# Generate paid-ad transcript

Generate and persist a transcript from one stored ad’s mirrored video.

## Endpoint: POST /v1/paid-ads/{platform}/{archiveId}/transcript

- Method: `POST`
- Path: `/v1/paid-ads/{platform}/{archiveId}/transcript`
- Authentication: Bearer API credential or OAuth access token
- Required scope: `content:write`
- MCP tool: `refresh`
- MCP action: `ad_transcript`

## Parameters

| Name | In | Type | Required | Description |
| --- | --- | --- | --- | --- |
| `platform` | path | meta | tiktok | yes | Advertising platform. |
| `archiveId` | path | string, 1–255 | yes | Provider archive id. |

## Known errors

- 400 invalid_request for an invalid filter, identity or limit.
- 401 authentication_required for a missing or invalid credential.
- 402 subscription_required when access is not entitled.
- 403 authorization_required for a missing scope or unavailable preview access.
- 403 workspace_not_accessible when x-genviral-tenancy names a workspace the credential cannot reach.
- 404 not_found when the record is outside the accessible catalogue.
- 429 rate_limited; honor Retry-After when present.
- 429 daily_call_limit; the plan's daily API and MCP calls are used until 00:00 UTC.
- 409 conflict when the write would duplicate an exclusive row.

## Request

The command has no request body. It uses only a stored, mirrored video; image-only ads return 422. A concurrent generation returns 409, while a ready transcript is returned from storage.

```
curl --request POST 'https://api.peekpanda.com/v1/paid-ads/meta/111222333/transcript' \
  --header "Authorization: Bearer $PEEKPANDA_TOKEN" \
  --header 'x-genviral-tenancy: personal'
```

## MCP request

Call refresh with action ad_transcript. Path fields go in params, filters in query, and a write body in body.

```
{
  "name": "refresh",
  "arguments": {
    "request": {
      "action": "ad_transcript",
      "params": {
        "platform": "<platform>",
        "archiveId": "<archiveId>"
      }
    }
  }
}

```

## Response

Returns the persisted transcript with word timing, language, and generation-time provenance when available.

```
{
  "text": "Turn saved recipes into a grocery list.",
  "words": null,
  "language": "en",
  "model": "transcription-model",
  "generatedAt": "2026-09-06T00:00:00.000Z"
}

```

## Source

Stored public advertising-library observations with source dates. Spend is unavailable unless the public record reports it.

## Refresh

An explicit generation uses the owned mirror and persists its result. Repeating after settlement returns the stored result without generating it again.