Skip to content

Fix Higgs Audio codec decode seams #436

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
0xShug0 merged 1 commit into from
Sep 4, 2026
+11 −8
Merged

Fix Higgs Audio codec decode seams #436

Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
19 changes: 11 additions & 8 deletions src/models/higgs_audio_tts/codec.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -62,7 +62,7 @@ constexpr int64_t kCodecDecodeCapacityBucketFrames = 32;
constexpr int64_t kCodecHopLength = 960;
constexpr int64_t kCodecPadSamples = kCodecHopLength / 2;
constexpr int64_t kCodecDecodeWindowFrames = 128;
constexpr int64_t kCodecDecodeOverlapFrames = 8;
constexpr int64_t kCodecDecodeContextFrames = 32;
constexpr int64_t kCodecFullDecodeMaxFrames = 512;
constexpr int64_t kResidualDilations[] = {1, 3, 9};

Expand Down Expand Up @@ -1584,11 +1584,14 @@ HiggsCodecDecodeOutput HiggsCodecRuntime::decode_codes(const std::vector<int32_t
std::vector<int32_t> window_codes;
int64_t emitted_frames = 0;
while (emitted_frames < frames) {
const int64_t window_begin =
std::max<int64_t>(0, emitted_frames - kCodecDecodeOverlapFrames);
const int64_t emit_begin = emitted_frames;
const int64_t emit_end =
std::min<int64_t>(frames, emitted_frames + kCodecDecodeWindowFrames);
const int64_t window_frames = emit_end - window_begin;
std::min<int64_t>(frames, emit_begin + kCodecDecodeWindowFrames);
const int64_t window_begin =
std::max<int64_t>(0, emit_begin - kCodecDecodeContextFrames);
const int64_t window_end =
std::min<int64_t>(frames, emit_end + kCodecDecodeContextFrames);
const int64_t window_frames = window_end - window_begin;
window_codes.resize(static_cast<size_t>(window_frames * kCodecCodebooks));
for (int64_t frame = 0; frame < window_frames; ++frame) {
const auto src =
Expand All @@ -1602,9 +1605,9 @@ HiggsCodecDecodeOutput HiggsCodecRuntime::decode_codes(const std::vector<int32_t
const auto window = run_window(
window_codes,
window_frames,
kCodecDecodeWindowFrames + kCodecDecodeOverlapFrames);
const int64_t trim_frames = emitted_frames - window_begin;
const int64_t emit_frames = emit_end - emitted_frames;
kCodecDecodeWindowFrames + 2 * kCodecDecodeContextFrames);
const int64_t trim_frames = emit_begin - window_begin;
const int64_t emit_frames = emit_end - emit_begin;
const int64_t sample_begin = trim_frames * kCodecHopLength;
const int64_t sample_count = emit_frames * kCodecHopLength;
if (sample_begin < 0 || sample_count <= 0 ||
Expand Down
Loading