# Veo 3.1音频结果门禁
Use after a Veo 3.1 native-audio generation returns a candidate clip. Python 3.9+ is required. The skill reads only redacted JSON evidence and never downloads or opens a media file.
Input one is the JSON contract from veo31-audio-cue-contract. Input two is a redacted result evidence object with status, audio_track_present, output_ref, file_sha256, file_bytes, and one observed row for each cue. Output is veo31_audio_result_gate.json and a Markdown review report.
python scripts/audit_audio_result.py <contract.json> <result.json> --out-dir <output-dir> --tolerance-ms 300.pass_for_human_review as a queue state, not a release approval.The script requires completed status, an audio track, a non-empty output reference, a 64-character file SHA256, positive byte count, exact cue coverage, an output fingerprint not present in prior_fingerprints, and observed cue offsets within the buyer-selected tolerance. It always emits release_ready=false; no threshold can replace human listening, visual review, or rights confirmation.
Keep the blocked report, correct the evidence or regenerate through the provider\'s normal controls, then rerun. Do not mark missing cues as present, reuse a hash from another file, or widen tolerance just to force a pass.
Use redacted references and hashes only. Do not put API keys, cookies, private voice samples, customer data, or unlicensed dialogue in the evidence. External generation, downloading, publishing, and payment remain manual and permission-gated.
The bundled valid fixture passes structural checks while remaining pending human review. The invalid fixture exits nonzero for coverage and duplicate fingerprint errors. A buyer accepts the skill only when the report, error paths, and human-review boundary are present.
请通过上方所属套餐查看交付内容与下载方式。
下载不等于安装成功,运行环境、平台适配和外部服务需按说明准备。
此处不单独展示技能价格,请以上方所属套餐价格为准。