Kedara Studios a48510a58d refactor: batch GPU decode calls in batch_stream_generate_voice_clone ...

Restructure the batch streaming loop and flush into a 3-phase approach:
1) collect decode windows for all active items, 2) perform a single
batched GPU decode call, 3) per-item post-processing (crossfade, trim).

Add decode_streaming_batch() to Qwen3TTSTokenizer to support B>1 batch
decoding. Remove examples/test_batch_streaming.py test script.

2026-02-09 01:57:30 +01:00

..

profile_talker.py

,,,

2026-01-26 06:16:51 +04:00

test_model_12hz_base.py

Initial commit

2026-01-22 17:57:04 +08:00

test_model_12hz_custom_voice.py

Initial commit

2026-01-22 17:57:04 +08:00

test_model_12hz_voice_design.py

Initial commit

2026-01-22 17:57:04 +08:00

test_optimized_no_streaming.py

reference changed

2026-01-26 23:14:50 +04:00

test_streaming_optimized.py

fix(streaming): add ref_code context for stable first chunks

2026-01-27 19:14:52 +04:00

test_streaming.py

minor

2026-01-26 23:24:18 +04:00

test_tokenizer_12hz.py

Initial commit

2026-01-22 17:57:04 +08:00