Fix distilled models

These were broken due to some logic errors in the codepath which
acquires models from huggingface.

Distilled large-v2 seems promising as a new default model.

3 files changed, 6 insertions, 7 deletions

diff --git a/GUI/GUI/GUI/Frame.cpp b/GUI/GUI/GUI/Frame.cpp
index 66c3d4e..908a54c 100644
--- a/GUI/GUI/GUI/Frame.cpp
+++ b/GUI/GUI/GUI/Frame.cpp
@@ -461,12 +461,12 @@ namespace {
         "base",

         "small.en",

         "small",

-        "yumfood/whisper_distil_medium_en_ct2",

+        "Systran/faster-distil-whisper-medium.en",

         "medium.en",

         "medium",

+        "Systran/faster-distil-whisper-large-v2",

         "large-v1",

         "large-v2",

-        "yumfood/whisper_distil_large_v2_ct2",

     };

     const size_t kNumModelChoices = sizeof(kModelChoices) / sizeof(kModelChoices[0]);

     constexpr int kModelDefault = 2;  // base.en

diff --git a/Scripts/text_to_text_demo.py b/Scripts/text_to_text_demo.py
index d5f0ada..4810361 100644
--- a/Scripts/text_to_text_demo.py
+++ b/Scripts/text_to_text_demo.py
@@ -7,8 +7,6 @@ from pythonosc import udp_client
 
 import generate_utils
 import osc_ctrl
-import paging
-import pythonosc
 import time
 
 class AppConfig:
diff --git a/Scripts/transcribe_v2.py b/Scripts/transcribe_v2.py
index 7655d4b..32652df 100644
--- a/Scripts/transcribe_v2.py
+++ b/Scripts/transcribe_v2.py
@@ -422,7 +422,7 @@ class Whisper:
         if cfg["use_cpu"]:
             model_device = "cpu"
 
-        download_it = os.path.exists(model_root)
+        already_downloaded = os.path.exists(model_root)
         if '/' in model_str:
             hf_hub_download(repo_id=model_str, filename='model.bin',
                     local_dir=model_root)
@@ -430,14 +430,15 @@ class Whisper:
                     local_dir=model_root)
             hf_hub_download(repo_id=model_str, filename='config.json',
                     local_dir=model_root)
-        if download_it:
+            already_downloaded = True
+        if already_downloaded:
             model_str = model_root
         self.model = WhisperModel(model_str,
                 device = model_device,
                 device_index = cfg["gpu_idx"],
                 compute_type = cfg["compute_type"],
                 download_root = model_root,
-                local_files_only = download_it)
+                local_files_only = already_downloaded)
 
     def transcribe(self, frames: bytes = None) -> typing.List[Segment]:
         if frames is None: