WAV alongside the editing copy. When something sounds wrong, compare the original, converted file and assembled timeline. This helps distinguish a generated-output problem from a conversion or editing problem. Changing the model, direction and text at once makes that distinction harder.

4. Validate the handoff before accepting a revision

A reusable review sequence is:

Match the requested correction to its line ID.
Check that the speech text contains the intended reading and the caption still has the approved spelling.
Listen for omitted words, names, numbers and sentence endings.
Join the line to its neighbors and check pauses and volume transitions.
Review the sequence with the actual visuals, captions and BGM.
Keep the approved file and its original output so later edits have a clear reference.

The source records 14 saved files and their combined duration. It does not claim that these checks produced a quantified improvement in intelligibility. The run also had a $0.50 cap and no automatic retries; that cap is a workflow limit, not a quoted production price or a verified invoice total.

What this changes in a production brief

If you are commissioning a narrated product video, provide official spellings and intended readings separately. Ask how a one-word correction will be identified, replaced and checked in the final mix. This gives the review a concrete object instead of a request to “make the voice better.”

Original production record · FablessBridge production examples

For a service explainer that needs coordinated narration, captions and motion, discuss your brief or email fabless-bridge@shian.space.

Disclosure: This article was drafted with AI assistance and adapted from FablessBridge's published production record. It reports that record rather than a new independent model test.

A Seven-Stage Workflow for Completing an AI-Assisted Video

佐藤光 — Wed, 14 Oct 2026 16:46:18 +0000

A folder of impressive clips is not yet a film. The sequence, text, sound and delivery format still need to work together.

This implementation note records the inputs, failure modes and checks needed to make the workflow reviewable.

Repair the actual failure

A scissors problem was addressed in the source image. A box handoff was split into starting and ending states. An HTML animation jitter was traced to layout sorting on every frame and changed to a one-time sort. These are different failure modes; repeated generation would not fix the layout code. Locate the defect before choosing a correction.

Approve work in stages

We worked through brief, representative images, motion, HTML text and animation, music and effects, review, and export. Image candidates established the cast and visual direction before motion. Exact on-screen text was handled separately. Stage approvals made it possible to identify whether a defect belonged in an input image, the generated movement or the edit.

Use one complete project as the unit

Our fictional Yoine launch sample is a 76-second film about a smart eye mask before manufacturing. The record includes 68 generated images, 37 clips and seven versions. Its story depicts a founder testing ideas with pages and ads; it is a fictional sample, not evidence of a live campaign or sales outcome.

Checks to carry into the next project

Approve representative images before generating all the clips.
Locate whether a failure belongs in input, motion or editing.
Review each delivery ratio as its own composition.

Keep a reproducible record of the configuration and inspect the final exported artifact. An automation success message is not a content review.

Watch the production examples

These assets come from the documented project. Fictional in-house samples illustrate our process; they are not evidence of client sales results.

Watch this production clip (MP4)

Source and production inquiries

Adapted from the original FableBridge production note (Japanese). Model specifications, estimates, prices and schedules are records from the source date, not guarantees for a new project. The source includes further conditions and references.

FableBridge, operated by Shian Co., Ltd. in Tokyo, produces product and service videos, landing pages and advertising campaigns. Share your audience, placement, preferred length and timeline through our consultation page or email fabless-bridge@shian.space. Explore our work on the official website.

Review AI Video in Five Stages Before Polishing Details

佐藤光 — Wed, 14 Oct 2026 16:40:33 +0000

When a first draft feels wrong, decide whether the concept is wrong before spending time polishing its details.

This implementation note records the inputs, failure modes and checks needed to make the workflow reviewable.

Check continuity and exactness separately

We used 24 storyboard images and 23 Kling motion intervals. Static logos improved exactness but weakened movement, so the edit had to recover continuity while retaining correct text. This is a useful split in review: inspect how shapes connect across the cut, then inspect product words and marks. One attractive frame cannot establish either across the whole sequence.

Show the solved state

The storyboard needed clearer scenes of the problem being resolved, including the 10-15 and 25-30 second ranges. A film can describe a problem accurately while failing to show the benefit. Review the storyboard for the intended change in the viewer's understanding before generating all the motion.

Agree on the direction first

For a 30-second Ekusid service video, the first three concepts used people. The chosen direction moved to a people-free paper-and-UI treatment. That is a concept decision, not a minor visual correction. A client should say which audience and message the draft misses, then identify the parts worth preserving.

Checks to carry into the next project

Separate a direction change from a detail correction.
Check whether the benefit is visible in the storyboard.
Give feedback with a time range and a reason.

Keep a reproducible record of the configuration and inspect the final exported artifact. An automation success message is not a content review.

Watch the production examples

These assets come from the documented project. Fictional in-house samples illustrate our process; they are not evidence of client sales results.

Watch this production clip (MP4)

Source and production inquiries

Use Claude Code to Orchestrate a Bounded Video Workflow

佐藤光 — Wed, 14 Oct 2026 16:40:06 +0000

Make it more natural is a creative intention. A timestamp and a concrete defect make it a reviewable production instruction.

This implementation note records the inputs, failure modes and checks needed to make the workflow reviewable.

Make feedback local

We compared old and new versions and gave notes with seconds: what happened, what should remain, and what should change. If an instruction could be read in several ways, ask the coordinator to restate its interpretation before execution. This helps avoid a broad rewrite when only one cut or one line of narration needs attention.

Bound the paid work

For the fictional 76-second Yoine sample, the generation estimate was about USD 23 against a USD 25 budget, excluding labor and Claude costs. A script should estimate before submitting, record each receipt and stop at the agreed limit. A network timeout should not trigger an unexamined paid retry. Keep costs attributable to a particular attempt and version.

Give the coordinator a defined role

We used Claude Code to organize requirements, call image and video APIs through scripts, build text and motion in HTML, arrange audio and export versions. It was the coordinator of those steps, rather than the video generation model itself. People still chose the creative direction and judged the output.

Checks to carry into the next project

Define deliverables and a spending limit.
Use timestamps and specific defects in revision notes.
Name the person who will watch and listen to the final export.

Keep a reproducible record of the configuration and inspect the final exported artifact. An automation success message is not a content review.

Watch the production examples

These assets come from the documented project. Fictional in-house samples illustrate our process; they are not evidence of client sales results.

Watch this production clip (MP4)

Source and production inquiries

Assemble 24fps Clips into a 30fps Remotion Composition

佐藤光 — Wed, 14 Oct 2026 15:23:58 +0000

Generating the shots was one stage. Making the product name, transitions and sound land on the intended frames was another.

This implementation note records the inputs, failure modes and checks needed to make the workflow reviewable.

Place sound on the same timeline

Music and effects were placed at frame positions, with fades where needed. The final mix in this record measured about -17.5 LUFS and -2.7 dB peak; these are project measurements rather than universal loudness targets. Review speech and music together after export, as a correct source track does not establish a correct final mix.

Keep copy and motion editable

We used Composition for the complete timeline and Sequence for the cuts, with exact text and diagrams rendered as code. Letter animations were offset by three frames. Where clip duration differed from the allotted cut, playback rate was adjusted from the source duration and cut length. Preserve the approved wording outside the generated footage so revisions do not require a new video generation.

Define the output timeline

Our fictional coffee sample uses five cuts in a 30-second film. Input clips were 24fps at 1080p; the output composition was 30fps and 900 frames, exported as H.264. The recorded environment was Node 24, Remotion 4.0.532 and Apple Silicon. Fixing the duration and frame rate first made the other timings explicit.

Checks to carry into the next project

Fix duration and fps before timing the layers.
Keep approved copy in an editable layer.
Inspect the exported file with audio enabled.

Keep a reproducible record of the configuration and inspect the final exported artifact. An automation success message is not a content review.

Watch the production examples

These assets come from the documented project. Fictional in-house samples illustrate our process; they are not evidence of client sales results.

Watch this production clip (MP4)

Source and production inquiries

Review a Generated Clip Before Paying for Another Attempt

佐藤光 — Wed, 14 Oct 2026 15:23:23 +0000

A disappearing hand made the end of one shot unusable. It did not make every frame of that shot unusable.

This implementation note records the inputs, failure modes and checks needed to make the workflow reviewable.

Separate generation from exact text

We kept product names and logos out of the generated image wherever exactness mattered, then added them in the editing stage with Remotion. Generated frames can look convincing while containing incorrect marks or words. Review those separately from camera motion and lighting, and keep the final copy as an editable layer.

Two errors, two responses

One shot moved in the wrong direction at a mailbox, so we generated it again. Another had a hand disappear at about 3.8 seconds. We retained the usable first 3.6 seconds and adjusted its speed to fit the edit. Those were different fixes for different errors, not a rule to slow down every problematic clip.

Select images before motion

For a fictional 30-second coffee subscription sample, we generated 15 Soul 2 images and six five-second Kling 3.0 Pro clips through Higgsfield. The estimated generation total was about USD 5.1, within a USD 10 budget. That does not include production labor. Comparing image candidates first helped us decide which compositions deserved a video attempt.

Checks to carry into the next project

Identify the first unusable frame.
Decide whether the error is in the input, motion or edit.
Estimate the next attempt before sending it.

Keep a reproducible record of the configuration and inspect the final exported artifact. An automation success message is not a content review.

Watch the production examples

These assets come from the documented project. Fictional in-house samples illustrate our process; they are not evidence of client sales results.

Watch this production clip (MP4)

Source and production inquiries

Keep the Composition Stable Across Sketch, Pencil and Generated Film

佐藤光 — Wed, 14 Oct 2026 15:16:50 +0000

The transition worked because the drawing and the film-like image described the same composition.

This implementation note records the inputs, failure modes and checks needed to make the workflow reviewable.

Review inputs and transitions

We reviewed 25 images before animating them, including faces, a car emblem and a creature design. Motion then revealed different problems: a residual white car and a mismatch in the road at the cut needed attention. An image review cannot replace a motion review. Inspect the first and last frames as well as the attractive middle frame.

Give each tool one job

Code drew the lines, Nano Banana Pro produced the still images, Kling v3 added movement, and ElevenLabs supplied music and effects. This was a fictional production sample rather than footage of real events. It shows how generated assets and editing can be combined; it does not imply that one prompt produced the completed film.

Build a shared composition

Our 37-second in-house sample follows a fictional notebook hero through rooftops, a chase, a creature and space before returning to dinner. We started with simple line drawings, generated film-like images from those compositions, then made pencil-shaded versions. Keeping the layout recognizable helped each visual style read as the same scene.

Checks to carry into the next project

Show a rough sequence, even if it is a stick drawing.
Name the details that must remain consistent.
Review transitions and endpoints after generating motion.

Keep a reproducible record of the configuration and inspect the final exported artifact. An automation success message is not a content review.

Watch the production examples

These assets come from the documented project. Fictional in-house samples illustrate our process; they are not evidence of client sales results.

Watch this production clip (MP4)

Source and production inquiries

Ship an AI Video with Display and Measurement Requirements

佐藤光 — Wed, 14 Oct 2026 15:16:20 +0000

An MP4 is a delivery file. Whether it works on a landing page or in an ad depends on decisions made around that file.

This implementation note records the inputs, failure modes and checks needed to make the workflow reviewable.

Connect the version to the result

Give each video version a clear identifier and connect it to the campaign URL or UTM values. Decide whether success means a consultation or another concrete conversion. If both pixel and server-side events are used, check deduplication so one inquiry is not counted twice. A playback count does not establish that the video generated qualified work.

Make the page usable before playback

For a landing page, prepare a poster image and decide when the video should load. Avoid loading every large clip immediately, especially below the fold. Test muted inline playback on the actual mobile page, and confirm that the page still communicates its offer before the visitor presses play. The video should support the page rather than delay its basic use.

Start at the placement

Choose aspect ratio, duration and safe areas for the intended placement. A horizontal landing-page film and a vertical social ad may need different compositions, not just different crops. Share the destination and launch date before storyboarding, then check the current requirements of the platform where the video will run.

Checks to carry into the next project

Name the page or ad placement.
Specify ratio, length and publication date.
Define an inquiry event before judging results.

Keep a reproducible record of the configuration and inspect the final exported artifact. An automation success message is not a content review.

Watch the production examples

These assets come from the documented project. Fictional in-house samples illustrate our process; they are not evidence of client sales results.

Watch this production clip (MP4)