auto4.55 KB
View saved version →
---
name: auto
description: Create AI music videos, lyric videos, and narrative music videos from existing songs and character photos with One More Shot (1ms). Use to generate music videos, transcribe lyrics, upload media, and retrieve finished videos or video history.
---
Use the connected 1ms MCP. Start with `get_account` and `get_generation_options` to obtain eligibility, current rates, supported styles, and input limits. The connection uses OAuth, independently of public API keys. Do not request or expose API keys. History may contain both public API and MCP jobs; use `request_origin` when distinguishing them.
Use authorized hosted URLs directly. For native host uploads, call `import_media` and poll `get_media` with backoff until ready. When a file is needed, call `request_upload` with `kind: image` for a singer/character photo or `kind: audio` for a song. This opens the authenticated picker inside the conversation without a Director project. Do not replace the tool call with a Markdown workspace link: links open a browser. Wait for the picker to return a ready import ID, then use it in generation inputs. If the host cannot send the selection back, the picker provides the import ID for the user to copy. Do not fabricate URLs or use local filesystem paths as hosted media.
When the user requests a quote, preparation only, or a pause before starting, call the appropriate preparation tool (`prepare_lyric_video`, `prepare_narrative_video` or `prepare_lyric_transcription`) only and stop; do not call the matching start tool (`start_lyric_video`, `start_narrative_video` or `start_lyric_transcription`) until the user requests it. For an unambiguous request to generate now, gather only missing creative inputs, then call the appropriate preparation tool (`prepare_lyric_video`, `prepare_narrative_video` or `prepare_lyric_transcription`) followed by the matching start tool (`start_lyric_video`, `start_narrative_video` or `start_lyric_transcription`) using the returned operation ID. Honor host approvals and any user limits. Default to basic quality and vertical 9:16. Narrative generation requires a prompt and 1–4 character photos with exactly one singer. Lyric and narrative videos use the entire uploaded audio. For a shorter video, upload the desired excerpt first; `trim_start` and `trim_end` are supported only for lyric transcription. Never silently generate the full song when a shorter excerpt was requested. Lyric and narrative style catalogs differ. Use transcription segments without rewriting timing fields. If the user supplies display lyrics and timing explicitly, preserve those words and timings in valid segments; do not claim they were transcribed from the audio.
Explain live credit rates and measured estimates when available. The initial queued response with zero charged credits is not free generation. MCP generation never triggers automatic credit purchases. Use existing credits only. If generation eligibility is missing, explain that the account lacks the required entitlement without promoting a purchase or upgrade. Do not create checkout flows or alter purchase history.
Preserve the operation ID and returned job ID. Retry uncertain submissions only with that same operation ID. Changed inputs require a new prepared operation. If reconciliation is required, stop submitting and report that state; never bypass it with a fresh operation.
Use `get_job` or `list_jobs` for progress and history. Respect Retry-After, avoid invented completion percentages, and distinguish cancellation requested from confirmed cancellation and actual refunded credits. `open_video_workspace` provides playback/download and a web fallback. Jobs live in API/plugin history, not the Auto gallery.
Finished video emails go to the account email (development delivery is restricted to the test recipient). The active UI may send one best-effort completion message when supported. Do not promise background messages in closed conversations. No transcription emai
Preview of the first 4,000 characters. Read full instructions →
