Add paired audio models for G031_dream_archive (20261105)
Adds paired model audio for G031_dream_archive: 28 WAV files and 14 updated task metadata files. SFX uses ElevenLabs Sound Effects v2 and Stable Audio 3.0 Large; Chinese dialogue uses MiniMax Speech2.8HD and ElevenLabs Multilingual v2. Each task meta.json includes model IDs, exact requests, task/actual durations, hashes, generation IDs, fees and review provenance. 2 tasks in this game use an explicitly recorded acoustic paraphrase shared by both models after original-prompt moderation refusals. Out-of-range SFX requests use the user-authorized endpoint minimum or maximum, with original and actual durations retained. Historical reuse is labeled; inputs and native audio are not presented as equal historical budgets or exact original-duration compliance. Existing primary WAVs and existing metadata fields are preserved. Audio and meta.json only; no deletions, game integration or formal quality scoring. Owner review and merge required.