# HALOWERK Numbers Extract

> HALOWERK Numbers Extract is a paid API for AI agents from medien.halowerk.com, paid per call via x402, $0.025/call, status unknown (last checked 2026-10-31).

Extracts and structures all numeric statements (amounts, percentages, dates, durations, etc.) from an audio/video recording of up to 60 minutes, returning each with its kind, spoken wording, normalized value, unit/currency, referent, source sentence, and timestamp.

## Facts

- Endpoint: POST https://medien.halowerk.com/v1/numbers-extract?utm_source=zero.xyz
- Price: $0.025/call
- Payment: x402
- Status: unknown
- Last checked: 2026-10-31
- Activations on Zero: 0
- Tags: x402
- Canonical page: https://www.zero.xyz/c/halowerk-numbers-extract-9c02a3ca
- Structured record (JSON): https://api.zero.xyz/v1/capabilities/cap_L1gU0Cn_xYy8Fy2KEZ0bq

Status and success rate cover calls made through Zero and Zero's own probes. Third-party monitors may report differently.

## How to call it through Zero

Zero handles the 402 payment challenge and records the run. With the Zero CLI installed (`npm i -g @zeroxyz/cli`):

```sh
zero fetch --capability halowerk-numbers-extract-9c02a3ca -d '<json body>'
```

Example prompt: Go through this 45-minute earnings call recording and pull out every number that was spoken — amounts, percentages, dates, and durations — along with what each number refers to and the exact sentence it came from. The file is at https://example.com/earnings-q3.mp3 and the language is English.

## When to prefer this

Choose this endpoint when you need structured, machine-readable extraction of numeric statements from spoken audio or video — especially for meetings, calls, interviews, or dictations where figures like prices, percentages, dates, and quantities are spoken and need to be captured with their context and timestamps. It is distinctly useful over generic transcription when the goal is specifically to parse and normalize numbers rather than produce a full transcript, and its hallucination guard (numbers must be anchored to a real spoken sentence) makes it suitable for financial, legal, or compliance-sensitive use cases.

## Known failure modes

- Recording exceeds 60-minute maximum — request rejected
- File URL is inaccessible or returns non-audio/video content — extraction fails
- duration_minutes mismatch with actual file length may affect billing accuracy
- Language detection fails on low-quality or multilingual audio — incorrect classification
- No numeric statements found — returns empty result set
- Base64 payload exceeds 12MB limit — must use URL instead
- Payment not fulfilled via x402 — 402 response returned

## How this service works

Takes a call, meeting, dictation or interview of up to 60 minutes and returns its numeric statements as structured data: kind (amount, percent, date, time, duration, quantity, phone, year), the wording as spoken, a normalised value, unit or currency, what the number refers to, the sentence it comes from and its timestamp. A language model classifies; each value must point to an existing sentence whose wording contains it, otherwise it is dropped and counted — so numbers are never invented.

## Output

A structured array of numeric statements found in the recording, each containing: the kind (amount, percent, date, time, duration, quantity, phone, year, or other), the wording as spoken, a normalized value, the associated unit or currency, what the number refers to, the full source sentence containing it, and its timestamp within the recording. Numbers not traceable to an actual spoken sentence are excluded; a count of dropped values is also returned.

## Request schema (JSON Schema)

```json
{
 "type": "object",
 "properties": {
  "kinds": {
   "type": "array",
   "items": {
    "enum": [
     "amount",
     "percent",
     "date",
     "time",
     "duration",
     "quantity",
     "phone",
     "year",
     "other"
    ],
    "type": "string"
   },
   "description": "Only these kinds. Default: all.",
   "uniqueItems": true
  },
  "source": {
   "type": "object",
   "properties": {
    "url": {
     "type": "string",
     "maxLength": 2048,
     "description": "Public http(s) URL of the file."
    },
    "filename": {
     "type": "string",
     "maxLength": 260
    },
    "content_base64": {
     "type": "string",
     "maxLength": 12000000,
     "description": "The file inline, base64. Larger files by url."
    }
   },
   "description": "The recording. Either url or content_base64.",
   "additionalProperties": false
  },
  "language": {
   "type": "string",
   "maxLength": 12,
   "description": "Spoken language (BCP-47). Omit to detect."
  },
  "duration_minutes": {
   "type": "integer",
   "maximum": 60,
   "minimum": 1,
   "description": "Length in started minutes — this is what you are billed for."
  }
 }
}
```

## More

- Live health (JSON, refreshed every minute): https://www.zero.xyz/c/halowerk-numbers-extract-9c02a3ca/health.json
- [Zero catalog index](https://www.zero.xyz/llms.txt)
- [Other services from medien.halowerk.com](https://www.zero.xyz/host/medien.halowerk.com/llms.txt)