From 8448a214def68db4d231890b630d747e1f2ada58 Mon Sep 17 00:00:00 2001
From: Gabriel de Marmiesse <gabriel@kyutai.org>
Date: Tue, 14 Apr 2026 16:12:55 +0200
Subject: [PATCH 1/4] Make `max_speakers` configurable within the json config

---
 moshi/moshi/models/loaders.py | 2 ++
 moshi/moshi/models/tts.py     | 7 ++++---
 2 files changed, 6 insertions(+), 3 deletions(-)

diff --git a/moshi/moshi/models/loaders.py b/moshi/moshi/models/loaders.py
index 6d494b2cb..21122d2e8 100644
--- a/moshi/moshi/models/loaders.py
+++ b/moshi/moshi/models/loaders.py
@@ -377,6 +377,8 @@ def get_moshi_lm(
     lm_kwargs = dict(lm_kwargs)
     assert lm_kwargs is not None
 
+    lm_kwargs.pop("max_speakers", None)
+
     if "conditioners" in lm_kwargs:
         lm_kwargs["condition_provider"] = get_conditioner_provider(
             lm_kwargs["dim"], device, lm_kwargs
diff --git a/moshi/moshi/models/tts.py b/moshi/moshi/models/tts.py
index da5b4c730..b1ad3e895 100644
--- a/moshi/moshi/models/tts.py
+++ b/moshi/moshi/models/tts.py
@@ -31,7 +31,7 @@
 DEFAULT_DSM_TTS_REPO = 'kyutai/tts-1.6b-en_fr'
 DEFAULT_DSM_TTS_VOICE_REPO = 'kyutai/tts-voices'
 PUBLIC_DATA_DSM_TTS_REPO = 'kyutai/tts-0.75b-en-public'
-
+DEFAULT_MAX_SPEAKERS = 5
 
 @dataclass
 class TokenIds:
@@ -377,7 +377,7 @@ class TTSModel:
     # The delay of the audio streams compared to the text stream.
     # Note that in addition, the acoustic tokens are delayed wrt the semantic ones.
     delay_steps: int
-    max_speakers: int = 5
+    max_speakers: int = DEFAULT_MAX_SPEAKERS
     multistream: bool = False
 
     # The following params can be overriden to customize generation.
@@ -415,6 +415,7 @@ def from_checkpoint_info(checkpoint_info: loaders.CheckpointInfo,
             lm=lm, mimi=mimi, tokenizer=tokenizer,
             voice_suffix=voice_suffix, voice_repo=voice_repo,
             machine=machine, delay_steps=delay_steps, multistream=multistream,
+            max_speakers=checkpoint_info.raw_config.get('max_speakers', DEFAULT_MAX_SPEAKERS),
             **kwargs)
         mimi_n_q = tts_model.n_q
 
@@ -646,7 +647,7 @@ def make_condition_attributes(
         if voices:
             voice_tensor = None
             mask = None
-            for idx in range(5):
+            for idx in range(self.max_speakers):
                 if idx < len(voices):
                     if voices[idx].suffix != ".safetensors":
                         raise ValueError(f"Voice file must be a .safetensors file, got {voices[idx]}")

From 02d8a772f10e5474572d6d51ce2261b200b0e727 Mon Sep 17 00:00:00 2001
From: Gabriel de Marmiesse <gabriel@kyutai.org>
Date: Tue, 14 Apr 2026 17:31:31 +0200
Subject: [PATCH 2/4] IIII

---
 moshi/moshi/models/tts.py | 1 +
 1 file changed, 1 insertion(+)

diff --git a/moshi/moshi/models/tts.py b/moshi/moshi/models/tts.py
index b1ad3e895..8a050ad8d 100644
--- a/moshi/moshi/models/tts.py
+++ b/moshi/moshi/models/tts.py
@@ -33,6 +33,7 @@
 PUBLIC_DATA_DSM_TTS_REPO = 'kyutai/tts-0.75b-en-public'
 DEFAULT_MAX_SPEAKERS = 5
 
+
 @dataclass
 class TokenIds:
     """

From d4aec5753aa625bc5436e91c25490e553d5e6329 Mon Sep 17 00:00:00 2001
From: Gabriel de Marmiesse <gabriel@kyutai.org>
Date: Tue, 14 Apr 2026 18:00:25 +0200
Subject: [PATCH 3/4] Fix

---
 moshi/moshi/models/tts.py | 3 ++-
 1 file changed, 2 insertions(+), 1 deletion(-)

diff --git a/moshi/moshi/models/tts.py b/moshi/moshi/models/tts.py
index 8a050ad8d..62a129dd4 100644
--- a/moshi/moshi/models/tts.py
+++ b/moshi/moshi/models/tts.py
@@ -412,11 +412,12 @@ def from_checkpoint_info(checkpoint_info: loaders.CheckpointInfo,
         machine = StateMachine(
             token_ids=token_ids, second_stream_ahead=second_stream_ahead,
             max_padding=max_padding, initial_padding=initial_padding)
+        max_speakers = checkpoint_info.tts_config.get('max_speakers', DEFAULT_MAX_SPEAKERS)
         tts_model = TTSModel(
             lm=lm, mimi=mimi, tokenizer=tokenizer,
             voice_suffix=voice_suffix, voice_repo=voice_repo,
             machine=machine, delay_steps=delay_steps, multistream=multistream,
-            max_speakers=checkpoint_info.raw_config.get('max_speakers', DEFAULT_MAX_SPEAKERS),
+            max_speakers=max_speakers,
             **kwargs)
         mimi_n_q = tts_model.n_q
 

From 0f05f6cae353a84ef1521a7ced231b9bfb5e475c Mon Sep 17 00:00:00 2001
From: Gabriel de Marmiesse <gabriel@kyutai.org>
Date: Tue, 14 Apr 2026 18:01:37 +0200
Subject: [PATCH 4/4] III

---
 moshi/moshi/models/loaders.py | 2 --
 1 file changed, 2 deletions(-)

diff --git a/moshi/moshi/models/loaders.py b/moshi/moshi/models/loaders.py
index 21122d2e8..6d494b2cb 100644
--- a/moshi/moshi/models/loaders.py
+++ b/moshi/moshi/models/loaders.py
@@ -377,8 +377,6 @@ def get_moshi_lm(
     lm_kwargs = dict(lm_kwargs)
     assert lm_kwargs is not None
 
-    lm_kwargs.pop("max_speakers", None)
-
     if "conditioners" in lm_kwargs:
         lm_kwargs["condition_provider"] = get_conditioner_provider(
             lm_kwargs["dim"], device, lm_kwargs