Skip to main content
POST
Voice chosen segments into downloadable clips
Voices the chosen segments (1-100) into one downloadable SegmentClip per segment, without rebuilding the video: a take stored under the job and never overwritten, at its natural duration (fit, overrun_seconds); any voiced clip counts as success. Text is voiced as written. When to use: to download or check chosen lines before a dub. A later changed_segments dub merges unchanged clips without charging them again (estimateDub lists them). A first or full dub voices and charges every segment; when only reused segments changed, the shortest is voiced again at its own speech price (minimum 1 minute). Side effects: Segments are voicing until the job ends; then audio is the new take and dub_state is outdated (not_dubbed without a dub), or the previous state on failure. ETags may change on any take: use clips[].segment_etag, or re-read before an edit. Cost: Spends max(1, round(s/60)) minutes (half to even), with s the speech seconds of the chosen segments; a free dubbing variant costs 0. Pass estimateSegmentGeneration’s total_minutes as expected_cost_minutes. If a segment fails, the job fails and refunds whole, usually with no clips; retry with a new key. Async: returns 202 and a segment_generation Job (Location); wait with GET /v3/jobs/{job_id}?wait_seconds=20 (poll hint 5 s); typically 30 s (p95 5 min). Idempotency: Idempotency-Key required; a same-key retry within 24 h returns the job’s current state. Plan: audio_download. Required scopes: dubbing:write, credits:spend Errors: demo_project_read_only (403), trial_project_dub_not_allowed (403), feature_not_in_plan (403), segment_not_found (404), transcript_not_ready (409), voices_not_assigned (409), project_not_ready (409), dub_in_progress (409), segment_busy (409), segments_missing_translation (422), instant_clone_not_available (422), cost_changed (409), insufficient_balance (402), too_many_concurrent_jobs (429)

Authorizations

Authorization
string
header
required

API key from Settings -> API keys, sent as Authorization: Bearer <key> (SDK environment variable RASK_API_KEY). Bound to one workspace and one acting user. Operations list the scopes they need (OpenAPI 3.1 scopes on a bearer scheme).

Scopes

dubbing:write
credits:spend

Headers

Idempotency-Key
string
required

Required. A unique key per logical request (SDKs send a bare UUIDv4; an RFC 9651 quoted string is accepted and unquoted); reuse it unchanged on retries. Kept 24 h per (workspace, acting user).

Required string length: 1 - 255
Pattern: ^[\x20-\x7E]+$

Path Parameters

project_id
string
required

Project id (from createProject or listProjects).

Maximum string length: 64

Body

application/json

Body of generateSegments.

segment_ids
string[]
required

Segments of the live transcript to voice (1-100).

Required array length: 1 - 100 elements
Maximum string length: 64
expected_cost_minutes
integer<int32>
required

Cap on the minutes this request may spend (any minute pool). Send the estimate's total_minutes. A higher current price fails with 409 cost_changed and spends nothing; an equal or lower price is charged as is.

Required range: 0 <= x <= 100000

Response

Accepted: Job.

Asynchronous work started through v3, and automatic follow-up work (auto-dub, on-screen auto-start, the stages of a localization run) on projects created through v3. Started by 202 responses (Location: /v3/jobs/{id}); wait with getJob wait_seconds. Finished jobs can be read for 90 days, then answer 404 job_not_found. Work started in the web app or API v2 has no job.

object
enum<string>
required

Always job.

Available options:
job
id
string
required

Job id.

workspace_id
string
required

Workspace that owns the job (and whose owner pays).

type
enum<string>
required

Job type. Open enum: new types are added over time.

  • media_import: registering uploaded or imported media (poll hint 3 s; cancelable)
  • project_processing: the project pipeline up to translation (poll hint 10 s)
  • translation: retranslating segments (poll hint 5 s)
  • segment_generation: paid voicing of chosen segments into downloadable clips (poll hint 5 s)
  • dub: dubbing or redubbing (poll hint 10 s)
  • lipsync: lip-sync (poll hint 30 s)
  • onscreen: on-screen text translation (poll hint 15 s; cancelable while queued)
  • localization: dub, lip-sync and on-screen as one run with child jobs (poll hint 15 s)
  • voice_clone: creating a voice clone (poll hint 10 s)
  • export: rendering an export (poll hint 15 s)
  • file_restore: restoring archived files (poll hint 300 s)
Available options:
media_import,
project_processing,
translation,
segment_generation,
dub,
lipsync,
onscreen,
localization,
voice_clone,
export,
file_restore
status
enum<string>
required

Job status. Monotonic: never goes backwards and never leaves a terminal state.

  • queued: accepted, not started
  • running: in progress
  • succeeded: finished; see result
  • failed: finished; see error
  • canceled: canceled, or its project was deleted (error.code = project_deleted)
Available options:
queued,
running,
succeeded,
failed,
canceled
progress
integer<int32> | null
required

0-100; null when the type has no honest progress.

Required range: 0 <= x <= 100
stage
enum<string> | null
required

Per-type stage; for a localization job, the running stage's.

Available options:
queued,
uploading,
transcribing,
separating_audio,
detecting_speakers,
matching_voices,
translating,
voicing,
merging,
downloading,
extracting_audio,
face_check,
waiting_for_worker,
rendering,
charging,
waiting_for_runner,
starting,
preparing,
checking,
publishing,
cloning,
restoring
queue_position
integer<int32> | null
required

Queue position (1 = next).

estimated_finish_at
string<date-time> | null
required

Estimated completion from the queue position and the type's typical duration (on-screen: the runner's ETA); null when terminal or unknown. An estimate, not a deadline.

overdue
boolean
required

True once the job runs past its type's overdue threshold: Rask is alerted. Keep waiting or check back later; do not start the same work again.

project_id
string | null
required

Project.

media_id
string | null
required

Media.

parent_job_id
string | null
required

The job that started this system job: the project_processing job of an auto-dub, the dub whose merge started an on-screen auto-start, the localization job of a stage, the project_processing job of a localization block, or the failed run a resumed localization job continues.

channel
enum<string>
required

Who started a job.

  • api: REST API v3
  • mcp: the Rask MCP server
  • system: automatic (auto-dub, on-screen auto-start, the stages of a localization run)
Available options:
api,
mcp,
system
params
object
required

Normalized request echo: a union on type (= Job.type).

details
object | null
required

Live per-type state.

stages
object[] | null
required

The stages of a localization job in run order; null for other types.

cost
object
required

Quoted at creation; charged turns true once the charge is committed. A localization job has no charge of its own: minutes is the sum of its stages charged and not refunded so far, charged is true once any stage is charged, refunded is true only when every charged stage was refunded, refunded_at is the latest refund. stages[].cost holds each stage; the quote is params.expected_cost_minutes.

result
object
required

Set when succeeded; a failed segment_generation job may also list clips already voiced.

error
object | null
required

Set when failed; on a canceled job only for the cancellation reason project_deleted (the project was deleted while the job ran; its work is discarded, refunded only for on-screen); instance is /v3/jobs/{id}, detail is generic. Each code's status, retry, recommendation and hint are in components.x-rask-job-errors. Codes by type: media_import source_download_failed, unsupported_media_type, media_corrupt, media_processing_failed, media_too_large (a voice sample above its limit); project_processing media_processing_failed, source_unavailable, no_audio_track, transcription_failed, no_speech_detected, audio_separation_failed, speaker_detection_failed, translation_failed, subtitle_import_rolled_back, project_processing_timeout, internal_error; translation translation_failed, translation_dispatch_failed, translation_timeout; segment_generation (cause_code) segment_generation_failed, segment_generation_timeout; localization localization_stage_failed, localization_stage_skipped (failed_stage, cause_code); dub (cause_code) dub_voiceover_failed, dub_merge_failed, dub_failed, dub_timeout; lipsync lipsync_failed, face_check_failed, no_face_detected, lipsync_rejected (cause_code), lipsync_timeout; onscreen onscreen_queue_timeout, onscreen_runner_lost, onscreen_runner_failed, onscreen_provider_unavailable, onscreen_finalize_timeout, onscreen_invalid_result, onscreen_publish_failed, onscreen_source_unavailable, onscreen_charge_failed; voice_clone voice_clone_failed, voice_clone_timeout; export export_failed; file_restore restore_failed. A failed job refunds its minutes (onscreen_charge_failed never charged). A failed localization job refunds only the failed stage, by its type's rule: succeeded stages stay charged, and skipped or unreached stages were never charged. A failed project_processing job refunds its project credit when a stage from transcription to translation failed, when it timed out and when a copy failed, but not for media_processing_failed, source_unavailable, no_audio_track or the internal_error of a new project.

created_at
string<date-time>
required

Creation time.

started_at
string<date-time> | null
required

Start time.

finished_at
string<date-time> | null
required

End time.

Path-absolute API links.