SilenceCut

Silences shorter than this are kept.

Audio below this level is treated as silence.

Keep this buffer around each silence edge.

High = VBR max / 24-bit / lossless where applicable.

Duration: 60.0s

✦ Tip — enter exact seconds. Max range shown is 600s; your file may be shorter. FFmpeg will clamp automatically.

Lower = more cuts, better for reels. Higher = fewer captions.

Caption auto-splits if it would exceed this length.

SRT = universal · SBV = YouTube · VTT = web/HTML5

Tiny gap prevents captions blending together on screen.

How it works:

• ✦ Whisper (OpenAI) — transcribes your audio via word-level timestamps
• ✦ Word grouping — splits into segments by word count + max duration
• ✦ Format export — outputs clean .srt / .sbv / .vtt ready to upload
• ✦ Requires — pip install openai-whisper on the server

WebP recommended — best size/quality ratio.

Lossy targets under 300–400 KB. Lossless = pixel-perfect, larger file.

82 hits the sweet spot for WebP under 400 KB at most resolutions.

Resize constrains longest edge. Aspect ratio always preserved.

Stripping EXIF removes GPS, camera info, reduces size slightly.

Size targeting for WebP:

• ✦ Quality 82 + Lossy — typically 80–200 KB for photos under 1920px
• ✦ Quality 70 + 1280px resize — reliably under 100 KB for web use
• ✦ Lossless WebP — exact pixel copy, usually 200 KB–1 MB depending on image
• ✦ AVIF — 30–50% smaller than WebP lossy, slower to encode

Each selected format produces a separate compressed file for download.

Lossy achieves smallest file size. JPEG is always lossy.

Starting quality. When max-size is set, this is adjusted automatically via binary search.

Server binary-searches quality to stay at or below target. PNG is lossless, size not guaranteed.

Removes GPS, camera data, ICC profile. Saves a few KB.

Compression Guide

• ✦ WebP Lossy q82 — ~80–200 KB for most photos
• ✦ AVIF — 30–50% smaller than WebP, slower
• ✦ JPEG — universal compat, always lossy
• ✦ PNG — lossless always, larger files
• ✦ Max Size + Lossy — server auto-adjusts quality
• ✦ SVG input — rasterised to chosen format

B&W = fastest, sharpest. Multicolor = separate pass per channel. Grayscale = layered tones.

Controls how corners are traced — lower = more angular, higher = smoother curves.

Removes pixel islands smaller than this. Increase for noisy/scanned images.

Higher = fewer path nodes, smoother SVG, slightly less accurate.

Pixels darker than this are traced as filled. Lower = capture lighter areas.

Best results:

• ✦ Logos & icons — B&W mode, smoothness 1.0, speckle 2
• ✦ Scanned line art — B&W mode, increase speckle to 10–20
• ✦ Flat color illustrations — Multicolor mode, smoothness 1.0
• ✦ Photos / complex scenes — not recommended, use Grayscale Steps instead