Klang raises €1.32M and releases an open Swedish speech-to-text model
The AMW Read
The release adds a Swedish dialect-focused open speech model to the player map, while its speed claim and commercial impact remain unverified.
Klang raises €1.32M and releases an open Swedish speech-to-text model
Swedish conversation AI company Klang has raised €1.32 million and released Pianissimo, an open speech-to-text model trained for Swedish and its dialects. Pianissimo builds on Nvidia Parakeet and runs locally on a standard laptop. Klang says its tests found accuracy comparable to the National Library of Sweden's KB Whisper Large while running roughly 60 times faster. The company plans to use the funding for AI development and international expansion.
The release puts a language-specific model on the player map for AI speech tools. Its commercial question is whether developers working with Swedish conversations value dialect coverage, local execution and speed enough to choose Pianissimo over an established alternative. The reported accuracy and speed comparison is a company claim; independent evaluation would be needed to establish how it holds across dialects, recording conditions and hardware.
For builders, the practical next step is to test Pianissimo on their own Swedish audio, measuring transcription errors and latency on the devices they intend to use. Investors can watch whether the open release leads to adoption and a clear path to revenue as Klang expands. The round funds that effort, but the information provided does not establish customer demand or commercial performance.