Repository Issue Activity (beta)

#3955 ggml-large-v3.bin: Running into an endless loop on transcription while ggml-medium.en.bin works correctly

#3954 Can whisper_full reuse the encoder output from language auto-detection for the first chunk?

#3953 Add REFUTE scientific critique + calibration benchmark

#3952 [feature request] Enhance server.cpp handle concurency

#3951 Streaming VAD segmentation degrades transcription accuracy on short phrases and boundary cuts

#3933 Missing validation of attacker-controlled n_fft in Parakeet model loader (src/parakeet.cpp) — DoS via uncaught exception

#3853 GigaAM support

#3949 Feature Request: Installable package via winget

#3648 running Whisper.cpp on Iphone 17 pro max

#3944 Stack buffer overflow in whisper_model_load: file-controlled n_dims written past ne[4]

#3942 [WhisperSmash] Heap-buffer-overflow in whisper_set_mel_with_state

#3937 Transcription of any language other than English doesn't work.

#3708 OpenCL for Adreno support

#3914 coreml: expose configurable MLComputeUnits for encoder fallback

#1051 Compile in Vietnamese language error

#587 Each sentence has a leading space when outputting to .txt file

#3935 Adreno GPU support broken after commit `896718e`

#3932 parakeet: duration argmax runs on log-probs that underflow to -inf, silently picking duration 0

#3930 Why Silero-VAD that is state-sponsored Russian software exist in this codebase?

#3924 Reachable assertion / abort in whisper_model_load() on an invalid ftype

#3923 Heap out-of-bounds read in log_mel_spectrogram() on very short audio (fewer than 201 samples)

#3118 feature request: support parakeet-tdt-0.6b-v2(https://huggingface.co/nvidia/parakeet-tdt-0.6b-v2)

#3918 Whisper-server exits when VAD finds no speech in silent WAV

#3576 Strix Halo ROCm 7 `HSA_STATUS_ERROR_MEMORY_APERTURE_VIOLATION`

#3779 Token-level garbled characters when transcribing Japanese with medium / large-v3-turbo models