Transcribr for FileMaker

Speech-to-text for
FileMaker. 100% local.

A single-file transcription tool that runs entirely on your machine. No cloud. No API keys. No data leaving the device. Just drop it in, call it from FileMaker, get text back.

whisper.cpp engine
Core ML accelerated
Zero dependencies
Notify Me See How It Works
Use Cases

What you can build

Meeting Notes

Record client meetings in FileMaker, get searchable transcripts automatically attached to the record.

Voice Notes

Let users dictate instead of type. Attach transcribed voice memos to any record in your solution.

Call Logging

Transcribe recorded phone calls and link the text to contact records. Fully searchable call history.

Dictation Interface

Real-time speech-to-text inside a FileMaker layout via web viewer. Live text as the user speaks.

Archive Processing

Batch-transcribe years of recorded meetings, voicemails, or interviews. Point at a folder, walk away.

Accessibility

Add captions or searchable text to audio content. Make your FileMaker solutions more accessible.

Why This Exists

The alternative is painful

Integrating speech-to-text today

✕ Sign up for a cloud API (OpenAI, Google, AWS)
✕ Manage API keys and authentication tokens
✕ Handle per-minute billing and usage limits
✕ Upload client audio to a third party
✕ Parse complex, inconsistent API responses
✕ Deal with rate limits, downtime, and API changes

With Transcribr

1
Download the executable for your platform
2
Run it — server starts on localhost:8001
3
Call Insert from URL from FileMaker
4
Parse the JSON response into your fields
Working transcription in your FileMaker solution in minutes, not weeks. No setup wizard. No configuration files. No cloud accounts.
Four Modes

Transcribe however you need to

One server, four ways to use it. Each maps directly to a FileMaker integration pattern.

Mode 01

Single File

Send an audio file, get a transcript back. Any format — MP3, WAV, M4A, OGG, FLAC and more. All files converted automatically.

FileMaker: One Insert from URL call. Returns JSON you parse with JSONGetElement. One script step, one result.
Mode 02

Batch Processing

Point it at a folder of audio files. It processes every file asynchronously and reports progress. Each result is stored and retrievable individually.

FileMaker: One API call to start the batch. Poll the status endpoint to show progress in your UI. Ideal for audio archive migration or overnight processing.
Mode 03

Watch Mode

Tell it to watch a folder. Any audio file that appears is automatically transcribed — no further interaction required. Set it and forget it.

FileMaker: Export recordings to the watched folder. A scheduled script checks for new transcripts on a timer and imports them automatically.
Mode 04

Real-Time Streaming

WebSocket connection for live audio. Stream from a microphone, get transcription results back as the speaker talks. Sub-sentence latency.

FileMaker: Connect via web viewer. Stream microphone audio, display live text. Ideal for dictation, call logging, or live meeting notes.
Five Models

Choose your trade-off

Select the model per request. Use tiny for live streaming, large for critical recordings. Mix and match as you go.

Model Speed Accuracy Memory Best For
tiny
Good ~75 MB Quick drafts, real-time streaming, low-spec machines
base Default
Better ~150 MB Good balance for most use cases
small
Good+ ~500 MB Improved accuracy when speed isn't critical
medium
Very good ~1.5 GB High-accuracy batch processing
large
Best ~3 GB Maximum accuracy for critical transcriptions
Response Format

What you get back

Every request returns clean JSON you can parse with native FileMaker functions. No plugins required.

{
  "transcription": "The full text as a single string.",
  "segments": [
    {
      "index": 0,
      "start_time": 0.00,
      "end_time": 4.52,
      "text": "The full text"
    },
    {
      "index": 1,
      "start_time": 4.52,
      "end_time": 7.10,
      "text": "as a single string."
    }
  ],
  "audio_duration_seconds": 7.10,
  "processing_time_seconds": 1.23,
  "model": "base"
}
Response Format

Clean JSON. Native to FileMaker.

Full Text
Complete transcription in one string. Drop it straight into a text field.
Timestamped Segments
Every phrase with start and end times in seconds. Build searchable, navigable transcripts.
Processing Metrics
Audio duration and transcription time. Monitor performance and set user expectations.
No Plugins
Parse everything with JSONGetElement, JSONListKeys, and native FileMaker functions only.
Under the Hood

Technical details

Engine
whisper.cpp — C++ implementation of OpenAI Whisper
Delivery
Single-file executable — no installation, no dependencies
API
REST + WebSocket, JSON responses
Audio Formats
MP3, WAV, M4A, OGG, FLAC, AAC, WMA + more — converted automatically
Platforms
macOS (Apple Silicon + Intel) and Linux. Windows coming soon.
Acceleration
Core ML on Apple Silicon — significantly faster inference
Network
Runs on localhost — no internet required
Documentation
Built-in Swagger UI for interactive API exploration
Coming Soon

Don't miss it

We're putting the finishing touches on Transcribr for FileMaker. Drop your email and we'll let you know the moment it's ready.

Notify Me
No spam. One email when it launches.