hl-valdemar/terebi
1
0
Fork 0
Code Issues Pull Requests Actions Packages Releases Wiki Activity

feat: stream YouTube video with coordinated audio and video inputs

Browse Source 
Resolve public YouTube URLs through external yt-dlp and play selected streams in the existing native window.

Independent bounded network readers keep one authored timeline. Successful seeks clear FFmpeg's sticky aborted-read AVIO error; manifest seekability uses native format flags rather than a top-level I/O handle.

Verified real YouTube playback, seeking, frame stepping, compact mode, cancellation, and the complete native regression checks.
This commit is contained in:
Valdemar Høgsvig Lorenzen
2026-10-10 22:18:24 +02:00
parent 25d674d84d
commit 8fb08db6f3
10 changed files with 2451 additions and 181 deletions
Download Patch File Download Diff File Expand all files Collapse all files
README.md
+26 -4
View File
@@ -16,6 +16,20 @@ The local app is `.build/terebi.app`. It links the installed FFmpeg libraries; i
Open a local file with **⌘O**, drop it onto the window, or use Finder's **Open With**. Successfully opened files start playback automatically, including when replacing the current file. **⇧⌘W** closes the file; closing the window quits and joins the worker.
## YouTube
YouTube playback optionally requires [yt-dlp](https://github.com/yt-dlp/yt-dlp), its EJS components, and a supported JavaScript runtime. Homebrew provides the extractor and its components; Deno is recommended:
```sh
brew install yt-dlp deno
```
Use **⇧⌘O → Open YouTube URL…**, the playback-actions menu, or drop a YouTube video link onto the window. Watch, short-link, Shorts, live, and embed URLs identify one video; playlists and channels are not opened. Resolution is cancellable by closing or replacing the media. Finder launches also search the usual Homebrew executable locations.
yt-dlp selects the best available SDR video and audio, including separate streams, with a combined-stream fallback. FFmpeg plays the selected HTTP(S) streams or native HLS manifests directly: no embedded browser, temporary full-video download, or muxing step. Playback, coordinated seeking, frame stepping, volume, and same-window compact mode use the existing native player. **Buffering…** freezes the playback clock without disabling pause, close, or replacement; live/non-seekable inputs disable seeking and frame stepping, and live sources show **LIVE**.
The app ignores yt-dlp configuration and plugins, does not read browser cookies, and does not update or fetch extractor components automatically. Private, age-restricted, region-blocked, or sign-in-required videos can fail. If YouTube extraction changes, run `brew upgrade yt-dlp deno`; local playback needs neither dependency.
## Playback UI
The layout follows [IINA's video-first window](https://iina.io/img/screenshot.png) and the supplied screenshots: a 32-point frosted titlebar with native traffic lights and file icon, plus a dark frosted 440 × 68-point floating controller. AppKit materials blur the video beneath the chrome; muted icons, an orange volume track, a narrow seek thumb, and zero-padded time labels follow the reference. Shared vertical padding and an optical offset for symbol/slider insets balance the controller's visible top and bottom margins in both play and pause states. The controller starts centered with a four-percent bottom inset and keeps its relative position when the window changes size.
@@ -50,11 +64,13 @@ Only implemented actions have controls; there are no placeholder playlist, subti
## Where to change things
- `Sources/Core/player.h`: the C interface and authoritative state snapshot. Playback commands enqueue work; poll state for completion. `seek_pending` is set synchronously when a seek or frame step is accepted and clears when all queued seeks/steps on that timeline finish. `player_step_frame` accepts only `-1` or `+1`, requires ready video, and publishes the selected frame's timestamp as `position`. Volume set/get is synchronous and uses a lock-free target gain. Callers own the opaque handle and must stop using it before destruction.
- `Sources/Core/player.cpp`: single demux/decode worker, bounded buffers, timestamps, audio-master synchronization, paused previews, exact frame stepping, decoder/resampler drain, seeking, errors, and shutdown. A seek stops output and flushes all old timeline state before resuming.
- `Sources/Core/player.h`: the C interface and authoritative state snapshot. Playback commands enqueue work; poll state for completion. `player_open` accepts local files; `player_open_network` copies selected HTTP(S) URLs, per-input headers, duration, and live status. `can_seek` reflects both selected inputs; `buffering` reports a stalled network clock. `seek_pending` is set synchronously on acceptance and clears when queued seeks/steps and both network input seeks finish. `player_step_frame` accepts only `-1` or `+1`, requires ready seekable video, and publishes the selected frame's timestamp as `position`. Volume set/get is synchronous and lock-free. Callers own the opaque handle and must stop using it before destruction.
- `Sources/Core/player.cpp`: demux/decode worker for local files and decode/timeline worker for network sources, bounded buffers, authored timestamps, audio-master synchronization, paused previews, exact frame stepping, independent decoder/resampler drain, seeking, errors, and shutdown. A seek stops output and flushes both selected inputs' old timeline state before resuming.
- `Sources/Core/network_input.hpp`: network-only demux readers with bounded encoded packet queues, asynchronous coordinated seeks, fifteen-second I/O deadlines, and interruption on seek/replacement/close/destruction. Remote input protocols exclude local-file access through manifests.
- `Sources/Core/output.hpp` and `Sources/macOS/Output.mm`: native output contract, allocation-free/nonblocking Core Audio callback and bounded audio ring, latency-aware presentation clock, retained frame mailbox, and Metal YUV conversion. The view can outlive the player safely.
- `Sources/App/TerebiApp.swift`: MainActor state polling, file panel/Finder/drop handling, autoplay, video host, AppKit glass and native slider cells, floating controls, visibility, video-fitted sizing, and native window integration. The shared `PlayerModel.open` path enqueues play after an accepted open, on the new file's command timeline; the C interface still opens paused for callers that need explicit playback control.
- `Sources/App/TerebiApp.swift`: MainActor state polling, file/YouTube panels and Finder/drop handling, cancellable URL resolution, autoplay, video host, AppKit glass and native slider cells, floating controls, visibility, video-fitted sizing, and native window integration. Both accepted open paths enqueue play on the new source's command timeline; the C interface still opens paused for callers that need explicit control.
- `Sources/App/CompactWindow.swift`: same-window compact geometry, saved normal window state, asynchronous fullscreen transitions and failure recovery, all-Spaces/floating behavior, magnetic edge snapping, and its persisted preference.
- `Sources/App/YouTubeResolver.swift`: YouTube URL and metadata validation, external yt-dlp/runtime discovery, SDR stream selection and headers, bounded nonblocking process output, sanitized extraction errors, and request-local cancellation.
- `Makefile`: compiler/linker commands and app bundle construction. No generated Xcode project or additional build dependency.
## Runnable check
@@ -69,6 +85,10 @@ Generates short fixtures and exercises the real playback interface and native ou
`make check-window` runs the compact-mode check, also included in `make check`. It exercises magnetic thresholds/corners, negative-origin screens, portrait/extreme aspect bounds, isolated preference persistence, actual hidden-window entry/resizing/snapping/restoration, view-host identity, keyboard eligibility, and a replay of AppKit's fullscreen-failure lifecycle. It requires a graphical macOS session but does not display windows or play audio. Visible fullscreen transitions, cross-Space/Stage Manager behavior, and hover/drag usability still require a check in the app.
`make check-youtube` checks URL/metadata/header trust boundaries and real subprocess output, bounded pipe draining, cancellation, and request isolation using isolated extractor fixtures. It needs neither installed yt-dlp/Deno nor external network access.
`make check-network` uses a range-capable loopback HTTP server and actual native playback. It checks combined/separate streams, identical stream indices across inputs, selected headers, coordinated seeks, exact VFR/B-frame stepping, asymmetric EOF and audio-tail seeks, recorded native HLS seeks, live/non-range seek rejection, buffering before the entire file arrives, permanently stalled primary/secondary reads, cancellation/replacement/destruction, HTTP failures, and recovery. Both new checks are included in `make check`.
Continuous audio uses resampled sample counts for its output timeline; packet timestamp quantization and rounded resampler delay must not insert silence or drop samples between blocks.
Volume persists across open, seek, and close. Live gain changes ramp over five milliseconds without callback allocation or locking; a stopped output adopts the requested gain before playback starts, so muted startup cannot leak audio.
@@ -77,8 +97,10 @@ Visual correctness, perceived A/V synchronization, and control usability still n
## Current limits
Local regular files, automatically selected best audio/video tracks, software decoding, and SDR output. PQ/HLG HDR is rejected rather than rendered incorrectly. No subtitles, selectable tracks, VideoToolbox decoding, rotation/deinterlacing, or network playback yet. Mid-track audio format changes report an error.
Local regular files and YouTube video URLs, automatically selected best audio/video tracks, software decoding, and SDR output. PQ/HLG HDR is rejected rather than rendered incorrectly. No subtitles, selectable tracks, VideoToolbox decoding, or rotation/deinterlacing. Mid-track audio format changes report an error. The URL UI accepts YouTube videos rather than arbitrary web pages or streaming services.
Forward frame stepping retains the live video decoder cursor instead of seeking and re-decoding the same keyframe group for each repeat. Backward stepping still scans from an earlier keyframe; long-GOP or unindexed files can therefore step backward slowly, without blanking the image or allocating a command per key repeat. Audio is skipped while stepping and rebuilt once on resume from the verified preroll anchor; resuming after a long forward run can also require a longer preroll. Intermediate frames needed by inter-frame dependencies are decoded and discarded; frames are not cached across the entire file. Every accepted step counts. Stepping requires usable decoded presentation timestamps and reports an error rather than inventing frame times when they are absent.
Audio buffering is bounded to one second plus one decoded block (up to four seconds); video is bounded to 64 frames/128 MiB. Extremely interleaved files can drop far-future video frames to avoid starving audio; the ceiling and upgrade path are marked at `queueVideo()`. Device/stream latency is accounted for, but device changes during playback are not handled yet.
Network demux queues are additionally bounded to 64 encoded packets/16 MiB per selected track, including packet side data. Each network reader can hold one pending packet while backpressured. Separate inputs retain one authored timestamp origin and one playback clock; buffering stalls that clock rather than skipping ahead. Network reads/seeks have fifteen-second deadlines; there is no automatic reconnection or signed-URL refresh.