add tts max length for kokoro (+1 squashed commits)

Squashed commits:

[c1c6feaf] add tts max length for kokoro

Author: LostRuins

koboldcpp.py

@@ -64,7 +64,7 @@
 extra_images_max = 4
 
 # global vars
-KcppVersion = "1.98"
+KcppVersion = "1.98.1"
 showdebug = True
 kcpp_instance = None #global running instance
 global_memory = {"tunnel_url": "", "restart_target":"", "input_to_exit":False, "load_complete":False, "restart_override_config_target":""}

otherarch/tts_adapter.cpp

@@ -478,6 +478,32 @@ std::string trim_words(const std::string& input, const std::string& separator, s
     return result.str();
 }
 
+static std::string TruncateToFirstNumberWords(const std::string& input, int limit) {
+    static const std::regex wordRegex(R"(\b[\w'-]+\b)");
+    std::sregex_iterator words_begin(input.begin(), input.end(), wordRegex);
+    std::sregex_iterator words_end;
+    int count = 0;
+    std::size_t cutoffPos = std::string::npos;
+    if(limit<=0)
+    {
+        return "";
+    }
+    for (auto it = words_begin; it != words_end; ++it) {
+        ++count;
+        if (count >= limit) {
+            // position AFTER the last matched word
+            cutoffPos = it->position() + it->length();
+            break;
+        }
+    }
+    if (cutoffPos == std::string::npos) {
+        // fewer than N words, return original
+        return input;
+    }
+    // Preserve everything up to and including the Nth word
+    return input.substr(0, cutoffPos);
+}
+
 static llama_context * ttc_ctx = nullptr; //text to codes ctx
 static llama_context * cts_ctx = nullptr; //codes to speech
 
@@ -562,6 +588,7 @@ bool ttstype_load_model(const tts_load_model_inputs inputs)
     }
 
     ttsdebugmode = inputs.debugmode;
+    tts_max_len = inputs.ttsmaxlen;
 
     // tts init
     if (is_ttscpp_file) {
@@ -577,8 +604,6 @@ bool ttstype_load_model(const tts_load_model_inputs inputs)
 
         nthreads = inputs.threads;
 
-        tts_max_len = inputs.ttsmaxlen;
-
         tts_model_params.use_mmap = false;
         tts_model_params.use_mlock = false;
         tts_model_params.n_gpu_layers = inputs.gpulayers; //offload if possible
@@ -692,6 +717,11 @@ static tts_generation_outputs ttstype_generate_ttscpp(const tts_generation_input
         }
     }
 
+    if(tts_max_len>0)
+    {
+        prompt = TruncateToFirstNumberWords(prompt,tts_max_len);
+    }
+
     if(ttsdebugmode==1 && !tts_is_quiet)
     {
         printf("\nUsing Speaker ID: %d, Voice: %s", speaker_seed, voiceused.c_str());