← Back to analysis status

Real video trial · 20 November 2026

The provider rejected the output schema. No video summary was returned.

The one authorized trial ran. Token counting succeeded, then the generation request returned HTTP 400 because our JSON schema was too complex for the provider’s output constraints. The exact error response was retained privately before parsing. No retry was made.

Claude Just Changed: The 6 Big Updates You Need to Know

Duncan Rogoff | Learn Claude Code · Open original video
Video ID: Of9ylhwSkJE · Requested range: 0:00–7:49 (0–469 seconds).

Attempt started at . This is a saved, read-only report. Opening it makes no provider request.

Actual requests and result

Why this attempt failed

The provider reported that the supplied schema produced too many serving states. Its documentation identifies nested array length limits and numeric bounds as possible contributors. Our request combined both. The response did not identify one offending field, so we cannot claim a specific field caused the rejection.

This is an InsightDesk request-schema compatibility failure—not a request to reconnect the key or add funds. The token-count result does not prove that the generation service acquired or understood the video.

Offline repair completed · 20 November 2026. We simplified the provider-facing schema by removing nested array and numeric bounds. Evidence, length, timestamp and modality checks remain mandatory in local validation. All 149 offline tests pass. No new provider request was made, so provider acceptance of the revised schema and video understanding remain unverified. Any further real test needs separate authorization; the permission for this attempt is consumed.

Evidence review and coverage

No verified insights are available. No generated content was returned, so speech fidelity, quotations, timestamp accuracy, visual evidence, filler removal and omissions could not be independently reviewed. No title-based summary is substituted.

The requested 469 seconds are not verified coverage. InsightDesk downloaded no local audio, transcript or video frames. Generation audiovisual access is unestablished; subtitle availability remains unknown. The other channels, long-form and no-creator-subtitle cases remain untested.

Response retention and timing

The bounded, credential-redacted original error response was saved encrypted in private database storage before envelope parsing. An authenticated recovery and integrity check succeeded. It survives replacement of the operator computer; there is no public response or draft download. This response contained no substantive content candidate.

Provider round trips totalled 5.220 seconds. The full runner took 22.241 seconds, including the operator’s durable-storage confirmation wait. That wait is not provider video-processing latency. Provider-internal acquisition and processing times were not reported.

Cost reserve versus actual charges

Reviewed standard rates: $0.30 per million input tokens across text, image, video and audio; $2.50 per million output tokens including thinking. Pricing reference. The documented no-charge YouTube URL preview was not relied on to reduce the reserve.

These reserves are price-based upper bounds, not an invoice, account-level spending cap or remaining credit balance. Hosting and storage costs are not included in the API reserve and are not metered here.

Current state

Authorization consumed; further paid calls locked. Public video analysis remains disabled. You do not need to submit another metadata check: it cannot resolve this schema failure. A faithful audiovisual report is still required before this becomes a usable analysis app.

Earlier trials · 19 November 2026 · Access & methodology · Contact InsightDesk