Note

Medium Risk
Adds a new audio.transcriptions.create API surface and updates generated client/docs/lockfiles, which could affect SDK behavior and dependency compatibility for consumers.

Overview
Regenerates the Python SDK with newer Fern tooling (cliVersion/generatorVersion) and bumps the package version to 5.21.0.

Adds a new audio namespace with audio.transcriptions.create (sync + async, plus raw clients/types) hitting POST v2/audio/transcriptions via multipart upload.

Updates generated reference docs and lockfile dependencies (e.g., huggingface-hub, hf-xet, certifi, charset-normalizer, tomli) and refreshes various type hints/usages in the docs (notably stream parameters and Sequence→List in signatures).

^{Written by Cursor Bugbot for commit c2fcc4f. This will update automatically on new commits. Configure here.}