Module export

Export: render the timeline with the CPU Compositor + Mixer and pipe it into ffmpeg.exe (-f rawvideo -pix_fmt rgba -s WxH -r fps -i pipe:0 + pre-mixed temp WAV), encoder chosen by the output extension / settings. Plus the lossless -c copy fast path for pure cuts of one source. Runs on a background thread; the UI polls Progress.

Structs

ExportOptions

GpuFrameRequest

One frame the export thread wants composited on the GPU. The UI thread owns the GL context, so it picks these up in App::update, renders layers and sends the pixels back.

GpuScratch 🔒

Decoder state for the GPU export path: the same layer decoding the player does, on this thread.

Progress

TempFile 🔒

Deleted on drop (a no-op once committed into place).

Enums

FrameSource

Where an export gets its pictures.

Constants

AUDIO_EXTS

CANCELLED 🔒

MIX_BLOCK 🔒

Audio mix block (frames). Same as playback::BLOCK on purpose: mixer::resample_add samples clip gains at the block ends and lerps between them, so a longer block here would smear fades, volume keyframes and audio crossfades shorter than the block — export must ramp like playback does.

Statics

ENCODERS 🔒

Functions

codec_args

ffmpeg output arguments (codec/quality) for an extension, given the user’s encoder preference and the available encoders. “auto”: mp4/mov/mkv/m4v → libx264 (+aac), webm → libvpx-vp9 (+libopus), gif → gif, avi → mpeg4 (+mp3), audio-only → sensible codec for the container. The preference is overridden where the container forbids it (webm: vp9/av1 only; mov: no vp9/av1). Hardware encoders (nvenc/qsv/amf) use -cq/-global_quality/-qp instead of -crf.

cpu_gaps 🔒

What this export cannot render, as a suffix for the progress line (empty when nothing is dropped).

detect_encoders

Encoder names ffmpeg reports (ffmpeg -hide_banner -encoders), cached after the first call. Empty if ffmpeg is missing.

ext_of 🔒

lossless_segments

If the project is a pure cut of exactly one video source (every clip from the same asset, audio clips exactly mirroring their linked video clip’s timing on each audio stream (muted tracks may be dropped), no gaps between clips, no text/image/sequence clips, no effects/retime/pan/fades, volume 1, all clips enabled, no transitions, no burnt-in subtitles), return the segments as (src_in, duration) in timeline order. Such a project can be written with -c copy.

mix_to_wav 🔒

Mix the whole timeline into a temp WAV (f32 stereo 48 kHz). Progress 0..0.1.

parse_encoders 🔒

run_export 🔒

run_lossless 🔒

spawn_job 🔒

Run job on a named thread; any Err (or panic) lands in Progress::finish.

start_export

Start a full (re-encoding) export on a background thread. Frames are rendered at project size with Compositor and audio mixed with Mixer into a temp WAV first (fast), then video frames are piped to ffmpeg’s stdin. Progress 0..1; cancel kills ffmpeg and removes the partial file. Audio-only extensions (mp3, wav, m4a, flac, ogg, aac, opus) skip video (-vn); gif skips audio.

start_lossless_cut

Lossless cut: ffmpeg -ss in -t dur -i src -c copy -avoid_negative_ts make_zero per segment (with -map dropping muted audio streams), then concat demuxer when there is more than one segment. Cuts land on keyframes (not frame-accurate) — that’s the trade for being instant.

stderr_tail 🔒

Drain ffmpeg’s stderr on a helper thread; returns the last ~2000 chars.

temp_output 🔒

Hidden sibling of out (same folder, same extension so ffmpeg still picks the muxer by name) that ffmpeg writes into; it is renamed over out only on success, so a cancel or failure never truncates or deletes an existing destination — or the source, when cutting in place.

wait_ffmpeg 🔒

Wait for ffmpeg, killing it if the job is cancelled. Non-zero exit → Err(stderr tail).

wav_header 🔒

44-byte RIFF/WAVE header: IEEE float 32-bit, stereo, 48 kHz, frames sample frames.