Fabio
c36510c21e
Some checks failed
Linux Build / Linux cpu (push) Has been cancelled
Linux Build / Linux cpu-clang (push) Has been cancelled
Linux Build / Linux vulkan (push) Has been cancelled
mac-build / macOS CPU (push) Has been cancelled
Nix Build / Nix (cpu) (push) Has been cancelled
Nix Build / Nix (vulkan) (push) Has been cancelled
windows-build / Windows CPU (push) Has been cancelled
42 lines
1.5 KiB
C++
42 lines
1.5 KiB
C++
#pragma once
|
|
|
|
#include "engine/framework/audio/dsp.h"
|
|
#include "engine/models/voxtral_realtime/assets.h"
|
|
#include "engine/models/voxtral_realtime/types.h"
|
|
|
|
#include <cstdint>
|
|
#include <memory>
|
|
#include <vector>
|
|
|
|
namespace engine::models::voxtral_realtime {
|
|
|
|
struct VoxtralRealtimeFrontendStreamState {
|
|
std::vector<float> cached_magnitude_frame;
|
|
int64_t cached_freq_bins = 0;
|
|
bool cached_frame_ready = false;
|
|
};
|
|
|
|
class VoxtralRealtimeFrontend {
|
|
public:
|
|
explicit VoxtralRealtimeFrontend(std::shared_ptr<const VoxtralRealtimeAssets> assets);
|
|
|
|
VoxtralRealtimeFeatures extract(const runtime::AudioBuffer & audio, bool first_chunk) const;
|
|
// A steady chunk carries `steady_tokens` audio tokens of 80 ms each; batching them into one
|
|
// extraction yields 8 * steady_tokens feature frames from a single STFT pass.
|
|
VoxtralRealtimeFeatures extract_stream_chunk(
|
|
const runtime::AudioBuffer & audio,
|
|
bool first_chunk,
|
|
VoxtralRealtimeFrontendStreamState & state,
|
|
int64_t steady_tokens = 1) const;
|
|
|
|
int64_t first_stream_chunk_samples() const;
|
|
int64_t steady_stream_chunk_samples(int64_t steady_tokens = 1) const;
|
|
int64_t first_stream_chunk_advance_samples() const;
|
|
int64_t steady_stream_chunk_advance_samples(int64_t steady_tokens = 1) const;
|
|
|
|
private:
|
|
std::shared_ptr<const VoxtralRealtimeAssets> assets_;
|
|
engine::audio::SparseMelFilterbank mel_filterbank_;
|
|
};
|
|
|
|
} // namespace engine::models::voxtral_realtime
|