H265 to ProRes Transcode

2025-03-12 14:45 UTC o3-mini-high Open in ChatGPT ↗

Below is one example of an ffmpeg command that converts an H.265 8K 360° video into ProRes while attempting to preserve both quality and metadata:

ffmpeg -i input.mkv -map_metadata 0 -c:v prores_ks -profile:v 4 -qscale:v 1 -pix_fmt yuv422p10le -c:a copy output.mov

Explanation

-i input.mkv
Specifies your source file. Replace “input.mkv” with the actual filename.
-map_metadata 0
Copies all metadata from the input file to the output, helping ensure that any embedded 360° or other metadata is preserved.
-c:v prores_ks
Selects the ProRes encoder. The prores_ks encoder is widely used for high-quality intermediate files.
-profile:v 4
Sets the encoder’s profile to ProRes HQ. (If you need a different ProRes variant, adjust this value accordingly.)
-qscale:v 1
Uses a very high-quality (nearly lossless) quality setting for the ProRes encoder. Lower values here mean higher quality.
-pix_fmt yuv422p10le
Sets the pixel format to a 10-bit 4:2:2 scheme, which is standard for ProRes and well-suited for high-resolution video.
-c:a copy
Copies the audio stream directly without re-encoding, thus preserving its quality and any associated metadata.
output.mov
Specifies the output file name and container. The MOV container is commonly used with ProRes.

Adjust filenames and parameters if needed for your particular workflow. This command should serve as a strong starting point for maintaining quality and metadata during your conversion.

Below is an ffmpeg command that downscales the 8K video (for example, to 4K resolution) and encodes it to 8-bit ProRes while copying metadata and audio:

ffmpeg -i input.mkv -map_metadata 0 -vf scale=3840:2160 -c:v prores_ks -profile:v 3 -qscale:v 1 -pix_fmt yuv422p -c:a copy output.mov

Explanation

-i input.mkv
Specifies your source file. Replace “input.mkv” with your actual filename.
-map_metadata 0
Copies all metadata from the input to the output, ensuring you keep embedded information.
-vf scale=3840:2160
Downscales the 8K video to 4K resolution (3840×2160). Adjust these dimensions as needed.
-c:v prores_ks
Uses the prores_ks encoder for ProRes output.
-profile:v 3
Chooses ProRes 422 (8-bit) instead of ProRes HQ (which is typically 10-bit).
-qscale:v 1
Sets a nearly lossless quality level for the video encoding.
-pix_fmt yuv422p
Sets the pixel format to 8-bit 4:2:2, matching the standard ProRes 422.
-c:a copy
Copies the audio stream without re-encoding.
output.mov
Specifies the output file name and container. The MOV container is common for ProRes.

Feel free to adjust the resolution or other parameters to best fit your workflow.

While ffmpeg supports CUDA acceleration for decoding H.265 (using decoders like hevc_cuvid), the ProRes encoders (e.g., prores_ks or prores_aw) remain CPU-bound. This means you can speed up the decoding stage with CUDA, but the encoding to ProRes itself won’t benefit from GPU acceleration.

For example, if you want to use CUDA for decoding, you could modify your command as follows:

ffmpeg -hwaccel cuda -c:v hevc_cuvid -i input.mkv -map_metadata 0 -vf scale=3840:2160 -c:v prores_ks -profile:v 3 -qscale:v 1 -pix_fmt yuv422p -c:a copy output.mov

Breakdown:

-hwaccel cuda -c:v hevc_cuvid
This tells ffmpeg to use CUDA for hardware-accelerated decoding of your H.265 input.
Rest of the command
The remaining options handle downscaling, encoding to ProRes 422 (8-bit), copying metadata, and preserving the audio stream.

Thus, you can leverage CUDA for decoding, but the ProRes encoding step will still be performed on the CPU.

the source is h265 .mp4. ffprobe shows this: Input #0, mov,mp4,m4a,3gp,3g2,mj2, from ‘C:\Users\GeorgeAdamon\Downloads\TheParty_8k_Mono.mp4’: Metadata: major_brand : isom minor_version : 512 compatible_brands: isomiso2mp41 creation_time : 2024-10-18T11:35:05.000000Z make : Insta360 model : Insta360 Pro2 encoder : Lavf58.13.101 description : {“info”:{“gyro_stabilized”:false,“initial_view_changed”:false}} Duration: 00:02:49.20, start: 0.000000, bitrate: 151956 kb/s Stream #0:00x1: Video: hevc (Main) (hvc1 / 0x31637668), yuv420p(tv), 7680x3840, 151440 kb/s, 29.97 fps, 29.97 tbr, 90k tbn (default) Metadata: creation_time : 2024-10-18T11:35:05.000000Z handler_name : VideoHandler vendor_id : [0][0][0][0] Side data: stereo3d: 2D spherical: equirectangular (0.000000/0.000000/0.000000) Stream #0:10x2: Audio: aac (mp4a / 0x6134706D), 48000 Hz, 4.0, fltp, 256 kb/s (default) Metadata: creation_time : 2024-10-18T11:35:05.000000Z handler_name : SoundHandler vendor_id : [0][0][0][0] Stream #0:20x3: Data: none (camm / 0x6D6D6163), 146 kb/s Metadata: creation_time : 2024-10-18T11:35:05.000000Z handler_name : CameraMetadataMotionHandler Unsupported codec with id 0 for input stream 2

I want it to work with Davinci

Below is an example command that uses CUDA to accelerate decoding of your H.265 MP4 while converting only the video and audio streams to an 8‑bit ProRes file that Davinci Resolve should accept:

ffmpeg -hwaccel cuda -c:v hevc_cuvid -i "C:\Users\GeorgeAdamon\Downloads\TheParty_8k_Mono.mp4" -map 0:v -map 0:a -map_metadata 0 -movflags use_metadata_tags -c:v prores_ks -profile:v 3 -vendor apl0 -qscale:v 1 -pix_fmt yuv422p -c:a copy output.mov

Explanation

-hwaccel cuda -c:v hevc_cuvid
Uses CUDA for hardware-accelerated decoding of the H.265 video stream.
-i "C:\Users\GeorgeAdamon\Downloads\TheParty_8k_Mono.mp4"
Specifies your input file (using quotes for Windows paths).
-map 0:v -map 0:a
Maps only the video and audio streams (ignoring the unsupported data stream).
-map_metadata 0 & -movflags use_metadata_tags
Copies over the global metadata—including your spherical/360° metadata—to the output file so that Davinci Resolve can recognize it.
-c:v prores_ks -profile:v 3 -vendor apl0 -qscale:v 1 -pix_fmt yuv422p
Encodes the video to ProRes 422 (8‑bit) with high quality. The -vendor apl0 flag is often used to improve compatibility with editing systems like Davinci Resolve.
-c:a copy
Copies the audio stream without re-encoding.

If you need to downscale the resolution (for instance, to 4K), you can insert a scale filter (e.g., add -vf scale=3840:2160 right after the input). Adjust the command as needed for your specific workflow.

If converting to ProRes causes unwanted resolution or scaling issues, you can try keeping the H.265 codec and simply rewrapping or re-encoding it in a way that “fixes” the compatibility problem for DaVinci Resolve.

Below are two approaches:

1. Remux the H.265 Streams into a Different Container

Sometimes DaVinci Resolve has issues with the MP4 container when it comes to 360° or H.265 files. You can rewrap (remux) the file into a MOV container without re-encoding the video or audio. This preserves the exact resolution and quality while keeping all the metadata:

ffmpeg -hwaccel cuda -c:v hevc_cuvid -i "C:\Users\GeorgeAdamon\Downloads\TheParty_8k_Mono.mp4" -map 0:v -map 0:a -map_metadata 0 -movflags use_metadata_tags -c:v copy -c:a copy output.mov

Key points:

-c:v copy -c:a copy: Copies the streams without re-encoding so the resolution remains exactly the same.
-movflags use_metadata_tags: Helps preserve metadata (including spherical/360° data) in the output MOV container.
-hwaccel cuda -c:v hevc_cuvid: Uses CUDA to accelerate decoding (although with copy mode, this is less critical).

2. Re-encode H.265 Using GPU Acceleration (NVENC)

If remuxing alone isn’t enough for DaVinci Resolve to handle the file, you can re-encode using H.265 with NVIDIA’s NVENC encoder. This way you keep the codec and resolution but may improve compatibility by having the file fully re-encoded:

ffmpeg -hwaccel cuda -c:v hevc_cuvid -i "C:\Users\GeorgeAdamon\Downloads\TheParty_8k_Mono.mp4" -map_metadata 0 -movflags use_metadata_tags -c:v hevc_nvenc -preset slow -rc vbr -cq 19 -c:a copy output.mp4

Key points:

-c:v hevc_nvenc: Re-encodes the video using NVIDIA’s H.265 encoder.
-preset slow -rc vbr -cq 19: Settings aimed at high-quality output; adjust these parameters if needed.
Keeps the original resolution and metadata, while potentially offering a file format that DaVinci Resolve can work with more reliably.

Choose the method that best fits your workflow. The first approach avoids any quality loss by not re-encoding, while the second provides a fully re-encoded H.265 file with GPU acceleration, which might be more compatible with some editing workflows in DaVinci Resolve.

When using NVENC you can speed up the encoding process by switching to a faster preset and slightly relaxing the quality settings. For instance, you could try:

ffmpeg -hwaccel cuda -c:v hevc_cuvid -i "C:\Users\GeorgeAdamon\Downloads\TheParty_8k_Mono.mp4" -map_metadata 0 -movflags use_metadata_tags -c:v hevc_nvenc -preset fast -rc vbr -cq 23 -c:a copy output.mp4

What Changed:

-preset fast
This replaces the slower preset, resulting in faster encoding.
-cq 23
A slightly higher (looser) quality setting can also speed up processing compared to a lower number like 19.

Keep in mind that even with NVENC, re-encoding 8K video can be demanding. If you still find it slow, you might consider:

Lowering the resolution (if acceptable)
Experimenting with other NVENC rate control modes (for example, using constant bitrate mode)
Checking if your GPU is fully supported and optimized for NVENC tasks

These adjustments should help increase the processing speed while still keeping the H.265 codec for compatibility with DaVinci Resolve.