sortformer

.gitignore

@@ -59,6 +59,10 @@ __pycache__/
 
 # WER evaluation data + generated working reports.
 /samples/wer/
+# Diarization eval corpora (AMI audio + fetched RTTMs); regenerable via
+# scripts/diar/ingest_ami.py + fetch_ami_forced_alignment.py. The committed
+# oracle case lives at samples/sortformer-2spk-mix.wav (outside samples/diar/).
+/samples/diar/
 /reports/*
 
 # Reviewable porting evidence (intake, porting log) is intentionally committed.

README.md

@@ -27,6 +27,7 @@ C/C++ speech-to-text inference library. Runs diverse STT model families via [GGU
 | Voxtral Realtime | `voxtral-mini-4b-realtime-2602` (streaming audio-LLM) | [docs/models/voxtral-realtime.md](docs/models/voxtral-realtime.md) |
 | MedASR | `medasr` (Conformer + CTC, English medical-dictation, gated) | [docs/models/medasr.md](docs/models/medasr.md) |
 | MOSS Transcribe-Diarize | `moss-transcribe-diarize` (audio-LLM; English + Chinese ASR with inline speaker diarization) | [docs/models/moss-transcribe-diarize.md](docs/models/moss-transcribe-diarize.md) |
+| Sortformer | `diar_streaming_sortformer_4spk-v2.1` (streaming speaker diarizer, up to 4 speakers; no transcription) | [docs/models/diar_streaming_sortformer_4spk-v2.1.md](docs/models/diar_streaming_sortformer_4spk-v2.1.md) |
 
 Per-variant model cards live under [`docs/models/`](docs/models/).
 

bindings/python/src/transcribe_cpp/__init__.py

@@ -54,6 +54,7 @@
 Task = Literal["transcribe", "translate"]
 Timestamps = Literal["none", "auto", "segment", "word", "token"]
 Diarize = Literal["default", "off", "on"]
+SortformerPreset = Literal["default", "very_high_latency", "high_latency", "low_latency"]
 CommitPolicy = Literal["auto", "on_finalize", "stable_prefix"]
 Feature = Literal[
     "initial_prompt", "temperature_fallback", "long_form",
@@ -81,8 +82,10 @@
     "MoonshineStreamingOptions",
     "ParakeetStreamOptions",
     "ParakeetBufferedStreamOptions",
+    "SortformerStreamOptions",
     "VoxtralRealtimeStreamOptions",
     "Backend",
+    "SortformerPreset",
     "KVType",
     "Task",
     "Timestamps",
@@ -799,6 +802,40 @@ def _apply(self, ext) -> None:
             ext.min_decode_interval_ms = self.min_decode_interval_ms
 
 
+class SortformerStreamOptions(FamilyExtension):
+    """Sortformer streaming operating-point options (run slot).
+
+    Sortformer is a diarizer: a run produces speaker segments, no text.
+    ``preset`` selects the latency / accuracy trade-off from the model's
+    published menu; ``"default"`` keeps the GGUF-shipped checkpoint
+    configuration. ``"very_high_latency"`` (~30 s lookahead) is the
+    offline-file operating point; ``"low_latency"`` (~1 s) is the
+    real-time point and costs substantially more compute per audio
+    second."""
+
+    _slot = "run"
+    _kind = _generated.TRANSCRIBE_EXT_KIND_SORTFORMER_STREAM
+    _struct = _generated.transcribe_sortformer_stream_ext
+    _init = "transcribe_sortformer_stream_ext_init"
+
+    _presets = {
+        "default": _generated.TRANSCRIBE_SORTFORMER_PRESET_DEFAULT,
+        "very_high_latency": _generated.TRANSCRIBE_SORTFORMER_PRESET_VERY_HIGH_LATENCY,
+        "high_latency": _generated.TRANSCRIBE_SORTFORMER_PRESET_HIGH_LATENCY,
+        "low_latency": _generated.TRANSCRIBE_SORTFORMER_PRESET_LOW_LATENCY,
+    }
+
+    def __init__(self, *, preset: SortformerPreset | None = None):
+        if preset is not None and preset not in self._presets:
+            raise ValueError(f"unknown sortformer preset {preset!r}; "
+                             f"expected one of {sorted(self._presets)}")
+        self.preset = preset
+
+    def _apply(self, ext) -> None:
+        if self.preset is not None:
+            ext.preset = self._presets[self.preset]
+
+
 # --- high-level handles ---------------------------------------------------
 
 

bindings/python/src/transcribe_cpp/_generated.py

@@ -13,7 +13,7 @@
 # Stable digest of the ABI surface below (structs, enums, macros, layout,
 # prototypes). A native provider package echoes this back so the API
 # package can reject an ABI-mismatched provider before dlopen.
-PUBLIC_HEADER_HASH = "d67a9bd78b964445"
+PUBLIC_HEADER_HASH = "fb2e64791dcbb70a"
 
 # === enum constants ===
 TRANSCRIBE_OK = 0
@@ -101,13 +101,18 @@
 TRANSCRIBE_STREAM_COMMIT_AUTO = 0
 TRANSCRIBE_STREAM_COMMIT_ON_FINALIZE = 1
 TRANSCRIBE_STREAM_COMMIT_STABLE_PREFIX = 2
+TRANSCRIBE_SORTFORMER_PRESET_DEFAULT = 0
+TRANSCRIBE_SORTFORMER_PRESET_VERY_HIGH_LATENCY = 1
+TRANSCRIBE_SORTFORMER_PRESET_HIGH_LATENCY = 2
+TRANSCRIBE_SORTFORMER_PRESET_LOW_LATENCY = 3
 TRANSCRIBE_WHISPER_PROMPT_FIRST_SEGMENT = 0
 TRANSCRIBE_WHISPER_PROMPT_ALL_SEGMENTS = 1
 
 # === macro constants (integer object-like macros) ===
 TRANSCRIBE_EXT_KIND_MOONSHINE_STREAMING_STREAM = 1414746957
 TRANSCRIBE_EXT_KIND_PARAKEET_BUFFERED_STREAM = 1396853584
 TRANSCRIBE_EXT_KIND_PARAKEET_STREAM = 1414744912
+TRANSCRIBE_EXT_KIND_SORTFORMER_STREAM = 1414743635
 TRANSCRIBE_EXT_KIND_VOXTRAL_REALTIME_STREAM = 1414746710
 TRANSCRIBE_EXT_KIND_WHISPER_RUN = 1314015319
 
@@ -148,6 +153,8 @@ class transcribe_parakeet_stream_ext(_c.Structure):
     pass
 class transcribe_parakeet_buffered_stream_ext(_c.Structure):
     pass
+class transcribe_sortformer_stream_ext(_c.Structure):
+    pass
 class transcribe_voxtral_realtime_stream_ext(_c.Structure):
     pass
 class transcribe_whisper_run_ext(_c.Structure):
@@ -173,6 +180,7 @@ class transcribe_whisper_chunk_trace(_c.Structure):
 transcribe_moonshine_streaming_stream_ext._fields_ = [("ext", transcribe_ext), ("min_decode_interval_ms", _c.c_int32)]
 transcribe_parakeet_stream_ext._fields_ = [("ext", transcribe_ext), ("att_context_right", _c.c_int32)]
 transcribe_parakeet_buffered_stream_ext._fields_ = [("ext", transcribe_ext), ("left_ms", _c.c_int32), ("chunk_ms", _c.c_int32), ("right_ms", _c.c_int32)]
+transcribe_sortformer_stream_ext._fields_ = [("ext", transcribe_ext), ("preset", _c.c_int)]
 transcribe_voxtral_realtime_stream_ext._fields_ = [("ext", transcribe_ext), ("num_delay_tokens", _c.c_int32), ("min_decode_interval_ms", _c.c_int32)]
 transcribe_whisper_run_ext._fields_ = [("ext", transcribe_ext), ("initial_prompt", _c.c_char_p), ("prompt_tokens", _c.POINTER(_c.c_int32)), ("n_prompt_tokens", _c.c_size_t), ("prompt_condition", _c.c_int), ("condition_on_prev_tokens", _c.c_bool), ("max_prev_context_tokens", _c.c_int32), ("temperature", _c.c_float), ("temperature_inc", _c.c_float), ("compression_ratio_thold", _c.c_float), ("logprob_thold", _c.c_float), ("no_speech_thold", _c.c_float), ("seed", _c.c_uint32), ("max_initial_timestamp", _c.c_float)]
 transcribe_whisper_chunk_trace._fields_ = [("struct_size", _c.c_uint64), ("t0_ms", _c.c_int64), ("t1_ms", _c.c_int64), ("temperature_used", _c.c_float), ("compression_ratio", _c.c_float), ("avg_logprob", _c.c_float), ("no_speech_prob", _c.c_float), ("no_speech_triggered", _c.c_bool), ("n_fallbacks", _c.c_int32)]
@@ -217,6 +225,7 @@ class transcribe_whisper_chunk_trace(_c.Structure):
     'transcribe_moonshine_streaming_stream_ext': {'size': 24, 'align': 8, 'offsets': {'ext': 0, 'min_decode_interval_ms': 16}},
     'transcribe_parakeet_stream_ext': {'size': 24, 'align': 8, 'offsets': {'ext': 0, 'att_context_right': 16}},
     'transcribe_parakeet_buffered_stream_ext': {'size': 32, 'align': 8, 'offsets': {'ext': 0, 'left_ms': 16, 'chunk_ms': 20, 'right_ms': 24}},
+    'transcribe_sortformer_stream_ext': {'size': 24, 'align': 8, 'offsets': {'ext': 0, 'preset': 16}},
     'transcribe_voxtral_realtime_stream_ext': {'size': 24, 'align': 8, 'offsets': {'ext': 0, 'num_delay_tokens': 16, 'min_decode_interval_ms': 20}},
     'transcribe_whisper_run_ext': {'size': 80, 'align': 8, 'offsets': {'ext': 0, 'initial_prompt': 16, 'prompt_tokens': 24, 'n_prompt_tokens': 32, 'prompt_condition': 40, 'condition_on_prev_tokens': 44, 'max_prev_context_tokens': 48, 'temperature': 52, 'temperature_inc': 56, 'compression_ratio_thold': 60, 'logprob_thold': 64, 'no_speech_thold': 68, 'seed': 72, 'max_initial_timestamp': 76}},
     'transcribe_whisper_chunk_trace': {'size': 48, 'align': 8, 'offsets': {'struct_size': 0, 't0_ms': 8, 't1_ms': 16, 'temperature_used': 24, 'compression_ratio': 28, 'avg_logprob': 32, 'no_speech_prob': 36, 'no_speech_triggered': 40, 'n_fallbacks': 44}},
@@ -365,6 +374,8 @@ def configure(lib):
     lib.transcribe_session_params_init.argtypes = [_c.POINTER(transcribe_session_params)]
     lib.transcribe_set_abort_callback.restype = None
     lib.transcribe_set_abort_callback.argtypes = [_c.c_void_p, _c.CFUNCTYPE(_c.c_bool, _c.c_void_p), _c.c_void_p]
+    lib.transcribe_sortformer_stream_ext_init.restype = None
+    lib.transcribe_sortformer_stream_ext_init.argtypes = [_c.POINTER(transcribe_sortformer_stream_ext)]
     lib.transcribe_speaker_segment_init.restype = None
     lib.transcribe_speaker_segment_init.argtypes = [_c.POINTER(transcribe_speaker_segment)]
     lib.transcribe_status_string.restype = _c.c_char_p

bindings/python/tests/test_family_ext.py

@@ -22,6 +22,7 @@
     t.MoonshineStreamingOptions,
     t.ParakeetStreamOptions,
     t.ParakeetBufferedStreamOptions,
+    t.SortformerStreamOptions,
     t.VoxtralRealtimeStreamOptions,
 ]
 
@@ -73,6 +74,19 @@ def test_parakeet_buffered_partial_overrides():
     assert built.right_ms == fresh.right_ms
 
 
+def test_sortformer_preset_maps_to_enum_value():
+    built = t.SortformerStreamOptions(preset="very_high_latency")._build()
+    assert built.preset == _generated.TRANSCRIBE_SORTFORMER_PRESET_VERY_HIGH_LATENCY
+    # None keeps the init default (DEFAULT = GGUF-shipped cfg).
+    default = t.SortformerStreamOptions()._build()
+    assert default.preset == _generated.TRANSCRIBE_SORTFORMER_PRESET_DEFAULT
+
+
+def test_sortformer_unknown_preset_rejected():
+    with pytest.raises(ValueError, match="preset"):
+        t.SortformerStreamOptions(preset="ultra_low_latency")  # type: ignore[arg-type]
+
+
 # --- model-gated: resolve_family validation + a real extension run ----------
 
 

bindings/rust/sys/src/transcribe_sys.rs

@@ -1,17 +1,18 @@
 // @generated by `cargo xtask bindgen` from include/transcribe/extensions.h
 // DO NOT EDIT BY HAND. Regenerate: `cargo xtask bindgen`.
-// Pinned to include/transcribe.abihash = d67a9bd78b964445
+// Pinned to include/transcribe.abihash = fb2e64791dcbb70a
 
 /// The public-ABI digest these bindings were generated against
 /// (sha256/16 over the normalized FFI surface). The load-time version
 /// gate and the CI drift check both anchor on this value.
-pub const PUBLIC_HEADER_HASH: &str = "d67a9bd78b964445";
+pub const PUBLIC_HEADER_HASH: &str = "fb2e64791dcbb70a";
 
 /* automatically generated by rust-bindgen 0.72.1 */
 
 pub const TRANSCRIBE_EXT_KIND_MOONSHINE_STREAMING_STREAM: u32 = 1414746957;
 pub const TRANSCRIBE_EXT_KIND_PARAKEET_STREAM: u32 = 1414744912;
 pub const TRANSCRIBE_EXT_KIND_PARAKEET_BUFFERED_STREAM: u32 = 1396853584;
+pub const TRANSCRIBE_EXT_KIND_SORTFORMER_STREAM: u32 = 1414743635;
 pub const TRANSCRIBE_EXT_KIND_VOXTRAL_REALTIME_STREAM: u32 = 1414746710;
 pub const TRANSCRIBE_EXT_KIND_WHISPER_RUN: u32 = 1314015319;
 impl transcribe_status {
@@ -1156,6 +1157,39 @@ unsafe extern "C" {
         ext: *mut transcribe_parakeet_buffered_stream_ext,
     );
 }
+impl transcribe_sortformer_preset {
+    pub const TRANSCRIBE_SORTFORMER_PRESET_DEFAULT: transcribe_sortformer_preset =
+        transcribe_sortformer_preset(0);
+    pub const TRANSCRIBE_SORTFORMER_PRESET_VERY_HIGH_LATENCY: transcribe_sortformer_preset =
+        transcribe_sortformer_preset(1);
+    pub const TRANSCRIBE_SORTFORMER_PRESET_HIGH_LATENCY: transcribe_sortformer_preset =
+        transcribe_sortformer_preset(2);
+    pub const TRANSCRIBE_SORTFORMER_PRESET_LOW_LATENCY: transcribe_sortformer_preset =
+        transcribe_sortformer_preset(3);
+}
+#[repr(transparent)]
+#[derive(Debug, Copy, Clone, Hash, PartialEq, Eq)]
+pub struct transcribe_sortformer_preset(pub ::std::os::raw::c_uint);
+#[repr(C)]
+#[derive(Debug, Copy, Clone)]
+pub struct transcribe_sortformer_stream_ext {
+    pub ext: transcribe_ext,
+    pub preset: transcribe_sortformer_preset,
+}
+#[allow(clippy::unnecessary_operation, clippy::identity_op)]
+const _: () = {
+    ["Size of transcribe_sortformer_stream_ext"]
+        [::std::mem::size_of::<transcribe_sortformer_stream_ext>() - 24usize];
+    ["Alignment of transcribe_sortformer_stream_ext"]
+        [::std::mem::align_of::<transcribe_sortformer_stream_ext>() - 8usize];
+    ["Offset of field: transcribe_sortformer_stream_ext::ext"]
+        [::std::mem::offset_of!(transcribe_sortformer_stream_ext, ext) - 0usize];
+    ["Offset of field: transcribe_sortformer_stream_ext::preset"]
+        [::std::mem::offset_of!(transcribe_sortformer_stream_ext, preset) - 16usize];
+};
+unsafe extern "C" {
+    pub fn transcribe_sortformer_stream_ext_init(ext: *mut transcribe_sortformer_stream_ext);
+}
 #[repr(C)]
 #[derive(Debug, Copy, Clone)]
 pub struct transcribe_voxtral_realtime_stream_ext {

bindings/rust/transcribe-cpp/src/family.rs

@@ -60,11 +60,55 @@ pub struct VoxtralRealtimeStreamOptions {
     pub min_decode_interval_ms: Option<i32>,
 }
 
+/// Sortformer streaming operating point (latency / accuracy trade-off).
+/// The menu is discrete (jointly-tuned bundles), not a latency dial;
+/// `Default` keeps the GGUF-shipped checkpoint configuration.
+#[derive(Debug, Clone, Copy, Default, PartialEq, Eq)]
+pub enum SortformerPreset {
+    #[default]
+    Default,
+    /// ~30.4 s algorithmic lookahead; the offline-file operating point.
+    VeryHighLatency,
+    /// ~10.0 s lookahead.
+    HighLatency,
+    /// ~1.04 s lookahead; the real-time point (compute-heavy per audio
+    /// second — many small windows).
+    LowLatency,
+}
+
+impl SortformerPreset {
+    fn to_sys(self) -> sys::transcribe_sortformer_preset {
+        match self {
+            SortformerPreset::Default => {
+                sys::transcribe_sortformer_preset::TRANSCRIBE_SORTFORMER_PRESET_DEFAULT
+            }
+            SortformerPreset::VeryHighLatency => {
+                sys::transcribe_sortformer_preset::TRANSCRIBE_SORTFORMER_PRESET_VERY_HIGH_LATENCY
+            }
+            SortformerPreset::HighLatency => {
+                sys::transcribe_sortformer_preset::TRANSCRIBE_SORTFORMER_PRESET_HIGH_LATENCY
+            }
+            SortformerPreset::LowLatency => {
+                sys::transcribe_sortformer_preset::TRANSCRIBE_SORTFORMER_PRESET_LOW_LATENCY
+            }
+        }
+    }
+}
+
+/// Sortformer diarizer run-extension knobs (run slot). Sortformer produces
+/// speaker segments, no text; read results via the speaker-segment
+/// accessors. `None` keeps the family default (the GGUF-shipped cfg).
+#[derive(Debug, Clone, Default, PartialEq, Eq)]
+pub struct SortformerStreamOptions {
+    pub preset: Option<SortformerPreset>,
+}
+
 /// A family extension for the run slot (offline `run`/`run_batch`).
 #[derive(Debug, Clone, PartialEq)]
 #[non_exhaustive]
 pub enum RunExtension {
     Whisper(WhisperRunOptions),
+    Sortformer(SortformerStreamOptions),
 }
 
 /// A family extension for the stream slot.
@@ -86,6 +130,7 @@ pub(crate) enum RunExtRaw {
         ext: Box<sys::transcribe_whisper_run_ext>,
         _prompt: Option<CString>,
     },
+    Sortformer(Box<sys::transcribe_sortformer_stream_ext>),
 }
 
 impl RunExtRaw {
@@ -95,6 +140,9 @@ impl RunExtRaw {
             RunExtRaw::Whisper { ext, .. } => {
                 (&**ext) as *const sys::transcribe_whisper_run_ext as *const sys::transcribe_ext
             }
+            RunExtRaw::Sortformer(e) => {
+                (&**e) as *const sys::transcribe_sortformer_stream_ext as *const sys::transcribe_ext
+            }
         }
     }
 }
@@ -128,6 +176,12 @@ impl RunExtension {
                     _prompt: prompt,
                 })
             }
+            RunExtension::Sortformer(o) => {
+                let mut ext: sys::transcribe_sortformer_stream_ext = unsafe { std::mem::zeroed() };
+                unsafe { sys::transcribe_sortformer_stream_ext_init(&mut ext) };
+                set(&mut ext.preset, o.preset.map(SortformerPreset::to_sys));
+                Ok(RunExtRaw::Sortformer(Box::new(ext)))
+            }
         }
     }
 }

bindings/rust/transcribe-cpp/src/lib.rs

@@ -65,7 +65,8 @@ pub use cancel::CancelToken;
 pub use error::{Error, Result};
 pub use family::{
     MoonshineStreamingOptions, ParakeetBufferedStreamOptions, ParakeetStreamOptions, RunExtension,
-    StreamExtension, VoxtralRealtimeStreamOptions, WhisperRunOptions,
+    SortformerPreset, SortformerStreamOptions, StreamExtension, VoxtralRealtimeStreamOptions,
+    WhisperRunOptions,
 };
 pub use logging::{disable_logging, init_logging};
 pub use model::{Capabilities, Model, ModelOptions, SessionLimits, SessionOptions};

bindings/swift/Sources/TranscribeCpp/ABIHash.swift

@@ -13,7 +13,7 @@ import CTranscribe
 extension Transcribe {
     /// sha256/16 of the normalized public FFI surface, pinned to the value in
     /// include/transcribe.abihash at the time this binding was last reviewed.
-    public static let pinnedHeaderHash = "d67a9bd78b964445"
+    public static let pinnedHeaderHash = "fb2e64791dcbb70a"
 
     /// The public-ABI digest this binding was reviewed against (16 hex chars).
     public static func headerHash() -> String { pinnedHeaderHash }

bindings/swift/Sources/TranscribeCpp/Family.swift

@@ -7,7 +7,7 @@ import CTranscribe
 // `transcribe_ext` pointer is handed to the run/begin call; the library copies
 // what it needs before returning.
 
-// MARK: - Run-slot extensions (whisper)
+// MARK: - Run-slot extensions (whisper, sortformer)
 
 public struct WhisperRunOptions: Sendable {
     public var initialPrompt: String?
@@ -46,12 +46,42 @@ public struct WhisperRunOptions: Sendable {
     }
 }
 
+/// Sortformer streaming operating point (latency / accuracy trade-off).
+/// The menu is discrete (jointly-tuned bundles), not a latency dial;
+/// `.default` keeps the GGUF-shipped checkpoint configuration.
+/// `.veryHighLatency` (~30 s lookahead) is the offline-file operating
+/// point; `.lowLatency` (~1 s) is the real-time point.
+public enum SortformerPreset: Sendable {
+    case `default`
+    case veryHighLatency
+    case highLatency
+    case lowLatency
+
+    var cValue: transcribe_sortformer_preset {
+        switch self {
+        case .default: return TRANSCRIBE_SORTFORMER_PRESET_DEFAULT
+        case .veryHighLatency: return TRANSCRIBE_SORTFORMER_PRESET_VERY_HIGH_LATENCY
+        case .highLatency: return TRANSCRIBE_SORTFORMER_PRESET_HIGH_LATENCY
+        case .lowLatency: return TRANSCRIBE_SORTFORMER_PRESET_LOW_LATENCY
+        }
+    }
+}
+
+/// Sortformer diarizer options (run slot). Sortformer produces speaker
+/// segments, no text; read results via the speaker-segment accessors.
+public struct SortformerStreamOptions: Sendable {
+    public var preset: SortformerPreset?
+    public init(preset: SortformerPreset? = nil) { self.preset = preset }
+}
+
 public enum RunExtension: Sendable {
     case whisper(WhisperRunOptions)
+    case sortformer(SortformerStreamOptions)
 
     var kind: UInt32 {
         switch self {
         case .whisper: return TRANSCRIBE_EXT_KIND_WHISPER_RUN
+        case .sortformer: return TRANSCRIBE_EXT_KIND_SORTFORMER_STREAM
         }
     }
 }
@@ -79,6 +109,11 @@ func withRunExtension<R>(
             c.initial_prompt = prompt
             return try withUnsafePointer(to: &c.ext) { try body($0) }
         }
+    case .sortformer(let o):
+        var c = transcribe_sortformer_stream_ext()
+        transcribe_sortformer_stream_ext_init(&c)
+        if let v = o.preset { c.preset = v.cValue }
+        return try withUnsafePointer(to: &c.ext) { try body($0) }
     }
 }