CHANGELOG.md

Changelog

v2.4.0 (2025-09-30)

Full Changelog: v2.3.0...v2.4.0

• feat(api): api update (6e0a4dd)
• feat(api): api update (cc7f909)
• chore(internal): update pydantic dependency (a4c5af5)
• chore(types): change optional parameter type from NotGiven to Omit (7c03fb8)
• chore: do not install brew dependencies in ./scripts/bootstrap by default (75d2397)
• chore: improve example values (5c0919c)

v2.3.0 (2025-09-03)

Full Changelog: v2.2.1...v2.3.0

• replace ava with leah (47bb854)
• avoid newer type syntax (773b08e)
• chore(internal): update pyright exclude list (8c1bd1d)
• feat(api): api update (7820c00)
• chore(internal): add Sequence related utils (f1f239b)

v2.2.1 (2025-08-21)

Full Changelog: v2.2.0...v2.2.1

• chore(internal): fix ruff target version (08c3067)
• chore: update @stainless-api/prism-cli to v5.15.0 (1634751)
• chore(internal): update comment in script (6649334)

v2.2.0 (2025-08-05)

Full Changelog: v2.1.0...v2.2.0

• feat: add support for resets in websocket speech sessions

v2.1.0 (2025-08-05)

Full Changelog: v2.0.0...v2.1.0

• fix(parsing): ignore empty metadata (2671330)
• fix(parsing): parse extra field types (e652a62)
• chore(project): add settings file for vscode (9d450b1)
• feat(api): api update (f0ef880)

July 17, 2025

2.0.0

• BREAKING CHANGES: The new v2 SDK provides more streaming functionality, a more modern, type-safe interface with better error handling, and improved performance. To migrate from the legacy v1 SDK, please update your code to use the new behavior or pin to a previous version if preferred. More details in the migration guide.

1.1.0

Jan 3, 2024

• synthesize_streaming will now return a buffer_empty boolean when extras are requested. This can be used to determine when the server has no more audio to send after the client has sent a flush message.
• A bug that caused create_voice to fail in Python 3.8 and earlier has been fixed.

1.0.1

Nov 18, 2023

• Changes the owner type in list_voices from 'lmnt' to 'system' to match the LMNT REST API.

1.0.0

Nov 14, 2023

• Uses the stable v1 LMNT REST API.
• BREAKING CHANGES: please update your code to use the new behavior or pin to a previous version if preferred.
  • Default audio encoding format in synthesize is now mp3 (previously wav). Format can be specified by adding format='wav' or format='mp3' to the synthesize call.
  • list_voices now returns a list of voice dictionaries for simplicity of return values and ease of use. Previously it returned a dictionary with key voices which contained a dictionary of voice dictionaries keyed by their voice id.
  • synthesize no longer returns just the binary audio data. It instead always returns a dictionary with keys audio, durations (optional), and seed (optional).
  • synthesize_streaming no longer returns an aiohttp WSMessage. It instead returns a dictionary with keys audio, durations (optional), and seed (optional).
• Adds support for creating, updating, and deleting voices as well as getting account details.

0.2.1

Oct 12, 2023

• Removes the default seed of synthesize (previously 0). When seed is unspecified, synthesize will now use a random seed.
• The durations dictionary has been updated. Now includes each word itself and changes the units of duration from the number of samples to seconds.

0.2.0

Sep 7, 2023

• Add support for durations.

0.1.0

Aug 11, 2023

• Initial release