Cap heats up a lot gets excessively hot when recording video

[AasheeshLikePanner]

Description

Cap heats up a lot / gets excessively hot when recording video with both camera (webcam/facecam) and microphone turned on. The laptop becomes noticeably warm within just a few minutes, fans spin up louder, and it feels uncomfortably hot to touch. This happens even during short recordings (3–10 minutes).

Expected: Smooth recording without abnormal heating / high CPU/GPU usage when using camera + mic together. It seems worse than when recording screen only (no camera/mic).

Reproduction

1. Open Cap desktop app
2. Start a new screen recording
3. Turn ON both Camera (webcam overlay) and Microphone
4. Click record (with or without capturing screen — happens either way)
5. Record for 3–10 minutes → device gets significantly hot

Additional Context

• Operating system, version: macOS tahoe 26.3.1
• Device (optional): macbook air m4

[linear]

CAP-662

[tembo]

This issue is closely related to #586 (Laggy system while recording), which reports similar high CPU usage and system lag during recording on a MacBook Pro M1 Max. In that issue, a maintainer noted that the encoding uses VTEncoder via AVFoundation which is supposed to be hardware-accelerated, but may require different configuration for high-resolution encoding.

Relevant code areas for investigation:

• Camera feed processing: crates/recording/src/feeds/camera.rs - handles camera frame capture and distribution
• macOS output pipeline: crates/recording/src/output_pipeline/macos.rs - manages encoding with channel pressure tracking
• VideoToolbox integration: crates/frame-converter/src/videotoolbox.rs - pixel format conversion using VTPixelTransferSession

Diagnostic tools available:

The codebase includes a camera benchmark that could help isolate whether the issue is in capture, conversion, or encoding:

cd apps/desktop
cargo run --example camera-benchmark --release

This measures frame rate, inter-frame timing, and full encoding pipeline performance.

Potential contributing factors:

1. The camera format selection in select_camera_format filters for formats under 2000x2000 and 30fps+, but simultaneous screen + camera encoding at high resolutions may still stress the encoder
2. Frame conversion between camera pixel formats and encoder input (UYVY/YUYV to NV12) happens on CPU via the converter pool
3. Multiple encoding streams (screen + camera) running in parallel compounds resource usage

Attaching diagnostic logs via the app's "Upload Logs" feature would help identify whether frame drops or encoder backpressure are occurring.

[tembo]

This issue is related to several existing reports about CPU usage during recording:

• #586 - Reports 100% CPU usage with GPU not being used during recording with webcam on MacBook Pro M1 Max. A contributor noted that Cap uses VTEncoder via AVFoundation which should be hardware accelerated, but may need different configuration for certain resolutions.
• #988 - Similar report of all available CPU being used on Apple M4 Max with laggy/patchy video output.
• #109 - Tracks CPU/memory optimization during recording, with reports of the issue reproducing specifically when recording with webcam + mic enabled.

Relevant code areas:

• Camera capture pipeline: crates/recording/src/sources/camera.rs - handles frame scaling via FFmpeg's software scaler when camera dimensions change
• Camera feed processing: crates/recording/src/feeds/camera.rs
• macOS camera backend: crates/camera-avfoundation/src/lib.rs
• Output encoding: crates/recording/src/output_pipeline/macos.rs

Diagnostic tools available:
The codebase includes a camera benchmark example at crates/recording/examples/camera-benchmark.rs that can help measure frame rate and encoding performance. Running this could help identify whether the bottleneck is in capture, color conversion, or encoding:

cargo run --release --example camera-benchmark

The camera source uses a software scaler (ffmpeg::software::scaling::Context) when frame dimensions don't match the expected output, which runs on CPU. If the webcam is delivering frames at a different resolution than expected, this could contribute to CPU load.

For debugging, logs from ~/Library/Logs/so.cap.desktop/ would help determine if frames are being dropped due to the encoder not keeping up (look for "encoder can't keep up" warnings).

[m13v]

this is almost certainly the encoding pipeline using software codecs (libx264/libx265) instead of VideoToolbox hardware encoders. software encoding on a MacBook Air M4 will pin the CPU and generate tons of heat because there's no active cooling fan to dissipate it.

we built a screen recording SDK that captures at 5 FPS continuously and the key to keeping thermals sane was using VTCompressionSession with H.265 hardware encoding. CPU stays under 5% and the encoding happens on Apple's dedicated media engine, which generates way less heat than running x264 on the efficiency cores.

the webcam + mic combo probably makes it worse because you're decoding the camera feed, compositing it onto the screen capture, then encoding the result, all in software. if any of those steps hit the CPU instead of the GPU/media engine, thermals spiral fast on fanless machines. worth checking Activity Monitor to see if the ffmpeg or encoding process is the one eating CPU.

[m13v]

our VTCompressionSession pipeline for low-CPU H.265 encoding on Apple Silicon, keeps thermals sane even on fanless machines: https://github.com/m13v/macos-session-replay/blob/main/Sources/SessionReplay/VideoChunkEncoder.swift

[richiemcilroy]

There's no version or logs here to work from. Please re-file with your current version, OS, and logs if the app is still running hot with camera and mic active.