Lorem ipsum dolor sit amet, consectetur adipiscing elit, sed do eiusmod tempor incididunt ut labore et dolore magna aliqua. Ut enim ad minim veniam, quis nostrud exercitation ullamco laboris nisi ut aliquip ex ea commodo consequat. Duis aute irure dolor in reprehenderit in voluptate velit esse cillum dolore eu fugiat nulla pariatur.
MiFID II and FCA call recording: compliance for voice transcription in finance
TL;DR: Financial firms operating under MiFID II and FCA jurisdictions must maintain searchable, high-accuracy records of all client-related voice communications, including remote and mobile calls under FCA Market Watch 66. Engineering and product teams building for these obligations typically implement dedicated cloud infrastructure with defined data residency, controls that prevent customer audio from being used to retrain models on Growth and Enterprise plans, and transcription accurate enough that the resulting records hold up under regulatory review. Transcription and speaker attribution errors are not product quality issues in this context. They are audit trail failures, and regulators treat them as such.
Why speech-to-text accuracy matters upstream of your LLM
TL;DR: Downstream LLM performance is ceiling-bounded by upstream transcription accuracy. A transcript with a meaningful error rate doesn't produce proportionally degraded summaries or CRM entries. It produces outputs where hallucinated names, inverted logic, and misattributed speaker turns compound silently into every downstream system that reads them. Prompt engineering cannot recover information the STT layer never captured. Gravite cut call quality review time by 93%, from 15 minutes to 1 minute per call, once transcript accuracy was high enough to trust the output without manual verification.
TL;DR: Telephony audio is constrained to 8kHz narrowband frequencies, stripping away the high-frequency spectral energy that generic 16kHz STT models require. Standard upsampling cannot recover phonemes that were never captured at the source, leading to transcription errors that compound silently into downstream NLU failures. Solaria-3 ranks #1 on Switchboard, the most challenging conversational telephony dataset, ahead of AssemblyAI, ElevenLabs, Deepgram, Mistral, and Speechmatics, ensuring your downstream LLM pipelines receive clean, structured data from real-world noisy call audio.
How to build a voice-to-text Discord bot with Gladia real-time transcription API
Published on Sep 21, 2023
Discord, the leading communication platform for gamers and communities, is designed for seamless communication with other users, be it through text channels, DMs, 1-1 calls or even collective voice channels.
Based on multiple request from our Discord members, we’ve built a custom JavaScript bot that makes use of Gladia’s live transcription API to transcribe speech in real time directly on the Discord server.
What can you do with Discord bot?
First, you can transcribe voice in real time directly on Discord’s voice channels. Ex. you’re streaming a game on Discord and want to access some learnings and tips received during the sessions. Or, you’re having your group gathers on the platform and want to be able to review the talking points after – just like with any other virtual meeting platform.
Beyond that, a bot like this could be used for real-time moderation to flag hate speech and ban users. With additional tools like ChatGPT, you could also create command-based notes to provide meeting summaries and helps you catch up with meetings you may have missed.
How to implement the Discord.js v14 bot + Gladia real-time transcription
Step 1: Register your bot
Create a Discord bot that you'd like to use for transcription. If you’ve never built one before, here’s a useful resource to help.
First, install all the required package by running:
npm install
Then, you will to setup the index.js script with your Discord keys, guild ID (Server ID), and the Voice Channel ID.
Step 2: Retrieve API key
Sign up for our speech-to-text API at app.gladia.io and obtain your API key. Documentation for Gladia live transcription can be found here.
Step 3: Code integration
Once everything is set up properly, simply run:
npm run start YOUR_GLADIA_TOKEN
Your bot should then join the channel corresponding to the channel ID you configured in the index.js file.
Step 4: Configure Discord permissions
Make sure your bot is invited on the server;
Give the bot the required voice permissions.
Bear in mind that the current v1 implementation of the bot is not fully optimized, so you might experiences inaccuracy regarding language changes & words.
We hope you enjoyed this short tutorial. Given how much audio data still goes to wasted, we’re always curious to explore the many ways in which transcription tech can be used to remedy that. Let us know if you went on to build a bot or used our API for others apps on Discord or beyond, we’d love to hear from you.
About Gladia
At Gladia, we built an optimized version of Whisper in the form of an API, adapted to real-life use cases and distinguished by exceptional accuracy, speed, extended multilingual capabilities and state-of-the-art features, including speaker diarization and word-level timestamps.
Contact us
Your request has been registered
A problem occurred while submitting the form.
Read more
Speech-To-Text
MiFID II and FCA call recording: compliance for voice transcription in finance
Speech-To-Text
Why speech-to-text accuracy matters upstream of your LLM
.png)
