Generation history

87 recovered generation attempts; 77 outputs, plus a local audio-treatment variant. The original 57-attempt ledger is below.

• Original ledger: 57 attempts, 48 outputs, 9 no-media requests.
• Two head-on edits: 59 attempts, 50 outputs. Backend and cost unknown; owner acceptance pending.
• Lip-sync round: 9 requests, 8 clips, 1 policy rejection; $5.12264 generation API usage. Archive total: 68 attempts, 58 outputs. Sync-3 refines a shared H3 base.
• Room-audio test: one OmniHuman V2 render using the second head-on image and existing V4 speech. Total: 69 attempts, 59 outputs. The treated version changes audio, not frames; it is not another paid generation.
• Subtle room-audio variant: existing wider H3 picture and V4 speech; $0 additional API spend ($0/min incremental). Original H3 remains $4.80/min. No new paid generation.
• Wider H3 speaking test: one6second clip, $0.48 generation API usage ($4.80/min). Prior image and audio reused; no new paid reviews.
• Named-model follow-up: 3 images plus 2 Google refinements, $0.78182 reported generation API usage; no new AI grades.
• Microphone and realism round: 12 built-in image generations with unknown USD charges. All remain available with individual votes and paraphrased revision notes. Read the revision story →

60 visual examples and 13 voice recordings have separate ballots. Context images, editorial copies and audiogram listening wrappers are not extra paid generations.

The pack includes 18 historical stills, 17 historical speaking clips and 13 voice recordings, plus later rounds. The image library includes the original phone image and two later edits. Full Google auditions, script segments and pacing tests remain available.

Generation attempts recovered in production case r006, including the appended Wan completions. Older built-in images have no reconstructed API charge. This is not invoice reconciliation.