aeneas, caption updates

3 files changed, 26 insertions, 9 deletions

diff --git a/roles/caption/tasks/aeneas.yml b/roles/caption/tasks/aeneas.yml
new file mode 100644
index 0000000..e280e7e
--- /dev/null
+++ b/roles/caption/tasks/aeneas.yml
@@ -0,0 +1,12 @@
+- name: Install packages
+  package:
+    name:
+      - espeak
+      - python3-espeak
+      - ffmpeg
+      - libespeak-dev
+- name: Install aeneas
+  ansible.builtin.pip:
+    name:
+      - numpy
+      - aeneas
diff --git a/roles/caption/tasks/main.yml b/roles/caption/tasks/main.yml
index 353b83b..3fa997c 100644
--- a/roles/caption/tasks/main.yml
+++ b/roles/caption/tasks/main.yml
@@ -17,6 +17,8 @@
       - tqdm
       - torchaudio
       - num2words
+- name: Set up aeneas
+  include: aeneas.yml
 - name: Create group
   group:
     name: "{{ emacsconf_group }}"
@@ -31,7 +33,7 @@
     path: "{{ emacsconf_caption_dir }}"
     state: directory
 - name: Copy the script for processing the files
-  tags: process-captions
+  tags: process-captions, wip
   template:
     src: process-captions.py
     dest: "{{ emacsconf_caption_dir }}/process-captions.py"
@@ -63,9 +65,9 @@
     group: "{{ emacsconf_group }}"
     mode: "g+rwX"
     recurse: true
-- name: Restart caption monitoring service
-  become: true
-  service:
-    name: captions
-    enabled: true
-    state: started
+# - name: Restart caption monitoring service
+#   become: true
+#   service:
+#     name: captions
+#     enabled: true
+#     state: started
diff --git a/roles/caption/templates/process-captions.py b/roles/caption/templates/process-captions.py
index f3f317d..6f6b013 100755
--- a/roles/caption/templates/process-captions.py
+++ b/roles/caption/templates/process-captions.py
@@ -47,6 +47,7 @@ TRIM_AUDIO = False
 MODEL = os.environ.get('MODEL', 'large')  # Set to tiny for testing
 WORK_DIR = "{{ emacsconf_caption_dir }}"
 JSON_FILE = os.path.join(WORK_DIR, 'talks.json')
+DO_SRV2 = False
 
 # ----------------------------------------------------------------
 
@@ -94,7 +95,7 @@ def get_files_to_work_on(directory):
         else:
             val['base'] = os.path.join(os.path.dirname(val['video'] or val['audio']),
                                        base_name(val['video'] or val['audio']))
-        if ALWAYS or (not 'vtt' in val or not 'srv2' in val):
+        if ALWAYS or (not 'vtt' in val or (DO_SRV2 and not 'srv2' in val)):
             if not 'audio' in val and 'video' in val:
                 # No audio, need to convert it
                 val = extract_audio(val)
@@ -141,6 +142,8 @@ def generate_captions(work):
     result = clean_up_timestamps(result)
     with open(new_file, 'w') as vtt:
         whisper.utils.write_vtt(result['segments'], file=vtt)
+    with open(work['base'] + '.txt')
+        whisper.utils.write_txt(result['segments'], file=txt)
     work['vtt'] = new_file
     if 'srv2' in work: del work['srv2']
     return work
@@ -209,7 +212,7 @@ if len(needs_work) > 0:
                 if work['audio']:
                     if ALWAYS or not 'vtt' in work:
                         work = generate_captions(work)
-                    if ALWAYS or not 'srv2' in work:
+                    if DO_SRV2 and (ALWAYS or not 'srv2' in work):
                         work = generate_srv2(work)
                         #     print("Aligning words", audio_file, datetime.datetime.now())
                         #     word_cuts = align_words(cuts)