> ## Documentation Index > Fetch the complete documentation index at: https://docs.eversince.ai/llms.txt > Use this file to discover all available pages before exploring further. # transcribe media > The words in audio and video, with timing. One ref or url: a timestamped transcript, sentence-level (word-level with include_words); media under five minutes answers in the call, longer returns a job and a second call after it finishes returns the words. Cached per library item, so a repeat read costs nothing; a url outside the workspace is read again every call. refs, up to 20: the words already stored on each file, never transcribing; a file nobody has read comes back unread. timeline_id: transcribes every uploaded clip on the timeline whose words are unknown and writes them onto the clips, which is what captions, get_transcript and remove_words read. format lines is the compact form for reading a long file whole; phrases marks the pauses, for choosing cut points. Cloud processing. ## OpenAPI ````yaml /openapi.json post /tools/transcribe_media openapi: 3.1.0 info: title: Eversince API version: '1' description: >- Every Eversince tool as one POST call, plus the routes beside them: the tool list, account and keys, uploads, webhooks and models. The prose is at https://docs.eversince.ai. servers: - url: https://eversince.ai/api/v1 security: - bearer: [] tags: - name: Discovery description: The tool list, the same on the MCP server, the REST API and the CLI. - name: Workspace description: Jobs, balances, settings, templates, skills and the overview. - name: Library description: >- Items, search, import, reading media, comments, boards, calendars, the brand kit and public links. - name: Timeline description: 'Video editing: timelines, clips, captions, sound, language and rendering.' - name: Canvas description: 'Still editing: canvases, slides, layers and rendering.' - name: Generation description: Image, video and audio generation, upscaling, cutouts and models. - name: Research description: 'What platforms publish: pulling and keeping posts.' - name: Account description: Balances, keys, workspaces and script sessions. - name: Uploads description: Files into the library by presigned upload or by URL. - name: Webhooks description: Job results posted to a URL as they complete. - name: Models description: Generation models and cost estimates. paths: /tools/transcribe_media: post: tags: - Library summary: transcribe media description: >- The words in audio and video, with timing. One ref or url: a timestamped transcript, sentence-level (word-level with include_words); media under five minutes answers in the call, longer returns a job and a second call after it finishes returns the words. Cached per library item, so a repeat read costs nothing; a url outside the workspace is read again every call. refs, up to 20: the words already stored on each file, never transcribing; a file nobody has read comes back unread. timeline_id: transcribes every uploaded clip on the timeline whose words are unknown and writes them onto the clips, which is what captions, get_transcript and remove_words read. format lines is the compact form for reading a long file whole; phrases marks the pauses, for choosing cut points. Cloud processing. operationId: transcribe_media parameters: - $ref: '#/components/parameters/IdempotencyKey' - $ref: '#/components/parameters/WorkspaceId' requestBody: required: true content: application/json: schema: type: object properties: ref: type: string description: An item ref, item id, or the job id that made it. url: type: string format: uri refs: description: >- Up to 20 audio or video items: their stored words, free. One that cannot be read comes back as its own row beside the rest. minItems: 1 maxItems: 20 type: array items: type: string timeline_id: description: >- Transcribe the uploaded clips on this timeline that have no words yet, and write the words onto them. type: string clip_ids: description: >- With timeline_id: only these clips, read again from scratch even if they have words, the way back from a transcript speech recognition got wrong. type: array items: type: string include_words: description: 'ref or url: word-level timing as well as sentences.' type: boolean format: description: >- sentences (default): each with start and end. lines: one timed line per sentence as one string, half the size, for reading a long file whole. phrases: speech broken at pauses, each with the gap after it, for choosing cut points. type: string enum: - sentences - lines - phrases start_time: description: >- ref or url: return only from this second. The whole file is still read. type: number minimum: 0 end_time: description: 'ref or url: return only up to this second.' type: number minimum: 0 min_gap_ms: description: >- format phrases: a pause this long or longer breaks one phrase from the next. Default 400; 700 and up finds only the breaths between takes. type: integer minimum: 0 maximum: 10000 speakers: description: >- format phrases: name who is talking on each phrase, from the voices the speakers tool picked out. A file nobody has been picked out of has none; this never listens to find out. type: boolean window: description: >- format phrases: only phrases touching this span of the file, in seconds. type: object properties: start: type: number minimum: 0 end: type: number minimum: 0 required: - start - end context: description: 'Optional: one sentence shown to the user beside this action.' type: string maxLength: 600 responses: '200': description: >- Done, or a job started for work that runs longer (follow it with get_jobs). content: application/json: schema: type: object required: - ok - data - meta properties: ok: const: true data: type: object additionalProperties: true meta: type: object properties: request_id: type: string cost: $ref: '#/components/schemas/Cost' cloud_processing: $ref: '#/components/schemas/CloudProcessing' media: type: array items: type: object additionalProperties: true default: $ref: '#/components/responses/Error' components: parameters: IdempotencyKey: name: Idempotency-Key in: header required: false schema: type: string description: >- Retry a paid dispatch safely: the same key returns the job already dispatched instead of charging again. WorkspaceId: name: x-workspace-id in: header required: false schema: type: string description: The workspace this one call acts in, when it is not the key's own. schemas: Cost: type: object description: AI credits the call cost. additionalProperties: true CloudProcessing: type: object description: Cloud processing the call used, and what is left. properties: ran_on: const: cloud minutes: type: number queued_minutes: type: number left_minutes: type: number bought_minutes: type: number note: type: string Error: type: object required: - ok - error properties: ok: const: false error: type: object required: - code - message - retryable properties: code: type: string message: type: string description: What went wrong, in words to act on. retryable: type: boolean suggested_action: type: string description: The next step that works. cloud_processing: $ref: '#/components/schemas/CloudProcessing' responses: Error: description: The call was refused or failed. content: application/json: schema: $ref: '#/components/schemas/Error' securitySchemes: bearer: type: http scheme: bearer description: >- An API key (es_live_…) from Settings, under API keys, or an OAuth access token. ````