whisper.cpp 0.3.1 → 0.3.3

checksums.yaml

@@ -1,7 +1,7 @@
 ---
 SHA256:
-  metadata.gz: 561b7b824c1f631537d681fb1e5a1eafe20546a8468705ecd2ba3989e0062c69
-  data.tar.gz: e047a8ab099358095100a55cd30aca398bbe8b8c4c55cfb0713cdf33e87d2e1a
+  metadata.gz: fb6cb421d7d00b2f9ff5f93f543bf468d5d8c0754befe434b25e1b0e2b3c80a4
+  data.tar.gz: cbd9ea39b8f408692b857cf0302d945392c7aec620834aca340e9c6cb6c900c8
 SHA512:
-  metadata.gz: 80c60958d31df322ac570c97ca6d758278f8711d9bfff371be34e8159ffcadc2facf149aa4c7b55de49a2d2d8812b570d0ddd68e543172301f930fe494a4243b
-  data.tar.gz: 02fd686ff1857adf234c53fa0719035e797a82bed4b165e54004c0b7c903728eb7d8c44273b3f46e6ff21a32df44e29986092b72ccfd853324749e639fa303ed
+  metadata.gz: 9a2671addade694795b2a76b1db257db8ffab3ebd9b3266df237a45a73d85b8f7c5f3ebae78bc33c4c38fb869397ca7f8082b909a5d8aebe0818f0174ac01fdd
+  data.tar.gz: d150ab03e621e4b2128ed34792d6968ad5e8fccf599150be240344f97b10c58ccbcac9a8d2022356d97c31c856004c18d0f8fc4edf32ceacfbe2dac14805de39

data/lib/whisper/model.rb

@@ -8,84 +8,80 @@ module Whisper
     def initialize(model_path)
       @model_path = model_path
       @ctx = nil
+      @state = nil
+      init_whisper_context
+      init_whisper_state
     end
 
     def transcribe_from_file(audio_file_path, format: 'plaintext', **params)
       # Load audio file and convert to float array
-      audio_data = Whisper::AudioProcessor.convert_to_float_array(audio_file_path)
-      transcribe_from_audio_data(audio_data, format: format, **params)
+      audio_data = Whisper::AudioProcessor.convert_to_float_array audio_file_path
+      transcribe_from_audio_data audio_data, format: format, **params
     end
 
     def transcribe_from_audio_data(audio_data, format: 'plaintext', **params)
-      init_whisper_context(params)
-
       # Prepare full params
-      full_params = default_full_params(params)
+      full_params = default_full_params params
 
       # Prepare audio data pointer
       n_samples = audio_data.size
-      samples_ptr = FFI::MemoryPointer.new(:float, n_samples)
-      samples_ptr.write_array_of_float(audio_data)
+      samples_ptr = FFI::MemoryPointer.new :float, n_samples
+      samples_ptr.write_array_of_float audio_data
 
-      # Call the whisper_full_parallel function
-      n_processors = params.fetch :n_processors, ENV['WHISPER_N_PROCS']&.to_i || 1
-      result = Whisper.whisper_full_parallel(@ctx, full_params, samples_ptr, n_samples, n_processors)
+      # Call the whisper_full_with_state function
+      result = Whisper.whisper_full_with_state @ctx, @state, full_params, samples_ptr, n_samples
       raise 'Transcription failed' if result != 0
 
       # Retrieve detected language
-      lang_id = Whisper.whisper_full_lang_id(@ctx)
-      language = Whisper.whisper_lang_str(lang_id)
+      lang_id = Whisper.whisper_full_lang_id_from_state @state
+      language = Whisper.whisper_lang_str lang_id
 
       # Retrieve the transcription output
-      n_segments = Whisper.whisper_full_n_segments(@ctx)
-      output = format_transcription(format, n_segments: n_segments)
+      n_segments = Whisper.whisper_full_n_segments_from_state @state
+      output = format_transcription format, n_segments: n_segments
 
-      TranscriptionResult.new(language, output)
+      TranscriptionResult.new language, output
     end
 
     def close
-      Whisper.whisper_free(@ctx) unless @ctx.nil?
+      Whisper.whisper_free_state @state unless @state.nil?
+      Whisper.whisper_free @ctx unless @ctx.nil?
     end
 
     private
 
-    def init_whisper_context(params)
+    def init_whisper_context params = {}
       return unless @ctx.nil?
 
       ctx_params = Whisper.whisper_context_default_params
 
-      # Set user-provided context params
-      user_ctx_params = params.fetch(:context_params, {})
-      user_ctx_params.each do |key, value|
-        if ctx_params.members.include?(field)
-          ctx_params[key] = value
-        else
-          warn "Unknown context_param field: #{field}"
-        end
+      params.select{ |k, _| ctx_params.members.include? k }.each do |key, value|
+        ctx_params[key] = value
       end
       ctx_params[:gpu_device] = ENV['WHISPER_GPU']&.to_i || 0
 
       # Initialize context
-      @ctx = Whisper.whisper_init_from_file_with_params(@model_path, ctx_params)
+      @ctx = Whisper.whisper_init_from_file_with_params @model_path, ctx_params
       raise 'Failed to initialize Whisper model' if @ctx.null?
     end
 
-    def default_full_params(params)
+    def init_whisper_state
+      @state = Whisper.whisper_init_state @ctx
+      raise 'Failed to initialize Whisper state' if @state.null?
+    end
+
+    def default_full_params params = {}
       # Get default full params
-      strategy = params.fetch(:sampling_strategy, Whisper::WHISPER_SAMPLING_GREEDY)
-      full_params = Whisper.whisper_full_default_params(strategy)
+      strategy = params.fetch :sampling_strategy, Whisper::WHISPER_SAMPLING_GREEDY
+      full_params = Whisper.whisper_full_default_params strategy
 
-        # Set translate to false to prevent translation to English
+      # Set translate to false to prevent translation to English
       full_params[:translate] = false
+      full_params[:language] = FFI::MemoryPointer.from_string 'auto'
 
       # Set user-provided full params
-      user_full_params = params.fetch(:full_params, {})
-      user_full_params.each do |key, value|
-        if full_params.members.include?(field)
-          full_params[key] = value
-        else
-          warn "Unknown full_param field: #{field}"
-        end
+      params.select{ |k, _| full_params.members.include? k }.each do |key, value|
+        full_params[key] = value
       end
 
       full_params
@@ -96,17 +92,17 @@ module Whisper
       case format.downcase
       when 'plaintext'
         n_segments.times do |i|
-          segment_text = Whisper.whisper_full_get_segment_text(@ctx, i)
+          segment_text = Whisper.whisper_full_get_segment_text_from_state @state, i
           output += segment_text
         end
       when 'srt'
         n_segments.times do |i|
-          start_time = Whisper.whisper_full_get_segment_t0(@ctx, i) / 100.0
-          end_time = Whisper.whisper_full_get_segment_t1(@ctx, i) / 100.0
-          segment_text = Whisper.whisper_full_get_segment_text(@ctx, i)
+          start_time = Whisper.whisper_full_get_segment_t0_from_state(@state, i) / 100.0
+          end_time = Whisper.whisper_full_get_segment_t1_from_state(@state, i) / 100.0
+          segment_text = Whisper.whisper_full_get_segment_text_from_state @state, i
 
           output += "#{i + 1}\n"
-          output += "#{format_time_srt(start_time)} --> #{format_time_srt(end_time)}\n"
+          output += "#{format_time_srt start_time} --> #{format_time_srt end_time}\n"
           output += "#{segment_text.strip}\n\n"
         end
       else
@@ -120,7 +116,7 @@ module Whisper
       minutes = ((seconds % 3600) / 60).to_i
       secs = (seconds % 60).to_i
       millis = ((seconds - seconds.to_i) * 1000).to_i
-      format('%02d:%02d:%02d,%03d', hours, minutes, secs, millis)
+      format '%02d:%02d:%02d,%03d', hours, minutes, secs, millis
     end
   end
 end

data/lib/whisper/version.rb

@@ -1,5 +1,5 @@
 module Whisper
 
-  VERSION = '0.3.1'
+  VERSION = '0.3.3'
 
 end

metadata

@@ -1,14 +1,14 @@
 --- !ruby/object:Gem::Specification
 name: whisper.cpp
 version: !ruby/object:Gem::Version
-  version: 0.3.1
+  version: 0.3.3
 platform: ruby
 authors:
 - Braulio Oliveira
 autorequire:
 bindir: bin
 cert_chain: []
-date: 2024-10-12 00:00:00.000000000 Z
+date: 2024-10-13 00:00:00.000000000 Z
 dependencies:
 - !ruby/object:Gem::Dependency
   name: ffi