diff --git a/app-phone/src/main/kotlin/pw/binom/viewmate/phone/stt/SttDebug.kt b/app-phone/src/main/kotlin/pw/binom/viewmate/phone/stt/SttDebug.kt index 63bbfe0..eaee2fd 100644 --- a/app-phone/src/main/kotlin/pw/binom/viewmate/phone/stt/SttDebug.kt +++ b/app-phone/src/main/kotlin/pw/binom/viewmate/phone/stt/SttDebug.kt @@ -23,16 +23,24 @@ object SttDebug { // через: adb push /data/local/tmp/... + run-as pkg cp val modelsDir = File(context.filesDir, "models") val vadModel = File(modelsDir, "silero_vad.onnx").absolutePath - val encoder = File(modelsDir, "small-encoder.int8.onnx").absolutePath - val decoder = File(modelsDir, "small-decoder.int8.onnx").absolutePath - val tokens = File(modelsDir, "small-tokens.txt").absolutePath + val convFrontend = File(modelsDir, "conv_frontend.onnx").absolutePath + val encoder = File(modelsDir, "encoder.int8.onnx").absolutePath + val decoder = File(modelsDir, "decoder.int8.onnx").absolutePath + val tokens = File(modelsDir, "tokenizer").absolutePath log("[stt] создаём VAD...") val segments = VadSegmenter.vadSegments(wav.samples, vadModel) log("[stt] фраз (VAD): ${segments.size}") - log("[stt] создаём Whisper-small (int8, CPU)...") - val stt = WhisperStt(encoder, decoder, tokens, numThreads = 4) + log("[stt] создаём Qwen3-ASR-0.6B (int8, CPU)...") + val stt = Qwen3AsrStt( + convFrontendPath = convFrontend, + encoderPath = encoder, + decoderPath = decoder, + tokenizerDir = tokens, + numThreads = 4, + hotwords = "тулсет,тулы,умеешь", + ) val parts = ArrayList() for ((i, seg) in segments.withIndex()) { val startMs = seg.first / 16