No clickbait, no unverified claims. We ingest primary sources, extract facts with exact quotes, structure scripts using the Ikegami explainer philosophy, enforce 4-role human reviews, and compile finished videos with Cloudflare-hosted Hypit containers.
A studio architecture that pairs generative AI speed with serious newsroom rigor.
URLs and pasted texts are frozen in Cloudflare R2. AI can only extract claims backed by verbatim substring quotes.
First-mention jargon explanations, intuitive unit conversions, and zero preachy superiority. Anti-AI rules guarantee natural speech.
Fact, language, rights/risk, and final audio-visual reviewers must all approve. Any script edit instantly invalidates prior approvals.
A containerized Hypit compiler turns SVML definitions into HTML/CSS motion, synchronized audio, and crisp 1080p MP4 deliverables.
Powered by Gemini 3.8 Flash and Flash-TTS via new-api, with strict JSON-schema enforcement on every turn.
Every ingestion, edit, approval, and render is logged in a cryptographic SHA-256 chain. Database edits are instantly detectable.
Automating repetitive grunt work so humans can focus on what matters: verifying truth and honing tone.
Add URLs or raw text. Robots.txt, redirects, and size limits are enforced before freezing the source text in R2 storage.
Gemini extracts factual candidates with exact quotes. A human fact-checker verifies quotes against the source before scripts can cite them.
Generate 60s shorts or 3–5m deep dives citing only verified claims. Deterministic linting catches absolute wording and sensitive terms.
Synthesizes calm, authoritative narrator voiceover in WAV format. Visual notes describe required screen graphics.
A Cloudflare container compiles the SVML composition, synchronizes subtitles, and encodes the finished MP4 preview.
Once all required reviewers approve, a production ZIP is assembled with video, subtitles, article, provenance, and upload checklists.
Our safeguards are enforced by deterministic software gates, not wishful thinking.
We focus on industry, technology, and daily life. Foreign diplomacy, elections, territorial disputes, and religion are flagged by linting.
The Ikegami style is studied as an educational methodology. Using Akira Ikegami's name or likeness to imply endorsement is a hard build error.
We forbid phrases like 'we visited' or 'our reporter saw' unless actual reporting took place. Everything is honestly framed as 'according to records'.
We do not blast hundreds of burner accounts. We generate clean, compliant packages for human upload onto accounts the team legitimately owns.
From initial pilot validation to media newsrooms and cross-border enterprise deployments.
No. Kaisetsu Studio is an independent platform that engineers the pedagogical explaining techniques pioneered by Mr. Ikegami into formal prompts, linting rules, and review workflows.
No. To protect your brand and comply with platform terms, the studio outputs a complete package (video, subtitles, description, checklist) for human review and manual upload.
Yes. You can select Japanese, Chinese, or English per topic; prompts, editorial linting, and subtitle wrapping automatically adapt to the target language.