services/whisper/utils/jumpcut.py

"""
jumpcut.py
---------------------------------------
Smart Jump Cut Engine (V8)

Purpose:
- Remove silence and filler pauses
- Improve pacing for short-form video
- Optimize retention curve
- Create TikTok/Reels-style fast cuts

Works fully on CPU (FFmpeg-based).
No GPU required.
"""

import subprocess
import os


# =====================================================
# CONFIG
# =====================================================

TEMP_SILENCE_FILE = "silence_detect.txt"
OUTPUT_FILE = "jumpcut_output.mp4"


# =====================================================
# SILENCE DETECTION
# =====================================================

def detect_silence(video_path):
    """
    Uses ffmpeg silencedetect to find pauses.
    """

    cmd = [
        "ffmpeg",
        "-i", video_path,
        "-af", "silencedetect=noise=-30dB:d=0.4",
        "-f", "null",
        "-"
    ]

    result = subprocess.run(cmd, stderr=subprocess.PIPE, text=True)

    return result.stderr


# =====================================================
# PARSE SILENCE TIMESTAMPS
# =====================================================

def parse_silence(log):
    """
    Extract silence start/end timestamps
    """

    silences = []

    start = None

    for line in log.split("\n"):

        if "silence_start" in line:
            try:
                start = float(line.split("silence_start:")[1].strip())
            except:
                continue

        if "silence_end" in line and start is not None:
            try:
                end = float(line.split("silence_end:")[1].split("|")[0].strip())
                silences.append((start, end))
                start = None
            except:
                continue

    return silences


# =====================================================
# BUILD FILTER (JUMP CUT LOGIC)
# =====================================================

def build_filter(silences, duration):
    """
    Converts silence ranges into ffmpeg trim filter
    """

    if not silences:
        return None

    segments = []
    last_end = 0

    for start, end in silences:

        if start > last_end:
            segments.append((last_end, start))

        last_end = end

    if last_end < duration:
        segments.append((last_end, duration))

    filters = []

    for i, (start, end) in enumerate(segments):
        filters.append(
            f"[0:v]trim=start={start}:end={end},setpts=PTS-STARTPTS[v{i}];"
            f"[0:a]atrim=start={start}:end={end},asetpts=PTS-STARTPTS[a{i}]"
        )

    video_concat = "".join([f"[v{i}]" for i in range(len(segments))])
    audio_concat = "".join([f"[a{i}]" for i in range(len(segments))])

    filters.append(
        f"{video_concat}{audio_concat}concat=n={len(segments)}:v=1:a=1[outv][outa]"
    )

    return ";".join(filters)


# =====================================================
# CORE ENGINE
# =====================================================

def smart_jumpcut(video_path):
    """
    Main jump cut engine
    """

    print("[JUMPCUT] Analyzing video...")

    # Step 1: detect silence
    log = detect_silence(video_path)

    silences = parse_silence(log)

    print(f"[JUMPCUT] Detected silences: {len(silences)}")

    # Step 2: get duration
    probe_cmd = [
        "ffprobe",
        "-v", "error",
        "-show_entries",
        "format=duration",
        "-of",
        "default=noprint_wrappers=1:nokey=1",
        video_path
    ]

    duration = float(subprocess.check_output(probe_cmd).decode().strip())

    # Step 3: build filter
    filter_complex = build_filter(silences, duration)

    if not filter_complex:
        print("[JUMPCUT] No silences found, returning original")
        return video_path

    # Step 4: render output
    output_path = OUTPUT_FILE

    cmd = [
        "ffmpeg", "-y",
        "-i", video_path,
        "-filter_complex", filter_complex,
        "-map", "[outv]",
        "-map", "[outa]",
        "-c:v", "libx264",
        "-preset", "ultrafast",
        "-c:a", "aac",
        output_path
    ]

    print("[JUMPCUT] Rendering optimized video...")

    subprocess.run(cmd, check=True)

    print("[JUMPCUT] Done:", output_path)

    return output_path


# =====================================================
# SIMPLE FAST MODE (FALLBACK)
# =====================================================

def fast_jumpcut(video_path):
    """
    Lightweight fallback:
    removes only large pauses quickly
    """

    output = "fast_jumpcut.mp4"

    cmd = [
        "ffmpeg", "-y",
        "-i", video_path,
        "-af", "silenceremove=start_periods=1:start_threshold=-30dB:stop_periods=-1",
        "-c:v", "libx264",
        "-preset", "ultrafast",
        "-c:a", "aac",
        output
    ]

    subprocess.run(cmd, check=True)

    return output


# =====================================================
# PUBLIC API
# =====================================================

def smart_jumpcut_engine(video_path, mode="smart"):
    """
    Entry point used by main.py
    """

    if mode == "fast":
        return fast_jumpcut(video_path)

    return smart_jumpcut(video_path)