Hey everyone,
We’ve just released a new Call Transcriptions module in FS PBX, bringing AI-powered transcription and summarization directly into your PBX — no external scripts, no manual uploads, no patchwork integrations.
Here’s what this release includes
Inside the CDR page, you can now click on any recorded call and instantly generate a high-accuracy transcript. Auto transcriptions are also available for all new calls
FS PBX uses AssemblyAI for transcriptions — and here’s why we chose them:
They also don’t accept URL-based audio, which means recordings would need to be uploaded — not ideal for privacy or compliance.
We still use them for summaries (see below).
Best transcription accuracy
Most reliable speaker diarization
Fastest processing times
HIPAA compliance out of the box
PII redaction built-in
Ability to send just a URL link instead of uploading call recordings
Webhook notifications, so FS PBX receives a completion event instantly (no polling, no wasted resources)
The privacy aspect was a major factor:
Because AssemblyAI accepts secure recording URLs, we avoid uploading sensitive audio files to third-party servers altogether. They also let us ensure transcripts are deleted on their side after processing, and PII redaction helps keep everything compliant.
Given all of this, AssemblyAI was the clear choice for launch. We will continue to monitor other LLMs, and as they become more mature, we will add them too.
You’ll get:
Perfect for quick reviews, QA, training, and follow-up workflows.
We’ve just released a new Call Transcriptions module in FS PBX, bringing AI-powered transcription and summarization directly into your PBX — no external scripts, no manual uploads, no patchwork integrations.
Here’s what this release includes
AI Voice Transcriptions (AssemblyAI)
Inside the CDR page, you can now click on any recorded call and instantly generate a high-accuracy transcript. Auto transcriptions are also available for all new calls
FS PBX uses AssemblyAI for transcriptions — and here’s why we chose them:
Why AssemblyAI?
During development, we tested multiple leading providers:OpenAI
We love OpenAI (and FS PBX integrates tightly with them), but when we began this project, they didn’t support speaker diarization. They later added it, so we gave them another shot — but in our testing, the transcription accuracy still wasn’t close to AssemblyAI’s output.They also don’t accept URL-based audio, which means recordings would need to be uploaded — not ideal for privacy or compliance.
We still use them for summaries (see below).
Google Gemini
Gemini had diarization, but accuracy wasn’t strong and processing times were inconsistent and slow.AssemblyAI
AssemblyAI consistently delivered the: Best transcription accuracy Most reliable speaker diarization Fastest processing times HIPAA compliance out of the box PII redaction built-in Ability to send just a URL link instead of uploading call recordings Webhook notifications, so FS PBX receives a completion event instantly (no polling, no wasted resources)The privacy aspect was a major factor:
Because AssemblyAI accepts secure recording URLs, we avoid uploading sensitive audio files to third-party servers altogether. They also let us ensure transcripts are deleted on their side after processing, and PII redaction helps keep everything compliant.
Given all of this, AssemblyAI was the clear choice for launch. We will continue to monitor other LLMs, and as they become more mature, we will add them too.
Smart AI Summaries (OpenAI)
Call summaries, on the other hand, were best handled by the cheapest OpenAI model, so we decided to use them in this case. Once a transcription completes, FS PBX automatically sends it to OpenAI for summarization.You’ll get:
- Key Points
- Action Items
- Sentiment
Perfect for quick reviews, QA, training, and follow-up workflows.
Video Walkthrough
We also published a complete video demo showing everything — recordings → transcription → summary → setup:
Last edited:
