Rumble 2 Flash Open
Reson8 tested the smallest possible speech model that maintains accuracy for complex, noisy audio while running on devices from voice notes to full call archives.
reson8.devCrawled Nov 20, 2026
Reson8 tested the smallest possible speech model that maintains accuracy for complex, noisy audio while running on devices from voice notes to full call archives.
How to measure whether a transcript is good enough.
The page describes how to train a speech model to understand specific languages.
Depth in the languages Europe actually speaks.
The mechanics of turning audio into reliable transcripts.
Speech recognition inside AI that talks back.
The Reson8 team shares benchmarks, speech models, and technical updates in automatic speech recognition (ASR).
364 speech recognition terms defined, A to Z.
An open reference on speech recognition: how ASR works, how to measure it, domain adaptation, European languages, voice agents, and handling voice data under GDPR.
Automatic speech recognition (ASR) converts spoken audio into written text by analyzing sound waves and using neural networks to predict words.
Voice AI software processes speech for input or output, handling audio transport, speech detection, speech-to-text, reasoning via large language models, and text-to-speech.
Reson8 pricing: Build, Launch, Scale and Enterprise plans. Rumble-2 speech-to-text from €0.14 an hour, billed per second, with EU-only processing.
Speech-to-text pricing varies by audio volume, streaming vs. batch processing, request rounding, and vendor-specific rates beyond the listed price per minute.
The speech-to-text system adapts to user vocabulary in real time and is designed for European languages.
Reson8's Data Processing Agreement: GDPR Article 28 terms for audio and personal data. EU-only processing, zero audio retention, no model training.
EU data residency ensures audio, transcripts, and metadata are stored and processed within the European Union or European Economic Area, aligning with GDPR restrictions on third-country transfers.
Under GDPR, voice recordings of identifiable individuals are personal data, and transcripts may also qualify. Voice data becomes biometric when processed to uniquely identify someone, often alongside other special-category data.
Resonate Labs B.V. (Reson8) outlines how it collects, uses, and protects personal data in its GDPR-compliant privacy policy for website visitors.
Handling voice data in regulated environments.
Zero data retention means a speech provider processes audio for transcription without storing it or related customer content post-request, meeting GDPR’s storage and data minimization principles.
A data processing agreement (DPA) is required under GDPR when one organization processes personal data on behalf of another, such as when call recordings or dictations are sent to a speech-to-text API.
Read Reson8 Labs' terms and conditions covering service agreements, data processing, intellectual property, liability, and acceptable use policies.