设计与多媒体

Speech To Text With Speakers

试用

Speech to Text With Speakers: Transcribe audio from file_id or public_url with three tiered actions for recordings up to 15, 30, or 60 minutes. Use when an a...

它能做什么

Speech to Text With Speakers: Transcribe audio from file_id or public_url with three tiered actions for recordings up to 15, 30, or 60 minutes. Use when an agent needs speech to text with speakers, transcribe meeting recordings, generate subtitles and captions for videos, convert voice memos to searchable text, transcribe podcast episodes, get task, task id, list tasks through AgentPMT-hosted remote tool calls. Discovery terms: speech to text with speakers, transcribe meeting recordings.

技能文档

Speech to Text With Speakers

Freshness

Last updated: 2026-07-22.

If the current date is more than 7 days after the last updated date, reinstall this skill from skills.sh or ClawHub before relying on endpoints, schemas, setup steps, or examples.

What This Tool Does

Turn any audio recording into clean, searchable text in seconds. Transcribe voice memos, meetings, interviews, podcasts, and webinars with accurate speech recognition that handles accents and background noise. Get plain text for quick reference, SRT or WebVTT subtitles for video captioning, or rich JSON output with word-level timestamps and speaker identification. Choose from three tiers based on recording length — up to 15, 30, or 60 minutes — and optionally enable speaker diarization to label who said what, profanity filtering, and alternative transcripts for maximum accuracy.

Product Instructions

Speech to Text

Transcribe audio with one tool. Transcribe actions run as background tasks: the submit response returns a task_id immediately, and short clips usually complete inline in that same response. Poll get_task for anything still processing.

Tool Call Format

{
  "action": "get_instructions"
}
{
  "action": "transcribe_quick",
  "file_id": "FILE_ID",
  "language_code": "en-US",
  "output_format": "text"
}
{
  "action": "transcribe_standard",
  "public_url": "https://example.com/meeting.m4a",
  "output_format": "vtt",
  "enable_word_timestamps": true,
  "enable_diarization": true
}
{
  "action": "transcribe_extended",
  "public_url": "https://example.com/interview.webm",
  "output_format": "json",
  "max_alternatives": 2
}
{
  "action": "get_task",
  "task_id": "TASK_ID"
}
{
  "action": "list_tasks",
  "limit": 20
}

Actions

  • transcribe_quick: audio up to 15 minutes. Price: 100 credits.
  • transcribe_standard: audio up to 30 minutes. Price: 150 credits.
  • transcribe_extended: audio up to 60 minutes. Price: 200 credits.
  • get_task: check a transcription task's progress and retrieve its result. Price: 1 credit.
  • list_tasks: list recent transcription tasks. Price: 1 credit.

Async task flow

  • Every transcribe action returns a task envelope: {action, task_id, status, ...}. ALWAYS check status before polling — short clips finish within the submit request and return status: "completed" with outputs inline, costing zero polls.
  • When status is "processing", poll get_task with the returned task_id every 10-15 seconds. progress advances as the job moves through download, validation, and recognition.
  • A completed task carries the full transcription payload in outputs[0]: the selected content inline (text, srt_content, vtt_content, or json_data), plus speakers, word_count, confidence_score, audio_metadata, and the File Manager artifact fields result_file_id/result_signed_url.
  • A failed task carries error. Tier-limit failures also carry error_details with recommended_actions — resubmit with the suggested larger tier.
  • If the service restarts mid-transcription, in-flight Google batch jobs resume automatically. Jobs that cannot resume are marked failed with an instruction to resubmit; they never sit in processing forever.

Billing

  • Transcribe actions charge on submission. A task that later fails (for example, audio too long for the tier) is not auto-refunded — pick the tier from the recording length you already know, and check error_details.recommended_actions before resubmitting.

Notes

  • Provide either file_id or public_url.
  • public_url must be an HTTPS URL and cannot point to private or internal network addresses.
  • Audio downloads are capped at 150MB. If a recording is larger, compress it to mp3 or m4a and retry.
  • If language_code is omitted, the tool defaults to en-US.
  • Supported output formats: text, srt, vtt, json.
  • Optional controls: enable_diarization, enable_word_timestamps, remove_filler_words, enable_profanity_filter, max_alternatives.
  • remove_filler_words defaults to true, which uses Google STT V2's cleaned transcript path.
  • Set remove_filler_words to false to preserve disfluencies through Vercel AI Gateway using the openai/whisper-1 gateway model slug. This path always requests word-level timestamps from the gateway for clipping workflows.
  • remove_filler_words=false does not support enable_diarization=true or max_alternatives greater than 1; use the default cleaned path for those features.
  • enable_diarization=true supports audio up to 20 minutes (a provider limit). Longer recordings fail with guidance: disable diarization, or split the audio into 20-minute segments and transcribe them individually.
  • With enable_diarization=true, text output is speaker-labelled one line per turn ([0:04] Speaker 0: ...), inline and in the stored transcription.txt, so it is readable without reformatting. Without diarization it is the flat transcript. Speaker numbers match the speaker_tag values in speakers/json_data.
  • During invocations with File Manager storage available, text and json results are stored for every successful transcription; srt and vtt results are stored when the generated subtitle content is non-empty. The corresponding filenames are transcription.txt, transcription.json, transcription.srt, and transcription.vtt.
  • Stored artifacts are returned through the result_file_id and result_signed_url fields in outputs[0]. Agents should use result_file_id for later File Manager operations.
  • If transcription succeeds but artifact storage fails, the inline result remains available and result_file_error explains that the File Manager file could not be created.

When To Use

  • Use this skill for Speech to Text With Speakers on AgentPMT.
  • Use it when an agent needs this specific tool's behavior, schema, inputs, outputs, and invocation shape.
  • Search and activation keywords: speech to text with speakers, transcribe meeting recordings, generate subtitles and captions for videos, convert voice memos to searchable text, transcribe podcast episodes, get task, task id, list tasks.
  • Supported action names: get_task, list_tasks, transcribe_extended, transcribe_quick, transcribe_standard.

Use Cases

  • Transcribe meeting recordings
  • Generate subtitles and captions for videos
  • Convert voice memos to searchable text
  • Transcribe podcast episodes
  • Create interview transcripts with speaker labels
  • Produce SRT or WebVTT subtitle files
  • Build searchable audio archives
  • Transcribe webinars and lectures
  • Analyze customer call recordings
  • Content repurposing from audio to text

Categories And Industries

No categories or industry tags are published for this tool.

Actions And Schema

Complete generated action schema: ./schema.md. Supported action count: 5. x402 availability: not enabled for this product.

  • get_task (action slug: get-task): Check a transcription task's progress and retrieve its result. Completed tasks carry the full transcription payload in outputs[0]. Price: 1 credits. Parameters: task_id.
  • list_tasks (action slug: list-tasks): List recent transcription tasks with status and progress. Price: 1 credits. Parameters: limit.
  • transcribe_extended (action slug: transcribe-extended): Start an asynchronous transcription of audio up to 60 minutes. Returns a task_id immediately (short clips complete inline in the same response); poll get_task for the completed transcript and File Manager artifact. Price: 200 credits. Parameters: enable_diarization, enable_profanity_filter, enable_word_timestamps, file_id, language_code, max_alternatives, output_format, public_url, plus 1 more.
  • transcribe_quick (action slug: transcribe-quick): Start an asynchronous transcription of audio up to 15 minutes. Returns a task_id immediately (short clips complete inline in the same response); poll get_task for the completed transcript and File Manager artifact. Price: 100 credits. Parameters: enable_diarization, enable_profanity_filter, enable_word_timestamps, file_id, language_code, max_alternatives, output_format, public_url, plus 1 more.
  • transcribe_standard (action slug: transcribe-standard): Start an asynchronous transcription of audio up to 30 minutes. Returns a task_id immediately (short clips complete inline in the same response); poll get_task for the completed transcript and File Manager artifact. Price: 150 credits. Parameters: enable_diarization, enable_profanity_filter, enable_word_timestamps, file_id, language_code, max_alternatives, output_format, public_url, plus 1 more.

Live Schema And Examples

Use the compact schema above for ordinary calls. Before a new production integration, or whenever parameters, enum values, nested objects, outputs, or examples are unclear, fetch live details first.

  • Exact schema: call agentpmt-tool-search-and-execution with action: "get_schema", and tool_id: "speech-to-text-with-speakers".
  • Detailed examples: call agentpmt-tool-search-and-execution with action: "get_instructions" and tool_id: "speech-to-text-with-speakers", or call this product with action: "get_instructions" when the product tool is already selected.
  • Treat returned live schema and instructions as more specific than this generated summary.

MCP schema lookup through the main AgentPMT MCP server:

{
  "method": "tools/call",
  "params": {
    "name": "AgentPMT-Tool-Search-and-Execution",
    "arguments": {
      "action": "get_schema",
      "tool_id": "speech-to-text-with-speakers"
    }
  }
}

For live examples, keep the same MCP tool and use these arguments:

{
  "action": "get_instructions",
  "tool_id": "speech-to-text-with-speakers"
}

Authenticated AgentPMT REST schema lookup body:

{
  "name": "agentpmt-tool-search-and-execution",
  "parameters": {
    "action": "get_schema",
    "tool_id": "speech-to-text-with-speakers"
  }
}

Authenticated AgentPMT REST live examples body:

{
  "name": "agentpmt-tool-search-and-execution",
  "parameters": {
    "action": "get_instructions",
    "tool_id": "speech-to-text-with-speakers"
  }
}

Call This Tool

Product slug: speech-to-text-with-speakers

Marketplace page: https://www.agentpmt.com/marketplace/speech-to-text-with-speakers

  • AgentPMT account route: first use ../agentpmt-account-mcp-rest-api-setup to connect the main MCP server or REST API for an Agent Group where this tool is enabled.
  • x402 route: not enabled for this product.
  • AgentPMT overview: use ../what-is-agentpmt for marketplace, Agent Group, workflow, MCP, REST, and payment concepts.

If those setup skills are not installed beside this product skill, use the downloads below.

Core AgentPMT setup skills:

  • What AgentPMT is: ../what-is-agentpmt
  • AgentPMT account MCP/REST setup: ../agentpmt-account-mcp-rest-api-setup

skills.sh install script:

npx skills add AgentPMT/agent-skills --skill what-is-agentpmt
npx skills add AgentPMT/agent-skills --skill agentpmt-account-mcp-rest-api-setup

MCP call shape after the main AgentPMT MCP server is connected:

{
  "method": "tools/call",
  "params": {
    "name": "Speech-to-Text-With-Speakers",
    "arguments": {
      "action": "get_task",
      "task_id": "example task id"
    }
  }
}

Use the exact tool name returned by tools/list; the name above is the expected readable form.

Authenticated AgentPMT REST call body:

{
  "name": "speech-to-text-with-speakers",
  "parameters": {
    "action": "get_task",
    "task_id": "example task id"
  }
}

Use the setup skill for the account connection details before making REST calls.

Response Handling

  • Treat the returned JSON as the source of truth for this tool call.
  • If the response includes warnings or correction targets, apply them before retrying.
  • If the response includes a passed or success-style boolean, use it as the workflow gate.
  • If validation fails or the response shape is unclear, call get_schema or get_instructions before retrying.
  • If get_task fails, preserve the request parameters and retry only after fixing schema, auth, or payment errors.

Security

  • Do not place account secrets, wallet private keys, mnemonics, signatures, or payment headers in prompts or logs.
  • Keep tool inputs scoped to the minimum content needed for the task.
  • Use the setup skills for credential handling; this product skill only defines product-specific behavior.

AgentPMT Reference

相关技能

Pocket: Search, list, and read the user's Pocket recordings including transcripts and AI summaries; create audio download links; save recording audio to file storage; upload new audio recordings; list tags. Use when an agent needs pocket, pocket recordings transcripts, search recordings by topic, pull meeting transcripts, get ai summaries of conversations, upload audio files for transcription, get audio download url, recording id through AgentPMT-hosted remote tool calls.

通过 MCP 把智能体接入 Speak AI,获得 113 个工具、5 个资源、3 个提示模板,完成音视频转写与分析。

32 次安装

Transcribe pre-recorded audio files or URLs with Gladia. Use when the user needs batch/async transcription, speaker diarization, subtitles (SRT/VTT), PII red...

5 次安装

AgentPMT Audit Logs: Look back over this account's own activity, read-only. Search and read past chat sessions, pull a full chat transcript together with the...

Transcribe audio and video with the transcribe.so CLI. Turns YouTube videos, podcasts (Apple Podcasts, Spotify, SoundCloud, Vimeo, Twitch, Loom), direct media URLs, and local audio or video files into speaker-labelled transcripts with timestamped segments, chapters, sections, cited Q&A, and subtitle files (SRT, VTT, karaoke VTT). Use when the user wants a transcript, show notes, chapters, subtitles, quotes, or answers grounded in a recording. 52 languages and dialects.

Transcribe audio to text using OATDA's unified audio API. Triggers when the user wants speech-to-text, transcription of meetings, podcasts, voice notes, subt...

15 次安装