Handy Tool 2.4.0

Added

• Your own STT server, with speakers (Advanced settings › STT server). Connect any OpenAI-compatible speech-to-text server on this PC or your network; Test lists its models. With "Tell the speakers apart" on, History keeps the take with speakers ("A: …" per turn) next to the plain text, with a Text / Speakers switch, and you choose which one is pasted, each with its own prefix and suffix. Your custom words are sent to the server and corrected in its answer.
• A second and third model if the first fails (Models › If your dictation model fails). When your dictation model cannot transcribe a take (server down, network dropped, request failed) the next one does; an online model is checked as soon as the take starts. History shows which model transcribed.
• Parakeet models from Hugging Face (Models › Parakeet from Hugging Face). Search, check and add newer or fine-tuned Parakeet ONNX models; they download file by file (resumable, checksum-checked) and run on this PC.
• A model's settings from its card: a gear on model cards with settings, and a click on the card of the model in use opens them.

Changed

• History's "Transcribe again" has a new icon.