Skip to main content
Glama

Sats4AI - Bitcoin-Powered AI Tools

by io.github.cnghockey
OverviewDiscussionsAdmin

transcribe_audio

Transcribe audio to text with WORD-LEVEL timestamps (timestamps:'word' returns per-word start/end times — subtitle alignment, karaoke captions, cutting video to speech) or segment timestamps. Uses Mistral Transcription — high-accuracy speech recognition that handles accents, background noise, and overlapping speakers. 13 languages: en, zh, hi, es, ar, fr, pt, ru, de, ja, ko, it, nl. Pass names (the people and places spoken in the audio) to help the speech model spell them right — free, same price. Up to 500 MB / 60 minutes per file. Async — returns requestId, poll with check_job_status(jobType='transcription'), then get_job_result. 10 sats/min. Privacy: audio and transcripts are ephemeral — processed, returned, and discarded. Never persisted. Pay per request with Bitcoin Lightning — no API key or signup needed. Requires create_payment with toolName='transcribe_audio'.

Input Schema

TableJSON Schema
NameRequiredDescriptionDefault
namesNoOptional: the names and places spoken in the audio (characters, people, towns), e.g. ["Siobhan", "Tadhg", "Chicoutimi"]. It helps the speech model spell them right. Free: the price does not change. We send up to 100At most 4000 characters of names. List names only: a name that sounds like an ordinary word can replace that word in the transcript ("Noone" for "noon"). If the model refuses the list, you still get the transcript, without it, and result.degraded includes "names".
diarizeNoIdentify different speakers (default false). Forces segment granularity upstream. If the speaker-label request fails, you still get the text, and result.degraded includes 'diarization'.
languageNoOptional. One of: detect (the default), en, zh, hi, es, ar, fr, pt, ru, de, ja, ko, it, nl. A locale tag (en-US) or the English name (French) also works; "auto" means detect.
paymentIdYesValid payment ID (must be paid)
timestampsNoTimestamp granularity in result.segments. 'word' returns per-word start/end times (subtitle alignment, karaoke captions, cutting video to speech). Default 'segment'. If the timestamp request fails, you still get the text, and result.degraded includes 'timestamps' (no segments, srt or vtt).
audioBase64YesBase64 encoded audio file
callback_idNoOptional correlation string echoed back in the webhook body. Max 128 chars.
callback_urlNoOptional HTTPS webhook we POST when the job finishes (HMAC-signed). Polling still works.

Schema Changelog

Changes observed during successful MCP inspections.

  1. Changed1 schema field changed
    • changedInput schema / properties / language / description
      Previous value: -"Language code (e.g., 'en', 'es')"New value: +"Optional. One of: detect (the default), en, zh, hi, es, ar, fr, pt, ru, de, ja, ko, it, nl. A locale tag (en-US) or the English name (French) also works; \"auto\" means detect."
  2. Changed1 schema field changed
    • addedInput schema / properties / names
      Added value: +{
+  "description": "Optional: the names and places spoken in the audio (characters, people, towns), e.g. [\"Siobhan\", \"Tadhg\", \"Chicoutimi\"]. It helps the speech model spell them right. Free: the price does not change. We send up to 100At most 4000 characters of names. List names only: a name that sounds like an ordinary word can replace that word in the transcript (\"Noone\" for \"noon\"). If the model refuses the list, you still get the transcript, without it, and result.degraded includes \"names\".",
+  "items": {
+    "type": "string"
+  },
+  "type": "array"
+}
  3. Changed2 schema fields changed
    • changedInput schema / properties / diarize / description
      Previous value: -"Identify different speakers (default false). Forces segment granularity upstream."New value: +"Identify different speakers (default false). Forces segment granularity upstream. If the speaker-label request fails, you still get the text, and result.degraded includes 'diarization'."
    • changedInput schema / properties / timestamps / description
      Previous value: -"Timestamp granularity in result.segments. 'word' returns per-word start/end times (subtitle alignment, karaoke captions, cutting video to speech). Default 'segment'."New value: +"Timestamp granularity in result.segments. 'word' returns per-word start/end times (subtitle alignment, karaoke captions, cutting video to speech). Default 'segment'. If the timestamp request fails, you still get the text, and result.degraded includes 'timestamps' (no segments, srt or vtt)."
  4. Changed4 schema fields changed
    • addedInput schema / properties / callback_id
      Added value: +{
+  "description": "Optional correlation string echoed back in the webhook body. Max 128 chars.",
+  "type": "string"
+}
    • addedInput schema / properties / callback_url
      Added value: +{
+  "description": "Optional HTTPS webhook we POST when the job finishes (HMAC-signed). Polling still works.",
+  "type": "string"
+}
    • addedInput schema / properties / diarize
      Added value: +{
+  "description": "Identify different speakers (default false). Forces segment granularity upstream.",
+  "type": "boolean"
+}
    • addedInput schema / properties / timestamps
      Added value: +{
+  "description": "Timestamp granularity in result.segments. 'word' returns per-word start/end times (subtitle alignment, karaoke captions, cutting video to speech). Default 'segment'.",
+  "enum": [
+    "none",
+    "segment",
+    "word"
+  ],
+  "type": "string"
+}
  5. First observed

TDQS

A4/5.0
Behavior4/5

Does the description disclose side effects, auth requirements, rate limits, or destructive behavior?

With no annotations, the description carries the full burden and does so well: it discloses async behavior (returns requestId, then poll), cost (10 sats/min, pay-per-request Lightning, no API key), size limits (500 MB / 60 min), and an explicit privacy guarantee that audio and transcripts are ephemeral and never persisted. It does not cover failure/error handling in prose, which keeps it at 4.

Agents need to know what a tool does to the world before calling it. Descriptions should go beyond structured annotations to explain consequences.

Conciseness4/5

Is the description appropriately sized, front-loaded, and free of redundancy?

Dense single block, but front-loaded with the capability and timestamps modes, and essentially every sentence carries operational information (engine, languages, limits, async flow, pricing, privacy, payment prerequisite). A bit of grouping or structure would improve scanability.

Shorter descriptions cost fewer tokens and are easier for agents to parse. Every sentence should earn its place.

Completeness4/5

Given the tool's complexity, does the description cover enough for an agent to succeed on first attempt?

For an 8-parameter async job tool with no annotations and no output schema, the description covers the lifecycle (payment → submit → poll → fetch), cost, and privacy, and hints at result shape (segments, srt/vtt, degraded flags). It stops short of fully describing the returned payload, which matters given there is no output schema.

Complex tools with many parameters or behaviors need more documentation. Simple tools need less. This dimension scales expectations accordingly.

Parameters4/5

Does the description clarify parameter syntax, constraints, interactions, or defaults beyond what the schema provides?

Schema coverage is 100%, so the baseline is 3, but the description adds genuine meaning beyond the schema: it explains that 'names' is free (same price), why to pass it, and ties 'word' timestamps to concrete downstream use cases. That is real semantic value on top of the structured fields.

Input schemas describe structure but not intent. Descriptions should explain non-obvious parameter relationships and valid value ranges.

Purpose4/5

Does the description clearly state what the tool does and how it differs from similar tools?

States a specific verb and resource ('Transcribe audio to text') plus differentiating scope details (word-level vs segment timestamps, Mistral engine, 13 languages). However, it never addresses the closely-named sibling transcribe_translate, so an agent has no explicit signal for when to prefer this tool over that one.

Agents choose between tools based on descriptions. A clear purpose with a specific verb and resource helps agents select the right tool.

Usage Guidelines4/5

Does the description explain when to use this tool, when not to, or what alternatives exist?

Gives concrete use contexts ('subtitle alignment, karaoke captions, cutting video to speech') and names the surrounding workflow tools it depends on (create_payment, check_job_status, get_job_result). No when-not guidance and no contrast with the translation sibling, so it stops short of a 5.

Agents often have multiple tools that could apply. Explicit usage guidance like "use X instead of Y when Z" prevents misuse.

All Sats4AI - Bitcoin-Powered AI Tools toolsInspect
Try in Browser
One endpoint for every MCP toolRayrun runs every MCP server behind one URL and holds the credentials. Each call is decided per client, tool, and argument, then recorded.ray.runAd

Glama MCP Gateway

Add one secure layer between your agents and this server.

Resources