#!/bin/bash
# ZipVoice AXCL 算力卡一键运行
# 用法: bash run_axcl.sh [standard|distill] [zh|en] [sentence|paragraph]

MODEL="${1:-distill}"
LANG="${2:-zh}"
MODE="${3:-sentence}"

BIN="cpp/install/axcl/zipvoice"
TOKEN="./resources/zipvoice_hf/zipvoice/tokens.txt"
VOCODER="./cpp/vocoder/vocos_full.axmodel"

cd "$(dirname "$0")"

if [ "$MODEL" = "distill" ]; then
    MODEL_DIR="./models/zipvoice_distill_ax650"
else
    MODEL_DIR="./models/zipvoice_ax650"
fi

if [ "$LANG" = "en" ]; then
    PWAV="./assets/moss_prompts/en_4_4p5s.wav"
    PTEXT="This is almost twice the current industry production level per train."
    DTEXT="This morning, a small train left the station, carrying sleepy passengers toward a bright coastal town."
    TFILE="./assets/paragraphs/en_scavenger.txt"
    REPO="--repo-dir ."
else
    PWAV="./assets/moss_prompts/zh_1_4p5s.wav"
    PTEXT="不管怎么样我和汤姆还是要感谢贝尔卡金的援手"
    DTEXT="今天午后天气很好，我打开窗户，听见远处有人聊天，水杯也轻轻晃了一下。"
    TFILE="./assets/paragraphs/zh_ginkgo.txt"
    REPO=""
fi

OUT="output_${LANG}_${MODEL}_${MODE}.wav"

echo "=========================================="
echo " ZipVoice AXCL"
echo "=========================================="
echo " MODEL:  $MODEL_DIR"
echo " LANG:   $LANG  MODE: $MODE"
echo " OUTPUT: $OUT"
echo "=========================================="

if [ "$MODE" = "paragraph" ]; then
    exec $BIN \
        --model-dir "$MODEL_DIR" --token-file "$TOKEN" \
        --prompt-wav "$PWAV" --prompt-text "$PTEXT" \
        --text-file "$TFILE" \
        --vocoder-model "$VOCODER" --output-wav "$OUT" --seed 42 $REPO
else
    exec $BIN \
        --model-dir "$MODEL_DIR" --token-file "$TOKEN" \
        --prompt-wav "$PWAV" --prompt-text "$PTEXT" \
        --text "$DTEXT" \
        --vocoder-model "$VOCODER" --output-wav "$OUT" --seed 42 $REPO
fi