Bump audio.cpp runtime to 2026-09-05 build (Higgs end-of-clip hiss fix) - #14546
Merged
Conversation
Rebuilt in support-files from upstream main a8fccb47, which includes 0xShug0/audio.cpp#454: the Higgs Audio v3 codec decoder now gets right-hand context at the end of the stream, removing the rising hiss that ended every clip. Same three families as before (index_tts2, higgs_audio_tts, fish_audio). New archive hashes go in at index 0, so existing installs show the amber 'update available' status and the Update button in the audio.cpp settings. Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
pull Bot
pushed a commit
to A-Archives-and-Forks/subtitleedit
that referenced
this pull request
Sep 5, 2026
…-clip hiss After the codec-side fix (audio.cpp SubtitleEdit#454, SE PR SubtitleEdit#14546) Higgs clips still ended in a rising broadband hiss in roughly half of all generations. Decoding identical seeded codes with 8 or 64 frames of tail context gave the same samples, and holding the final frame's codes decoded to a steady hiss at the same level, so the noise is in the codes the language model emits, not in the codec. What decides it is the reference WAV: the model ends its clip the way the reference ends. References cut mid-noise or ending in room tone gave hissy endings; references ending in silence gave clean ones. Speak now conditions on a cached copy of the reference whose tail is trimmed at the pipeline's peak-relative threshold, faded out over 50 ms and followed by 400 ms of digital silence (CloneReferenceTail, prepared/<name>.wav next to the voice, keyed on size + mtime + recipe version). On 48 seeded clips per treatment the median level of the last four frames went from -46 dBFS to -64 dBFS and loud endings from 18 to 3, with no runaway generations in 192 requests. Preparation is best-effort and falls back to the file as is. Also retries once when audio.cpp reports "reached max_tokens before EOC" (one runaway in ~370 requests; no seed is sent so the retry samples afresh). Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
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.
Rebuilt the audio.cpp TTS runtime in support-files as audiocpp-indextts25-2026-09-05 from upstream main
a8fccb47, which includes 0xShug0/audio.cpp#454: the Higgs Audio v3 codec decoder now gets right-hand context at the end of the stream, so clips no longer end in a rising hiss.Same families as the previous build (IndexTTS-2.5, Higgs Audio v3, Fish S2 Pro), all seven platform archives incl. CUDA.
IndexTts25AudioCppDownloadServiceDownloadHashManager(mac hash verified locally against the downloaded archive), so existing installs show the amber update dot in the TTS engine list and the Update button in the audio.cpp settings dialog.🤖 Generated with Claude Code