GPU Normalize Audio for Tdarr

USE AT YOUR OWN RISK

THIS IS WIP

IT MIGHT BURN YOUR HOUSE ;-)

TLDR

GPU Normalize Audio is a Tdarr FlowPlugin that tries to move FFmpeg loudnorm-style audio normalization onto the GPU while matching Tdarr's normal CPU-only Normalize Audio output.

What it does today:

• Normalizes every audio stream in a file, not only the first one.
• Keeps video, subtitles, chapters, metadata, attachments, and data streams.
• Uses CUDA for the loudness stats/apply path and streams decode/encode through FFmpeg.
• Avoids huge raw PCM bridge files for normal gpuSourcePort jobs.
• Keeps decoded audio parity as the top priority, even when that costs speed.
• Release smoke testing covers multiple input codec/layout cases, including AAC stereo, MP3 stereo, AC3 5.1, E-AC-3 5.1, DTS 5.1, and multi-audio inputs.

What we are trying to do:

• Match CPU Normalize Audio output first.
• Make the GPU path faster over time without cheating parity.
• Keep improving long-media speed; the current release beats CPU on tested long combined 5.1 plus generated 2-channel fallback jobs.

Plugin UI

Install

Download the latest GitHub release zip and extract it into your Tdarr plugins folder.

Keep the version folder. Do not flatten it.

Correct layout:

FlowPlugins/CommunityFlowPlugins/audio/gpuNormalizeAudio/1.2.8/

Tdarr loads:

FlowPlugins/CommunityFlowPlugins/audio/gpuNormalizeAudio/1.2.8/index.js

Tdarr Docker CUDA library path hint:

-e LD_LIBRARY_PATH=/usr/lib64:/usr/lib/x86_64-linux-gnu

Set this on the Tdarr Docker container if CUDA loading fails because the container resolves the wrong libcuda.so.1 first, for example with libcuda.so.1: file too short.

Defaults:

• Add Generated 2-Channel Track=true
• Track Order=eng,en
• Normalize ONLY Languages=eng,en (clear it or use all to normalize every language)
• Remove Other Languages=true
• Fallback To Undetected Audio=true
• Only Generate 2-Channel For First Language=true
• Max Concurrent Jobs=1
• Audio Bitrate=192k
• Integrated Loudness I=-18.0
• Loudness Range LRA=7.0
• True Peak TP=-2.0
• Max Gain dB=15
• Debug Logging=false

Performance (i9 9900k @ 5ghz vs. Nvidia 1050 TI)

Latest release: v1.2.8.

1.2.8 keeps the streaming gpuSourcePort path, promotes the accepted long-media optimization stack, fixes a post-success generated-stereo fallback teardown crash, and preserves exact decoded parity against Tdarr CPU Normalize Audio on the validated matrix. By default the release UI prefers English for both Track Order and Normalize ONLY Languages (eng,en), Remove Other Languages is on, and Fallback To Undetected Audio is on. Clear Normalize ONLY Languages or enter all to normalize every audio stream. If selected languages match no streams while removal is on, undetected audio is copied unchanged when present; if no desired or undetected audio would remain, the job fails instead of producing a no-audio output.

The table below uses a 5.1 source that has no existing stereo track. The GPU job writes both the normalized original 5.1 stream and the generated normalized 2-channel fallback, then both decoded streams are compared against Tdarr CPU Normalize Audio output. SRC_CPU time is split the same way: the original CPU Normalize Audio job for the 5.1 stream, plus a CPU Normalize Audio reference for the generated 2-channel source.

Speed vs CPU uses measured SRC_CPU/GPU wall time: above 1.0x is faster than CPU, below 1.0x is slower than CPU.

60min TLDR: original 5.1 2.842x faster (1123.3s CPU vs 395.3s GPU), generated 2ch 1.068x faster (283.4s CPU vs 265.4s GPU), combined job 3.522x faster (1406.7s CPU vs 399.4s GPU), parity pass for both decoded streams.

30min TLDR: original 5.1 2.633x faster (528.6s CPU vs 200.8s GPU), generated 2ch 1.065x faster (140.4s CPU vs 131.9s GPU), combined job 3.298x faster (669.0s CPU vs 202.8s GPU), parity pass for both decoded streams. 1.2.8 is a long-media combined-wall speed and stability release over 1.2.7.

60min

Workload	CPU reference	GPU 1.2.8	Speed vs CPU	Result
Original 5.1	1123.3s	395.3s	2.842x
Generated 2ch	283.4s	265.4s	1.068x
Combined	1406.7s	399.4s	3.522x

30min

Workload	CPU reference	GPU 1.2.8	Speed vs CPU	Result
Original 5.1	528.6s	200.8s	2.633x
Generated 2ch	140.4s	131.9s	1.065x
Combined	669.0s	202.8s	3.298x

10min

Workload	CPU reference	GPU 1.2.8	Speed vs CPU	Result
Original 5.1	168.6s	70.4s	2.395x
Generated 2ch	49.9s	44.4s	1.124x
Combined	218.5s	71.1s	3.073x

Short smoke clips are used only for parity checks and are intentionally excluded from the main README performance tables. Longer media is the intended target and is where the GPU path now catches and passes CPU on the tested 5.1 plus generated-2ch workload.

Codec smoke coverage for 1.2.8 covers AAC stereo, MP3 stereo, AC3 5.1, E-AC-3 5.1, DTS 5.1, and a multi-audio sample. The multi-audio smoke also verifies language-scope behavior: blank means all languages, omitted means release default English with other languages removed, Remove Other Languages=false copies non-selected audio unchanged, and no language match fails safely when no desired or undetected audio would remain.

2-channel fallback behavior:

Setting	Default	Behavior
Add Generated 2-Channel Track	true	Adds normalized generated stereo track(s) according to the selected 2-channel scope.
Track Order	eng,en	Comma-separated language priority used before choosing the generated 2-channel source. The default prefers English, while unlisted languages keep source order after listed languages. This uses source metadata and cannot fix missing or wrong language tags.
Normalize ONLY Languages	eng,en	Comma-separated language filter for streams to normalize. Clear it or enter all to normalize every audio stream.
Remove Other Languages	true	When Normalize ONLY Languages is a real language list, deletes audio streams outside that list. Disable it to copy other languages unchanged.
Fallback To Undetected Audio	true	If selected languages do not exist and removal is on, keep undetected audio streams unchanged when present. If no desired or undetected audio would remain, fail the job instead of outputting a no-audio file.
Only Generate 2-Channel For First Language	true	Ensures the first Track Order language has stereo. If that language lacks stereo, creates one generated track from the best same-language source, preferring 5.1/6-channel over 7.1/8-channel. Disable it to create one for every non-stereo audio stream/language.

Compared with releases before 1.1.11, the default tested contract when Add Generated 2-Channel Track=true includes generated 2-channel fallback work, and parity is checked for both the original audio stream and the generated fallback. Older release rows below used the previous primary-stream speed table, so treat the timing comparison as release history rather than a strict same-workload benchmark.

Version	10min GPU Time	30min GPU Time	Exact Stats Time	Exact Apply Time	Speed Change vs Previous
1.2.8	71.1s	202.8s	26.1s	124.2s
1.2.7	86.1s	241.0s	26.9s	140.0s
1.2.6	58.1s	172.2s	22.7s	133.1s
1.2.5	120.3s	359.2s	19.6s	100.0s
1.2.4	134.8s	400.8s	20.3s	122.1s
1.2.3	144.2s	405.3s	215.4s	121.2s
1.2.2	177.5s	450.6s	210.6s	124.0s
1.2.1	150.9s	438.5s	244.3s	131.9s
1.2.0	155.7s	452.3s	261.5s	138.1s
1.1.13	155.3s	456.8s	262.2s	138.7s
1.1.12	196.5s	568.4s	222.0s	169.5s
1.1.11	217.2s	634.9s	252.9s	162.8s
1.1.10	114.9s	334.0s	108.5s	97.9s

Choose 1.2.8 for current parity, the fastest validated long combined jobs, the stereo-fallback crash fix, accepted long-media scheduling/defaults, fallback to undetected audio when language tags are missing, safe no-audio failure behavior, and release/diagnostic tooling hardening. Use 1.2.7 only if you need the previous speed-release snapshot.

For 1.2.3, the exact-stats workers run as an overlapped pair. The history row lists the slower worker time, not the sum of both overlapped worker counters.

Versions

Version	Use
1.2.8	Current release. Promotes the accepted long-media optimization defaults, fixes a post-success generated-stereo fallback teardown crash, and matches CPU decoded output for normalized 5.1 plus generated 2-channel fallback on the validated matrix.
1.2.7	Previous release. Runs same-source original plus generated 2-channel fallback routes in parallel on long jobs, keeps short-media CPU first-pass for codec parity, and matches CPU decoded output for normalized 5.1 plus generated 2-channel fallback on the full validated matrix.
1.2.6	Previous release. Uses original-only GPU first-pass acceleration to speed up long 5.1 plus generated 2-channel fallback jobs, keeps fallback to undetected audio and safe no-audio failure behavior, and matches CPU decoded output for normalized 5.1 plus generated 2-channel fallback on the full validated matrix.
1.2.5	Previous release. Speeds up long 5.1 plus generated 2-channel fallback jobs, keeps fallback to undetected audio and safe no-audio failure behavior, and matches CPU decoded output for normalized 5.1 plus generated 2-channel fallback on the full validated matrix.
1.2.4	Previous release. Adds fallback to undetected audio, fails safely if selected language removal would leave no audio, keeps faster-than-CPU long combined jobs, and matches CPU decoded output for normalized 5.1 plus generated 2-channel fallback on the full validated matrix.
1.2.3	Previous release. Improves long combined-job speed, keeps default-on removal of audio languages outside Normalize ONLY Languages, adds release manifest checks and process cleanup hardening, and matches CPU decoded output for normalized 5.1 plus generated 2-channel fallback on the full validated matrix.
1.2.2	Previous release. Adds default-on removal of audio languages outside Normalize ONLY Languages, keeps opt-out copy behavior, promotes current speed defaults, and matches CPU decoded output for normalized 5.1 plus generated 2-channel fallback on validated cases.
1.2.1	Previous release. Adds language-only normalization, no-match pass-through, cleaner default logs, deploy cleanup, and matches CPU decoded output for normalized 5.1 plus generated 2-channel fallback on the required matrix.
1.2.0	Previous release. Fixes generated-stereo first-language scope, keeps streaming gpuSourcePort, and matches CPU decoded output for normalized 5.1 plus generated 2-channel fallback on the required matrix.
1.1.13	Previous release. Speeds up long generated-stereo fallback jobs again, keeps streaming gpuSourcePort, and matches CPU decoded output for normalized 5.1 plus generated 2-channel fallback on the required matrix.
1.1.12	Older release. Speeds up long generated-stereo fallback jobs, keeps streaming gpuSourcePort, and matches CPU decoded output for normalized 5.1 plus generated 2-channel fallback on the required matrix.
1.1.11	Keeps streaming gpuSourcePort, matches CPU decoded output for normalized 5.1 plus generated 2-channel fallback on tested long media, and adds language-priority controls for first-language versus all-language 2-channel fallback generation.
1.1.10	Uses smaller streaming chunks, stats decode prefetch, and safe unsafe-output-feedback skipping after threshold lock; required parity matrix passed. Faster than CPU on all tested required limiter-heavy primary-stream cases.
1.1.9	Parallelizes safe feedback-skip apply windows while preserving exact state and decoded parity; required parity matrix passed. Faster than CPU on tested 10min/30min limiter-heavy media.
1.1.8	Skips safe output-feedback accumulation after the feedback threshold is active while preserving normal limiter/output rendering; required parity matrix passed. Still slower than CPU on long limiter-heavy media.
1.1.7	Slot-accumulated safe feedback, parallel unsafe feedback, smaller chunks, and asymmetric risk planning speed up limiter-heavy 5.1 jobs; required parity matrix passed. Still slower than CPU on long limiter-heavy media.
1.1.6	Segmented exact safe feedback speeds up limiter-heavy 5.1 jobs, required parity matrix passed. Still slower than CPU on long limiter-heavy media.
1.1.5	Modularized GPU runtime, required parity matrix passed. Performance is effectively unchanged from 1.1.4; still slower than CPU on long limiter-heavy media.
1.1.4	Faster exact stats and limiter-active GPU path, required parity matrix passed. Still slower than CPU on long limiter-heavy media.
1.1.3	Faster limiter-active GPU path, required parity matrix passed. Still slower than CPU on long limiter-heavy media.
1.1.2	Streaming two-pass release. Avoids huge raw PCM bridge files and fixed long-case parity.
1.1.1	Added guarded GPU normalize concurrency.
1.1	Older optimized exact GPU line.
1.0	First stable CPU-output matching line.
0.0.x	Old pre-stable snapshots. Some are known not to match CPU normalizer output. Use only for rollback/debug.