ref #9 : add API documentation in whisper.h

main.cpp

@@ -230,25 +230,25 @@ int main(int argc, char ** argv) {
 
             // print result
             if (!wparams.print_realtime) {
-                fprintf(stderr, "\n");
+                printf("\n");
 
                 const int n_segments = whisper_full_n_segments(ctx);
                 for (int i = 0; i < n_segments; ++i) {
                     const char * text = whisper_full_get_segment_text(ctx, i);
 
                     if (params.no_timestamps) {
-                        fprintf(stderr, "%s", text);
+                        printf("%s", text);
                         fflush(stdout);
                     } else {
                         const int64_t t0 = whisper_full_get_segment_t0(ctx, i);
                         const int64_t t1 = whisper_full_get_segment_t1(ctx, i);
 
-                        fprintf(stderr, "[%s --> %s]  %s\n", to_timestamp(t0).c_str(), to_timestamp(t1).c_str(), text);
+                        printf("[%s --> %s]  %s\n", to_timestamp(t0).c_str(), to_timestamp(t1).c_str(), text);
                     }
                 }
             }
 
-            fprintf(stderr, "\n");
+            printf("\n");
 
             // output to text file
             if (params.output_txt) {

whisper.h

@@ -31,33 +31,81 @@ extern "C" {
     // C interface
     //
 
-    // TODO: documentation will come soon
+    //
+    // Basic usage:
+    //
+    //     #include "whisper.h"
+    //
+    //     ...
+    //
+    //     struct whisper_context * ctx = whisper_init("/path/to/ggml-base.en.bin");
+    //
+    //     if (whisper_full(ctx, wparams, pcmf32.data(), pcmf32.size()) != 0) {
+    //         fprintf(stderr, "failed to process audio\n");
+    //         return 7;
+    //     }
+    //
+    //     const int n_segments = whisper_full_n_segments(ctx);
+    //     for (int i = 0; i < n_segments; ++i) {
+    //         const char * text = whisper_full_get_segment_text(ctx, i);
+    //         printf("%s", text);
+    //     }
+    //
+    //     whisper_free(ctx);
+    //
+    //     ...
+    //
+    // This is a demonstration of the most straightforward usage of the library.
+    // "pcmf32" contains the RAW audio data in 32-bit floating point format.
+    //
+    // The interface also allows for more fine-grained control over the computation, but it requires a deeper
+    // understanding of how the model works.
+    //
 
     struct whisper_context;
 
     typedef int whisper_token;
 
+    // Allocates all memory needed for the model and loads the model from the given file.
+    // Returns NULL on failure.
     WHISPER_API struct whisper_context * whisper_init(const char * path_model);
+
+    // Frees all memory allocated by the model.
     WHISPER_API void whisper_free(struct whisper_context * ctx);
 
+    // Convert RAW PCM audio to log mel spectrogram.
+    // The resulting spectrogram is stored inside the provided whisper context.
+    // Returns 0 on success
     WHISPER_API int whisper_pcm_to_mel(
             struct whisper_context * ctx,
             const float * samples,
             int n_samples,
             int n_threads);
 
+    // This can be used to set a custom log mel spectrogram inside the provided whisper context.
+    // Use this instead of whisper_pcm_to_mel() if you want to provide your own log mel spectrogram.
     // n_mel must be 80
+    // Returns 0 on success
     WHISPER_API int whisper_set_mel(
             struct whisper_context * ctx,
             const float * data,
             int n_len,
             int n_mel);
 
+    // Run the Whisper encoder on the log mel spectrogram stored inside the provided whisper context.
+    // Make sure to call whisper_pcm_to_mel() or whisper_set_mel() first.
+    // offset can be used to specify the offset of the first frame in the spectrogram.
+    // Returns 0 on success
     WHISPER_API int whisper_encode(
             struct whisper_context * ctx,
             int offset,
             int n_threads);
 
+    // Run the Whisper decoder to obtain the logits and probabilities for the next token.
+    // Make sure to call whisper_encode() first.
+    // tokens + n_tokens is the provided context for the decoder.
+    // n_past is the number of tokens to use from previous decoder calls.
+    // Returns 0 on success
     WHISPER_API int whisper_decode(
             struct whisper_context * ctx,
             const whisper_token * tokens,
@@ -65,10 +113,15 @@ extern "C" {
             int n_past,
             int n_threads);
 
+    // Token sampling methods.
+    // These are provided for convenience and can be used after each call to whisper_decode().
+    // You can also implement your own sampling method using the whisper_get_probs() function.
+    // whisper_sample_best() returns the token with the highest probability
+    // whisper_sample_timestamp() returns the most probable timestamp token
     WHISPER_API whisper_token whisper_sample_best(struct whisper_context * ctx, bool need_timestamp);
     WHISPER_API whisper_token whisper_sample_timestamp(struct whisper_context * ctx);
 
-    // return the id of the specified language, returns -1 if not found
+    // Return the id of the specified language, returns -1 if not found
     WHISPER_API int whisper_lang_id(const char * lang);
 
     WHISPER_API int whisper_n_len          (struct whisper_context * ctx); // mel length
@@ -76,10 +129,13 @@ extern "C" {
     WHISPER_API int whisper_n_text_ctx     (struct whisper_context * ctx);
     WHISPER_API int whisper_is_multilingual(struct whisper_context * ctx);
 
+    // The probabilities for the next token
     WHISPER_API float * whisper_get_probs(struct whisper_context * ctx);
 
+    // Token Id -> String. Uses the vocabulary in the provided context
     WHISPER_API const char * whisper_token_to_str(struct whisper_context * ctx, whisper_token token);
 
+    // Special tokens
     WHISPER_API whisper_token whisper_token_eot (struct whisper_context * ctx);
     WHISPER_API whisper_token whisper_token_sot (struct whisper_context * ctx);
     WHISPER_API whisper_token whisper_token_prev(struct whisper_context * ctx);
@@ -87,16 +143,19 @@ extern "C" {
     WHISPER_API whisper_token whisper_token_not (struct whisper_context * ctx);
     WHISPER_API whisper_token whisper_token_beg (struct whisper_context * ctx);
 
+    // Task tokens
     WHISPER_API whisper_token whisper_token_translate ();
     WHISPER_API whisper_token whisper_token_transcribe();
 
+    // Performance information
     WHISPER_API void whisper_print_timings(struct whisper_context * ctx);
 
     ////////////////////////////////////////////////////////////////////////////
 
+    // Available decoding strategies
     enum whisper_decode_strategy {
-        WHISPER_DECODE_GREEDY,
+        WHISPER_DECODE_GREEDY,      // Always select the most probable token
-        WHISPER_DECODE_BEAM_SEARCH,
+        WHISPER_DECODE_BEAM_SEARCH, // TODO: not implemented yet!
     };
 
     struct whisper_full_params {
@@ -129,18 +188,23 @@ extern "C" {
 
     WHISPER_API struct whisper_full_params whisper_full_default_params(enum whisper_decode_strategy strategy);
 
-    // full whisper run - encode + decode
+    // Run the entire model: PCM -> log mel spectrogram -> encoder -> decoder -> text
+    // Uses the specified decoding strategy to obtain the text.
     WHISPER_API int whisper_full(
             struct whisper_context * ctx,
             struct whisper_full_params params,
             const float * samples,
             int n_samples);
 
+    // Number of generated text segments.
+    // A segment can be a few words, a sentence, or even a paragraph.
     WHISPER_API int whisper_full_n_segments(struct whisper_context * ctx);
 
+    // Get the start and end time of the specified segment.
     WHISPER_API int64_t whisper_full_get_segment_t0(struct whisper_context * ctx, int i_segment);
     WHISPER_API int64_t whisper_full_get_segment_t1(struct whisper_context * ctx, int i_segment);
 
+    // Get the text of the specified segment.
     WHISPER_API const char * whisper_full_get_segment_text(struct whisper_context * ctx, int i_segment);
 
 #ifdef __cplusplus