Deepfake Narrative Creation

A deepfake video or audio clip is rarely persuasive on its own — it needs a *narrative scaffold*: a plausible transcript, a backstory explaining when and why the footage exists, news-style write-ups, corroborating social posts, witness accounts, and even a "fact-check" that appears to verify it. This technique uses LLMs to mass-produce that scaffold so a synthetic clip lands inside a coherent, self-reinforcing information environment. It works because human credibility judgments depend heavily on context and consistency across sources; an LLM can fabricate that surrounding context faster than audiences or moderators can dismantle it.

• Media forensics on the core artifact: Apply deepfake-detection and signal-forensic analysis (lip-sync, blink/lighting inconsistencies, audio artifacts) to the clip itself
• C2PA / Content Credentials verification: Check for missing or broken provenance manifests; authentic recordings increasingly carry capture credentials
• Narrative-coherence anomaly detection: Fabricated scaffolds appear unusually fast and fully formed; flag synchronized emergence of transcript, articles, and "witnesses"
• Cross-source independence checks: Verify that "independent" corroborations are not stylometrically identical or co-timed (sign of single-source generation)

This technique is the text counterpart to T9 synthetic media and deepfakes — it exists specifically to wrap T9 output in context. It draws on authority impersonation (T8-AT-001) for fake official responses, synthetic evidence (T8-AT-002) for supporting documents, and disinformation seeding (T8-AT-007) for coordinated reactions.