ascii-video

Sample frames from the source video, map per-cell luminance to characters from a density-sorted palette, and sample original pixel colors for hue. Render each frame onto a pixel canvas with pre-rasterized font bitmaps, then pipe raw RGB frames to ffmpeg for H.264 encoding.

Extract per-frame audio features with SciPy: 6-band FFT, RMS energy, spectral centroid, and beat detection with exponential decay. Drive value field generators, particle triggers, and shader intensity from these features, then encode frames to MP4 with the audio muxed.

No GPU is required. The pipeline runs on CPU with NumPy-vectorized effects and Pillow font rasterization, using concurrent.futures workers for parallel encoding. A 1080p 24fps frame renders in roughly 100-200ms, with quality profiles adapting to available cores and RAM.

ASCII characters are small bright dots on black backgrounds, so linear brightness multipliers clip highlights while leaving frames dark. Use percentile-based adaptive tonemapping with gamma correction instead: default gamma 0.75, lower to 0.50-0.60 for solarize or posterize scenes.

Setting stderr=subprocess.PIPE on a long-running ffmpeg process deadlocks once the buffer fills at about 64KB. Redirect stderr to a log file instead, and pipe only raw RGB frames through stdin for the encode.

Not all Unicode glyphs render in every monospace font, and Pillow's textbbox returns wrong heights on macOS. Validate palettes at initialization by rendering each character and dropping blanks, and compute cell height from font.getmetrics() ascent plus descent.