popiskill-audio-stt-local-v1

name	popiskill-audio-stt-local-v1
description	Transcribe audio or video through the PopiArt runtime baseline for local-first speech-to-text. Use this when the user wants a PopiArt-managed STT path for transcripts, captions, or analysis without manually running local scripts.
tags	["official","runtime","audio","stt","local"]
version	v1
model_type	audio
estimated_duration_s	120
default_profile	true
profile_description	Official PopiArt runtime baseline for local-first speech-to-text transcription.

PopiArt STT Local

This is an official PopiArt runtime catalog skill.

• Popiart_skillhub owns the public skill definition.
• popiartServer owns registration, execution, jobs, artifacts, and transcript outputs.
• PopiNewAPI owns any managed speech-model routing that sits behind the server boundary.
• popiartcli owns source upload ergonomics, run, jobs, and artifact retrieval.

Use it when the goal is to:

• transcribe one audio or video source into text
• produce a quick transcript or caption draft through PopiArt
• keep the workflow inside PopiArt instead of manually running local STT scripts

Do not use it for:

• text-to-speech output
• full dubbing workflows
• direct provider-key experiments outside PopiArt

Required input

• one media source:
  • source_artifact_id, or
  • audio_url, or
  • video_url

Optional input

• language
• backend
• model
• diarization
• timestamps
• format
• prompt
• notes

Workflow

1. Prefer source_artifact_id when the source clip already comes from PopiArt.
2. Use audio_url or video_url only when the source already lives at a stable URL.
3. Run the skill through popiart.
4. Wait for completion.
5. Pull transcript artifacts or read direct text fields from the result.

Command pattern

popiart run popiskill-audio-stt-local-v1 --input @params.json --wait

Inline example:

popiart run popiskill-audio-stt-local-v1 --input '{"source_artifact_id":"art_123","language":"zh","timestamps":true}' --wait

Payload template

{
  "source_artifact_id": "art_123",
  "language": "zh",
  "backend": "auto",
  "model": "default",
  "diarization": false,
  "timestamps": true,
  "format": "text"
}

Output handling

After the job finishes:

• read job_id
• inspect text when the runtime returns inline transcript text
• inspect artifact_ids for transcript, subtitle, or segment artifacts
• use popiart artifacts pull <artifact-id> when a local transcript file is needed

Operating guidance

• For local media files, upload first with popiart artifacts upload ./clip.wav --role source.
• video_url is acceptable when the runtime extracts audio server-side.
• If the user needs spoken audio output instead, switch to popiskill-audio-tts-multimodel-v1.