Cuts background noise (AC, fans, electrical hum) and softens sibilance in meeting, lecture, and interview recordings and video, then normalizes loudness to podcast (-16), YouTube (-14), or broadcast (-23 LUFS). Point to a quiet patch and profile mode removes only that noise. Also trims leading/trailing silence and builds a before/after report (spectrogram, LUFS). On a 60-second 1962 Kennedy clip, the noise floor fell 5.6 dB and loudness moved from -17.1 to -15.9 LUFS. Only ffmpeg, no API key.
Drop your raw lecture, talking-head, or vlog footage into the chat and get back an MP4 with the silences and filler words ("um," "uh") cut out. You also get FCPXML, EDL, and CSV edit lists so you can keep editing at full source quality in Premiere or DaVinci Resolve. Runs in the cloud, nothing to install. Measured: a 1:50 synthetic-voice sample came back at 0:57 (20 cuts); an 11-minute clip processed in 3:13. This isn't a highlight picker.
Upload 2–10 takes of the same Short. BestTake finds the cleanest usable delivery of each line and builds one finished cut automatically.
Compresses AI filler text — preambles, transitions, repetitions — at 3 intensity levels (Lite / Full / Ultra) while enforcing a protected scope (URLs, numbers, code, reasoning) that is never deleted, so "be concise" stops silently destroying what matters.
Automatically cuts silences and filler words like "um" and "uh" from talking-head, lecture and vlog footage. ffmpeg silence detection (plus optional local STT filler detection) works out the segments to keep, then it renders a frame-accurate mp4 plus EDL and FCPXML for Premiere or DaVinci at source quality. In a measured sample, 47.3% of a 33.87 s source was removed with an 11.6 ms audio/video length gap. It strips dead time, it does not pick highlights. Works with ffmpeg only, no API key.
Stop losing gigs to false AI-detector flags — TrueVoice rewrites your draft into natural human prose and shows every AI tell it removed.
テーマを伝えるだけで15〜35秒の縦型ショート動画を生成するツールキット。VOICEVOX＋Remotion＋ffmpegの完全無料スタックで、何本作ってもランニングコスト0円。台本生成から字幕付きレンダリングまで一気通貫。
Drop a video or audio file into the chat and get back clean SRT/VTT subtitles plus an MP4 with the captions burned in. After transcribing, it re-splits cues by line length, characters-per-second, and on-screen time so they're easy to read. The agent translates them itself, and if the cue count or timecodes drift from the original, it won't ship the file. No install, no API key. Real run: a 42s clip went from 7 raw Whisper segments to 12 clean cues; a 2m25s clip transcribed in 9s, burned in 29s.