com.blockether.vis.ext.channel-tui.cinema — com.blockether/vis-channel-tui 0.1.2

Headless session → MP4 screencast. Replays a persisted session's transcript through the REAL TUI render pipeline (screen/render-frame! painting into a Lanterna VirtualTerminal — no TTY needed), and encodes the captured frames to .mp4:

pure-JVM H.264 via jcodec's AWTSequenceEncoder (each frame rendered to a BufferedImage with Java2D). Pixel-exact because WE draw the glyphs — no external player font drift. No native deps, GraalVM-friendly.

The replay is HUMANIZED rather than a flat scroll-through: every disclosure COLLAPSED, the session is re-enacted turn by turn as if it were happening live — a user turn is TYPED IN character by character; before each answer a WORK beat holds the live Vis is calling the provider… spinner in the assistant's place; then the finished answer JUMPS IN and SLOWLY scrolls into view. See session->frames.

The heavy TUI stack (screen, Lanterna) is reached via requiring-resolve so merely loading this namespace stays cheap.

export! clj

(export! session-id opts)

Render session-id to an MP4 screencast file.

opts: :out output path (default: <session-id>.mp4 in the cwd) :cols :rows :fps :theme :font-size :type-cps :work-ms :jump-hold-ms :scroll-rows-per-frame :post-user-ms :post-assist-ms — see defaults

Returns {:path :format :frames :video-ms :cols :rows}.

frames->mp4! clj

(frames->mp4! {:keys [cols rows frames]}
              out
              {:keys [font-size fps] :or {font-size 18 fps 8}})

Encode a session->frames result to out (an MP4 File/path) via jcodec's pure-Java AWTSequenceEncoder. Returns the output File.

session->frames clj

(session->frames session-id opts)

Replay session-id headless and return {:cols :rows :frames [{:grid :delay-ms} …] :video-ms}.

opts overrides defaults. The replay is HUMANIZED: rather than dumping the whole transcript and scrolling through it, it RE-ENACTS the session as if it were happening live, one message at a time, with every disclosure COLLAPSED (the natural resting view a live session shows, not an expanded dump):

• a USER turn is TYPED IN character by character — the bubble grows as the text lands, like a human at the keyboard — then a short beat; • before each ASSISTANT answer there is a WORK beat: the LIVE progress bubble holds in the assistant's place (⠿ Vis is calling the provider… 1.2s / Esc to cancel, spinner animating, elapsed clock ticking) so it reads as real work being done; • the finished answer then JUMPS IN whole (a bubble popping in near the bottom of the viewport), holds a beat, then SLOWLY scrolls into view; • a beat to read it, and on to the next turn.

Pacing is by FRAME COUNT at a constant fps (the encoder ignores per-frame delays): every 'beat' is that many repeated frames, and the whole clip length scales with the session's content. scroll/follow carries no :pos, so each captured frame SNAPS to its target — deterministic, no stray ease.