Don't let span chunking change the resampler's output - #915
Open
benface wants to merge 1 commit into
Open
Conversation
`fill_input_buffer` treats `current_span_len()` as the samples left in the span, but it's the span's total length, so it stops reading at the end of every span. Vorbis packets are smaller than a resampler chunk, so every chunk ends up short, gets marked `partial_len`, and rubato pads it with zeros. The same audio comes out different depending on how the decoder split it. Stop only where the rate or channel count actually changes, and check that before reading anything from the next span. When reading stops there, rebuild the resampler. That fixes a second bug: a span that changed the sample rate was still being resampled at the old one. 441 frames at 44.1k then 441 at 22.05k, converted to 48k, gave 962 samples instead of 1440. Drop the counter the boundary detection was using — reading stops at the change itself now, so there's nothing left to count. Only the passthrough still needs it. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
benface
force-pushed
the
fix/resampler-span-boundaries
branch
from
August 6, 2026 01:17
1623f94 to
6c15966
Compare
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Resampler output changes depending on how the input happens to be split into spans, which it shouldn't.
fill_input_bufferstops at every span boundary and marks the read partial, so rubato zero-pads it as if the stream had ended. Decoders report a span per packet, so this happens constantly — a 48 kHz file on a 44.1 kHz device hisses the whole way through. It now reads across spans and only marks partial at a real end or a format change.The resamplers were also tracking input position against span length to spot boundaries, but they read ahead into their own buffer, so that count never matched where the input actually was. Filling already stops at a format change, so the tracking is gone — the passthrough keeps its own, where it's accurate.
Tests cover the chunking, plus a sample rate change and a channel count change at a boundary. The chunking one fails on master.