Skip to content

Local UI to run and train LLMs and diffusion models. Supports GGUF, MLX, Qwen3.8, DeepSeek-V4.1, MiniMax-H3, Gemma 4, FLUX and more.

Studio: clone with 13 more audio.cpp models, convert with Tone-Color VC (#12822)

Etherlcdb5d55c33c8
diff --git a/studio/backend/core/inference/audio_cpp_models.py b/studio/backend/core/inference/audio_cpp_models.py
index e7dc37b0c9c..020cfbfee8b 100644
--- a/studio/backend/core/inference/audio_cpp_models.py
+++ b/studio/backend/core/inference/audio_cpp_models.py
@@ -569,6 +569,15 @@ _CLONE_FAMILIES: tuple[AudioCppFamily, ...] = (
),
AudioCppFamily("voxcpm2", "tts", clone = CloneSpec("optional")),
AudioCppFamily("fish_audio", "tts", clone = CloneSpec("required")),
+ AudioCppFamily("breeze_tts", "tts", clone = CloneSpec("required")),
+ AudioCppFamily("omnivoice", "tts", clone = CloneSpec("required")),
+ AudioCppFamily("voxcpm1", "tts", clone = CloneSpec("required")),
+ AudioCppFamily("dots_tts", "tts", clone = CloneSpec("optional")),
+ AudioCppFamily("higgs_audio_tts", "tts", clone = CloneSpec("optional")),
+ # MOSS ignores a transcript; Irodori refuses one ("unknown Irodori-TTS request option").
+ AudioCppFamily("moss_tts_local", "tts", clone = CloneSpec("unused")),
+ AudioCppFamily("moss_tts_nano", "tts", clone = CloneSpec("unused")),
+ AudioCppFamily("irodori_tts", "tts", clone = CloneSpec("unused")),
AudioCppFamily("echo_tts", "tts", server_task = "clon", speaks = False, clone = CloneSpec("unused")),
AudioCppFamily(
"confucius4_tts", "tts", server_task = "clon", speaks = False, clone = CloneSpec("unused")
@@ -677,6 +686,13 @@ _CONVERT_FAMILIES: tuple[AudioCppFamily, ...] = (
speaks = False,
convert = ConvertSpec(source_rate = 16000, target_rate = 16000),
),
+ AudioCppFamily(
+ "tone_color_vc",
+ "tts",
+ server_task = "vc",
+ speaks = False,
+ convert = ConvertSpec(source_rate = 22050, target_rate = 22050),
+ ),
)
@@ -695,7 +711,9 @@ _FAMILY_LIST: tuple[AudioCppFamily, ...] = (
AudioCppFamily("kitten_tts", "tts", needs_espeak = True),
AudioCppFamily("piper_tts", "tts", needs_espeak = True),
AudioCppFamily("inflect_v2", "tts", needs_espeak = True),
- AudioCppFamily("pocket_tts", "tts", request_defaults = {"voice": "alba"}),
+ AudioCppFamily(
+ "pocket_tts", "tts", request_defaults = {"voice": "alba"}, clone = CloneSpec("unused")
+ ),
AudioCppFamily("supertonic", "tts", request_defaults = {"voice": "F1"}),
AudioCppFamily("qwen3_tts", "tts"),
*_CLONE_FAMILIES,
@@ -703,20 +721,13 @@ _FAMILY_LIST: tuple[AudioCppFamily, ...] = (
*(
AudioCppFamily(name, "tts")
for name in (
- "moss_tts_nano",
- "moss_tts_local",
"moss_tts_v15",
"moss_ttsd",
"chatterbox_turbo",
- "voxcpm1",
"neutts",
"magpie_tts",
- "higgs_audio_tts",
- "irodori_tts",
- "breeze_tts",
- "dots_tts",
+ # Its spec lists clone, but from a short reference clip it may not keep the voice.
"dramabox",
- "omnivoice",
"vibevoice",
"glm_tts",
"outetts",
diff --git a/studio/backend/tests/test_audio_cpp_clone_requests.py b/studio/backend/tests/test_audio_cpp_clone_requests.py
index 2a57bba500c..e8af6d05c95 100644
--- a/studio/backend/tests/test_audio_cpp_clone_requests.py
+++ b/studio/backend/tests/test_audio_cpp_clone_requests.py
@@ -69,6 +69,9 @@ MODELS = {
"f5": lambda: _model("f5_tts", "F5-TTS-GGUF"),
"kokoro": lambda: _model("kokoro_tts", "Kokoro-82M-GGUF"),
"vibevoice": lambda: _model("vibevoice", "VibeVoice-1.5B-GGUF"),
+ "omnivoice": lambda: _model("omnivoice", "OmniVoice-GGUF"),
+ "dots": lambda: _model("dots_tts", "DotTTS-SOAR-GGUF"),
+ "irodori": lambda: _model("irodori_tts", "Irodori-TTS-v4-Small-GGUF"),
}
@@ -141,6 +144,11 @@ SPEECH, TASKS = "/v1/audio/speech", "/v1/tasks/run"
{"reference_text": T}),
("f5", {"reference_text": T, "speed": 1.3}, {"speed": 1.3, "reference_text": T, "options": ABSENT}),
("f5", {"reference_text": T, "speed": 9}, {"speed": 2.0}),
+ ("omnivoice", {"reference_text": T}, {**BASE, "reference_text": T}),
+ ("dots", {"reference_text": T}, {**BASE, "reference_text": T}),
+ ("dots", {}, BASE),
+ # Irodori refuses a transcript as an unknown option.
+ ("irodori", {"reference_text": T}, BASE),
])
# fmt: on
def test_clone_request_body(key, kwargs, expect):
@@ -185,7 +193,12 @@ def test_status_fields_name_the_clone_workflow_and_transcript_rule():
assert (f["audio_workflows"], f["audio_reference_text"]) == (["clone"], "required")
assert f["audio_required_inputs"] == []
assert f["audio_clone"]["reference_text_waived"] == [["x_vector_only_mode", ["true"]]]
- for key, workflows, rt in (("vox", ["speak", "clone"], "optional"), ("kokoro", ["speak"], None)):
+ for key, workflows, rt in (
+ ("vox", ["speak", "clone"], "optional"),
+ ("omnivoice", ["speak", "clone"], "required"),
+ ("irodori", ["speak", "clone"], "unused"),
+ ("kokoro", ["speak"], None),
+ ):
f = acb.model_info_fields(MODELS[key]())
assert (f["audio_workflows"], f["audio_reference_text"]) == (workflows, rt)
cosy = acb.model_info_fields(MODELS["cosy"]())["audio_clone"]
diff --git a/studio/backend/tests/test_audio_cpp_convert_requests.py b/studio/backend/tests/test_audio_cpp_convert_requests.py
index eed20b96742..4bbf20cc277 100644
--- a/studio/backend/tests/test_audio_cpp_convert_requests.py
+++ b/studio/backend/tests/test_audio_cpp_convert_requests.py
@@ -199,9 +199,10 @@ def test_a_target_the_family_cannot_take_is_refused(family, target, match):
acc.convert_request(_model(family), mode = "speech", source = SOURCE, target = target)
-def test_meanvc2_sends_audio_voice_ref_and_seed():
[email protected]("family", ["meanvc2", "tone_color_vc"])
+def test_a_reference_family_sends_audio_voice_ref_and_seed(family):
request = acc.convert_request(
- _model("meanvc2"), mode = "speech", source = SOURCE, target = TARGET, pitch = 4, seed = 7
+ _model(family), mode = "speech", source = SOURCE, target = TARGET, pitch = 4, seed = 7
)
assert request == {"audio": SOURCE, "voice_ref": TARGET, "seed": "7"}
@@ -412,6 +413,9 @@ def test_status_fields_describe_convert_per_family():
assert chatterbox["audio_convert"]["pitch"] == {}
assert chatterbox["audio_convert_rules"] == {"source_rate": 16000, "target_rate": 24000}
assert acb.model_info_fields(_model("meanvc2"))["audio_convert"]["pitch"] == {}
+ tone_color = acb.model_info_fields(_model("tone_color_vc"))
+ assert tone_color["audio_workflow_tasks"] == {"convert": "vc"}
+ assert tone_color["audio_convert_rules"] == {"source_rate": 22050, "target_rate": 22050}
kokoro = acb.model_info_fields(_model("kokoro_tts"))
assert kokoro["audio_convert"] is None and kokoro["audio_options_by_workflow"] is None
diff --git a/studio/backend/tests/test_audio_cpp_models.py b/studio/backend/tests/test_audio_cpp_models.py
index 422d4de10ec..8fe236e3545 100644
--- a/studio/backend/tests/test_audio_cpp_models.py
+++ b/studio/backend/tests/test_audio_cpp_models.py
@@ -524,6 +524,15 @@ _CLONE_TABLE = {
"firered_audio": ("tts", False, "optional"),
"vevo2": ("tts", False, "unused"),
"miotts": ("tts", False, "unused"),
+ "breeze_tts": ("tts", True, "required"),
+ "omnivoice": ("tts", True, "required"),
+ "voxcpm1": ("tts", True, "required"),
+ "dots_tts": ("tts", True, "optional"),
+ "higgs_audio_tts": ("tts", True, "optional"),
+ "moss_tts_local": ("tts", True, "unused"),
+ "moss_tts_nano": ("tts", True, "unused"),
+ "irodori_tts": ("tts", True, "unused"),
+ "pocket_tts": ("tts", True, "unused"),
}
@@ -534,6 +543,7 @@ _CONVERT_TABLE = {
"meanvc2": ("vc", {"speech": "vc"}, "audio", 16000),
"chatterbox": ("clon", {"speech": "vc"}, "audio", 16000),
"vevo2": ("tts", {"speech": "vc", "singing": "svc"}, "audio", 24000),
+ "tone_color_vc": ("vc", {"speech": "vc"}, "audio", 22050),
}
@@ -613,10 +623,13 @@ def test_edit_families_bind_the_edit_workflow():
edit_names = ["DotTTS-Edit-GGUF", "dots-tts-edit-q8_0.gguf"]
dots_edit = acm.family_policy("dots_tts", names = edit_names)
binding = dots_edit.workflows["edit"]
- assert list(dots_edit.workflows) == ["speak", "edit"]
+ assert list(dots_edit.workflows) == ["speak", "clone", "edit"]
assert (binding.server_task, binding.endpoint) == ("tts", "tasks")
for names in (["DotTTS-MF-GGUF"], ["DotTTS-SOAR-GGUF"], []):
- assert list(acm.family_policy("dots_tts", names = names).workflows) == ["speak"], names
+ assert list(acm.family_policy("dots_tts", names = names).workflows) == [
+ "speak",
+ "clone",
+ ], names
vevo2, firered = acm.FAMILIES["vevo2"], acm.FAMILIES["firered_audio"]
assert list(firered.workflows) == ["clone", "edit"]
# Vevo2 also converts a voice.
diff --git a/studio/backend/tests/test_audio_cpp_umbrella_catalog.py b/studio/backend/tests/test_audio_cpp_umbrella_catalog.py
index 7e5595acb69..bc596d3f53f 100644
--- a/studio/backend/tests/test_audio_cpp_umbrella_catalog.py
+++ b/studio/backend/tests/test_audio_cpp_umbrella_catalog.py
@@ -20,7 +20,7 @@ UMBRELLA = {
"Apollo-GGUF": ("apollo", []),
"AudioSR-GGUF": ("audiosr", []),
"BS-RoFormer-ep368-GGUF": ("bs_roformer", ["separate"]),
- "Breeze-TTS-2-GGUF": ("breeze_tts", ["speak"]),
+ "Breeze-TTS-2-GGUF": ("breeze_tts", ["speak", "clone"]),
"Canary-180M-Flash-GGUF": ("canary_asr", ["transcribe"]),
"Chatterbox-GGUF": ("chatterbox", ["clone", "convert"]),
"Chatterbox-Turbo-GGUF": ("chatterbox_turbo", ["speak"]),
@@ -30,9 +30,9 @@ UMBRELLA = {
"ControlFoley-GGUF": ("controlfoley", ["music"]),
"CosyVoice3-GGUF": ("cosyvoice3", ["clone"]),
"CrisperWhisper2.0-GGUF": ("crisperwhisper", []),
- "DotTTS-Edit-GGUF": ("dots_tts", ["speak", "edit"]),
- "DotTTS-MF-GGUF": ("dots_tts", ["speak"]),
- "DotTTS-SOAR-GGUF": ("dots_tts", ["speak"]),
+ "DotTTS-Edit-GGUF": ("dots_tts", ["speak", "clone", "edit"]),
+ "DotTTS-MF-GGUF": ("dots_tts", ["speak", "clone"]),
+ "DotTTS-SOAR-GGUF": ("dots_tts", ["speak", "clone"]),
"DramaBox-GGUF": ("dramabox", ["speak"]),
"FireRedAudio-GGUF": ("firered_audio", ["clone", "edit"]),
"FireRedTTS3-Base-GGUF": ("fireredtts3", ["clone"]),
@@ -45,21 +45,21 @@ UMBRELLA = {
"HTDemucs-GGUF": ("htdemucs", ["separate"]),
"HeartMuLa-GGUF": ("heartmula", ["music"]),
"Higgs-Audio-v3-STT-GGUF": ("higgs_audio_stt", ["transcribe"]),
- "Higgs-Audio-v3-TTS-4B-GGUF": ("higgs_audio_tts", ["speak"]),
+ "Higgs-Audio-v3-TTS-4B-GGUF": ("higgs_audio_tts", ["speak", "clone"]),
"Hviske-v5.3-GGUF": ("hviske_asr", ["transcribe"]),
"IndexTTS2-GGUF": ("index_tts2", ["clone"]),
"IndexTTS2.5-GGUF": ("index_tts2", ["clone"]),
"Inflect-Micro-v2-GGUF": ("inflect_v2", ["speak"]),
- "Irodori-TTS-500M-v3-GGUF": ("irodori_tts", ["speak"]),
- "Irodori-TTS-600M-v3-VoiceDesign-GGUF": ("irodori_tts", ["speak"]),
- "Irodori-TTS-v4-Small-GGUF": ("irodori_tts", ["speak"]),
+ "Irodori-TTS-500M-v3-GGUF": ("irodori_tts", ["speak", "clone"]),
+ "Irodori-TTS-600M-v3-VoiceDesign-GGUF": ("irodori_tts", ["speak", "clone"]),
+ "Irodori-TTS-v4-Small-GGUF": ("irodori_tts", ["speak", "clone"]),
"KittenTTS-GGUF": ("kitten_tts", ["speak"]),
"Kokoro-82M-GGUF": ("kokoro_tts", ["speak"]),
"Kroko-ASR-GGUF": ("kroko_asr", ["transcribe"]),
"KugelAudio-0-Open-GGUF": ("kugelaudio", []),
"MMS-Forced-Aligner-GGUF": ("mms_forced_aligner", []),
- "MOSS-TTS-Local-v1.5-GGUF": ("moss_tts_local", ["speak"]),
- "MOSS-TTS-Nano-100M-GGUF": ("moss_tts_nano", ["speak"]),
+ "MOSS-TTS-Local-v1.5-GGUF": ("moss_tts_local", ["speak", "clone"]),
+ "MOSS-TTS-Nano-100M-GGUF": ("moss_tts_nano", ["speak", "clone"]),
"MOSS-Transcribe-Diarize-GGUF": ("moss_transcribe_diarize", ["transcribe"]),
"MOSS-VoiceGenerator-GGUF": ("moss_voicegen", ["speak"]),
"MagpieTTS-Multilingual-357M-GGUF": ("magpie_tts", ["speak"]),
@@ -77,11 +77,11 @@ UMBRELLA = {
"Niagara-ASR-GGUF": ("niagara_asr", ["transcribe"]),
"OWSM-CTC-GGUF": ("owsm_ctc", []),
"OWSM-GGUF": ("owsm", []),
- "OmniVoice-GGUF": ("omnivoice", ["speak"]),
+ "OmniVoice-GGUF": ("omnivoice", ["speak", "clone"]),
"Parakeet-TDT-0.6B-v3-GGUF": ("parakeet_tdt", ["transcribe"]),
"PersonaPlex-GGUF": ("personaplex", []),
"Piper-TTS-GGUF": ("piper_tts", ["speak"]),
- "PocketTTS-GGUF": ("pocket_tts", ["speak"]),
+ "PocketTTS-GGUF": ("pocket_tts", ["speak", "clone"]),
"PulseVAD-GGUF": ("pulsevad", []),
"Qwen3-ASR-0.6B-GGUF": ("qwen3_asr", ["transcribe"]),
"Qwen3-ASR-1.7B-GGUF": ("qwen3_asr", ["transcribe"]),
@@ -100,12 +100,12 @@ UMBRELLA = {
"Stable-Audio-3-Small-Music-GGUF": ("stable_audio", ["music"]),
"Stable-Audio-3-Small-SFX-GGUF": ("stable_audio", ["music"]),
"Supertonic-3-GGUF": ("supertonic", ["speak"]),
- "Tone-Color-VC-GGUF": ("tone_color_vc", []),
+ "Tone-Color-VC-GGUF": ("tone_color_vc", ["convert"]),
"UniverSR-GGUF": ("universr", []),
"Vevo2-GGUF": ("vevo2", ["clone", "edit", "convert"]),
"VibeVoice-1.5B-GGUF": ("vibevoice", ["speak"]),
"VibeVoice-ASR-GGUF": ("vibevoice_asr", ["transcribe"]),
- "VoxCPM1-GGUF": ("voxcpm1", ["speak"]),
+ "VoxCPM1-GGUF": ("voxcpm1", ["speak", "clone"]),
"VoxCPM2-GGUF": ("voxcpm2", ["speak", "clone"]),
"Voxtral-Mini-4B-Realtime-2602-GGUF": ("voxtral_realtime", ["transcribe"]),
}
@@ -114,7 +114,6 @@ UMBRELLA = {
SPEC_TASKS = {
"gigaam_asr": ["asr"],
"maya1": ["tts"],
- "tone_color_vc": ["vc"],
"crisperwhisper": ["asr", "align"],
"kugelaudio": ["tts"],
"owsm": ["asr"],
diff --git a/studio/backend/tests/test_audio_edit_requests.py b/studio/backend/tests/test_audio_edit_requests.py
index 28a0b30a356..c993999c6e6 100644
--- a/studio/backend/tests/test_audio_edit_requests.py
+++ b/studio/backend/tests/test_audio_edit_requests.py
@@ -243,7 +243,11 @@ def test_status_fields_carry_the_edit_rules():
fields = acb.model_info_fields(_model(family))
assert fields["audio_edit"] == _RULES.get(family), family
assert ("edit" in fields["audio_workflows"]) == (family in _RULES), family
- assert acb.model_info_fields(_model("dots_tts"))["audio_workflows"] == ["speak", "edit"]
+ assert acb.model_info_fields(_model("dots_tts"))["audio_workflows"] == [
+ "speak",
+ "clone",
+ "edit",
+ ]
assert acb.model_info_fields(_model("vevo2"))["audio_workflows"] == ["clone", "edit"]
diff --git a/studio/backend/tests/test_audio_workflows.py b/studio/backend/tests/test_audio_workflows.py
index 55acc35c7ec..1a972b9a9c6 100644
--- a/studio/backend/tests/test_audio_workflows.py
+++ b/studio/backend/tests/test_audio_workflows.py
@@ -172,8 +172,8 @@ def test_status_and_load_carry_the_convert_fields():
("RVC-GGUF", "rvc", ["convert"]),
("SeedVC-MLX-GGUF", "seed_vc", ["convert"]),
("MeanVC2-GGUF", "meanvc2", ["convert"]),
- ("DotTTS-Edit-GGUF", "dots_tts", ["speak", "edit"]),
- ("DotTTS-MF-GGUF", "dots_tts", ["speak"]),
+ ("DotTTS-Edit-GGUF", "dots_tts", ["speak", "clone", "edit"]),
+ ("DotTTS-MF-GGUF", "dots_tts", ["speak", "clone"]),
("Vevo2-GGUF", "vevo2", ["clone", "edit", "convert"]),
("FireRedAudio-GGUF", "firered_audio", ["clone", "edit"]),
],
diff --git a/studio/frontend/src/features/audio/audio-cpp-catalog.ts b/studio/frontend/src/features/audio/audio-cpp-catalog.ts
index f274764f047..833c3a65e90 100644
--- a/studio/frontend/src/features/audio/audio-cpp-catalog.ts
+++ b/studio/frontend/src/features/audio/audio-cpp-catalog.ts
@@ -62,8 +62,8 @@ export const AUDIO_CPP_MODELS: readonly AudioCppModel[] = [
{ id: folder("KittenTTS-GGUF"), task: "tts", needsEspeak: true },
{ id: folder("Piper-TTS-GGUF"), task: "tts", needsEspeak: true },
{ id: folder("Inflect-Micro-v2-GGUF"), task: "tts", needsEspeak: true },
- { id: folder("PocketTTS-GGUF"), task: "tts" },
- { id: folder("MOSS-TTS-Nano-100M-GGUF"), task: "tts" },
+ { id: folder("PocketTTS-GGUF"), task: "tts", workflows: ["speak", "clone"] },
+ { id: folder("MOSS-TTS-Nano-100M-GGUF"), task: "tts", workflows: ["speak", "clone"] },
{ id: folder("Supertonic-3-GGUF"), task: "tts" },
{ id: folder("Chatterbox-Turbo-GGUF"), task: "tts" },
{ id: folder("VoxCPM2-GGUF"), task: "tts", workflows: ["speak", "clone"] },
@@ -73,28 +73,30 @@ export const AUDIO_CPP_MODELS: readonly AudioCppModel[] = [
{ id: folder("CosyVoice3-GGUF"), task: "tts", workflows: ["clone"] },
{ id: folder("Qwen3-TTS-12Hz-1.7B-CustomVoice-GGUF"), task: "tts" },
{ id: folder("Qwen3-TTS-12Hz-1.7B-VoiceDesign-GGUF"), task: "tts" },
- { id: folder("DotTTS-Edit-GGUF"), task: "tts", workflows: ["speak", "edit"] },
+ { id: folder("DotTTS-Edit-GGUF"), task: "tts", workflows: ["speak", "clone", "edit"] },
{ id: folder("Vevo2-GGUF"), task: "tts", workflows: ["clone", "edit", "convert"] },
{ id: folder("FireRedAudio-GGUF"), task: "tts", workflows: ["clone", "edit"] },
{ id: folder("SeedVC-MLX-GGUF"), task: "tts", workflows: ["convert"] },
{ id: folder("RVC-GGUF"), task: "tts", workflows: ["convert"] },
{ id: folder("MeanVC2-GGUF"), task: "tts", workflows: ["convert"] },
- { id: folder("Breeze-TTS-2-GGUF"), task: "tts" },
- { id: folder("DotTTS-MF-GGUF"), task: "tts" },
- { id: folder("DotTTS-SOAR-GGUF"), task: "tts" },
+ { id: folder("Tone-Color-VC-GGUF"), task: "tts", workflows: ["convert"] },
+ { id: folder("Breeze-TTS-2-GGUF"), task: "tts", workflows: ["speak", "clone"] },
+ { id: folder("DotTTS-MF-GGUF"), task: "tts", workflows: ["speak", "clone"] },
+ { id: folder("DotTTS-SOAR-GGUF"), task: "tts", workflows: ["speak", "clone"] },
+ // Speak only: its spec lists clone, but from a short reference clip it may not keep the voice.
{ id: folder("DramaBox-GGUF"), task: "tts" },
- { id: folder("Higgs-Audio-v3-TTS-4B-GGUF"), task: "tts" },
- { id: folder("Irodori-TTS-500M-v3-GGUF"), task: "tts" },
- { id: folder("Irodori-TTS-600M-v3-VoiceDesign-GGUF"), task: "tts" },
- { id: folder("Irodori-TTS-v4-Small-GGUF"), task: "tts" },
- { id: folder("MOSS-TTS-Local-v1.5-GGUF"), task: "tts" },
+ { id: folder("Higgs-Audio-v3-TTS-4B-GGUF"), task: "tts", workflows: ["speak", "clone"] },
+ { id: folder("Irodori-TTS-500M-v3-GGUF"), task: "tts", workflows: ["speak", "clone"] },
+ { id: folder("Irodori-TTS-600M-v3-VoiceDesign-GGUF"), task: "tts", workflows: ["speak", "clone"] },
+ { id: folder("Irodori-TTS-v4-Small-GGUF"), task: "tts", workflows: ["speak", "clone"] },
+ { id: folder("MOSS-TTS-Local-v1.5-GGUF"), task: "tts", workflows: ["speak", "clone"] },
{ id: folder("MOSS-VoiceGenerator-GGUF"), task: "tts" },
{ id: folder("MagpieTTS-Multilingual-357M-GGUF"), task: "tts" },
{ id: folder("Maya1-GGUF"), task: "tts" },
{ id: folder("NeuTTS-2E-GGUF"), task: "tts" },
- { id: folder("OmniVoice-GGUF"), task: "tts" },
+ { id: folder("OmniVoice-GGUF"), task: "tts", workflows: ["speak", "clone"] },
{ id: folder("VibeVoice-1.5B-GGUF"), task: "tts" },
- { id: folder("VoxCPM1-GGUF"), task: "tts" },
+ { id: folder("VoxCPM1-GGUF"), task: "tts", workflows: ["speak", "clone"] },
{ id: folder("Fish-Audio-S2-Pro-GGUF"), task: "tts", workflows: ["speak", "clone"] },
{ id: folder("Confucius4-TTS-GGUF"), task: "tts", workflows: ["clone"] },
{ id: folder("FireRedTTS3-Base-GGUF"), task: "tts", workflows: ["clone"] },
@@ -175,7 +177,6 @@ export const AUDIO_CPP_UNOFFERED_FOLDERS: Readonly<Record<string, string>> = {
"Sortformer-Diar-4spk-v1-GGUF": "speaker diarization has no page",
"PulseVAD-GGUF": "voice activity detection has no page",
"MuScriptor-Small-GGUF": "music transcription has no page",
- "Tone-Color-VC-GGUF": "Studio has no conversion request for it yet",
"Samsone-GGUF": "describes audio rather than transcribing it",
};
@@ -196,9 +197,9 @@ export function isSpeakAndCloneFamilyId(id: string | null | undefined): boolean
}
// Families that clone and convert (chatterbox, vevo2) and the convert-only ones (rvc, seed_vc,
-// meanvc2), by repo name the way the backend's family_from_names reads it.
+// meanvc2, tone_color_vc), by repo name the way the backend's family_from_names reads it.
const CLONE_AND_CONVERT_FAMILY_HINT = /vevo[-_]?2|chatterbox(?![-_]?turbo)/i;
-const CONVERT_ONLY_FAMILY_HINT = /rvc|seed[-_]?vc|meanvc[-_]?2/i;
+const CONVERT_ONLY_FAMILY_HINT = /rvc|seed[-_]?vc|meanvc[-_]?2|tone[-_]?color/i;
export function isCloneAndConvertFamilyId(id: string | null | undefined): boolean {
return CLONE_AND_CONVERT_FAMILY_HINT.test(id ?? "");
diff --git a/studio/frontend/tests/audio-cpp-catalog.test.ts b/studio/frontend/tests/audio-cpp-catalog.test.ts
index 330d97f0913..ac552ef858d 100644
--- a/studio/frontend/tests/audio-cpp-catalog.test.ts
+++ b/studio/frontend/tests/audio-cpp-catalog.test.ts
@@ -135,7 +135,7 @@ test("recommended ids are unique and name their Hub repo or package folder", ()
});
test("folders of the shared repo the pickers leave out are never seeded and say why", () => {
- assert.equal(Object.keys(AUDIO_CPP_UNOFFERED_FOLDERS).length, 19);
+ assert.equal(Object.keys(AUDIO_CPP_UNOFFERED_FOLDERS).length, 18);
for (const [folder, reason] of Object.entries(AUDIO_CPP_UNOFFERED_FOLDERS)) {
assert.equal(audioCppModelFor(`${AUDIO_CPP_REPO}/${folder}`), null, folder);
assert.match(folder, /-GGUF$/);
@@ -188,6 +188,7 @@ test("voice conversion models are seeded with the pages they run on", () => {
["RVC-GGUF", ["convert"], "Voice conversion"],
["SeedVC-MLX-GGUF", ["convert"], "Voice conversion"],
["MeanVC2-GGUF", ["convert"], "Voice conversion"],
+ ["Tone-Color-VC-GGUF", ["convert"], "Voice conversion"],
["Chatterbox-GGUF", ["clone", "convert"], "Voice cloning and conversion"],
["Vevo2-GGUF", ["clone", "edit", "convert"], "Voice cloning and conversion"],
["IndexTTS2-GGUF", ["clone"], "Voice cloning"],
diff --git a/studio/frontend/tests/audio-picker-filter.test.ts b/studio/frontend/tests/audio-picker-filter.test.ts
index 6ed3cc62e1b..065c7331d45 100644
--- a/studio/frontend/tests/audio-picker-filter.test.ts
+++ b/studio/frontend/tests/audio-picker-filter.test.ts
@@ -90,6 +90,7 @@ test("voice conversion models list on Convert only; Chatterbox and Vevo2 on Clon
["RVC-GGUF", ["convert"]],
["SeedVC-MLX-GGUF", ["convert"]],
["MeanVC2-GGUF", ["convert"]],
+ ["Tone-Color-VC-GGUF", ["convert"]],
["Chatterbox-GGUF", ["clone", "convert"]],
["Vevo2-GGUF", ["clone", "edit", "convert"]],
] as const) {
@@ -291,7 +292,12 @@ test("the clone-only name hint follows the backend's speaks=False families", asy
});
test("Hub search rows for conversion families outside the catalog list on Convert", () => {
- for (const id of ["someone/RVC-v2-GGUF", "x/Seed-VC-GGUF", "x/MeanVC2-GGUF"]) {
+ for (const id of [
+ "someone/RVC-v2-GGUF",
+ "x/Seed-VC-GGUF",
+ "x/MeanVC2-GGUF",
+ "x/Tone-Color-VC-GGUF",
+ ]) {
const row = { id, task: "text-to-speech" };
assert.equal(audioRowMatchesWorkflow(row, "convert"), true, id);
assert.equal(audioRowMatchesWorkflow(row, "clone"), false, id);
@@ -340,6 +346,10 @@ test("every seeded audio GGUF lists on exactly the pages it runs on", () => {
}
assert.deepEqual(listed("Fish-Audio-S2-Pro-GGUF"), ["speak", "clone"]);
assert.deepEqual(listed("MioTTS-1.7B-GGUF"), ["clone"]);
+ assert.deepEqual(listed("OmniVoice-GGUF"), ["speak", "clone"]);
+ assert.deepEqual(listed("DotTTS-Edit-GGUF"), ["speak", "clone", "edit"]);
+ assert.deepEqual(listed("DramaBox-GGUF"), ["speak"]);
+ assert.deepEqual(listed("Tone-Color-VC-GGUF"), ["convert"]);
assert.deepEqual(listed("FireRedTTS3-Instruct-GGUF"), ["clone"]);
assert.deepEqual(listed("MOSS-VoiceGenerator-GGUF"), ["speak"]);
assert.deepEqual(listed("HeartMuLa-GGUF"), ["music"]);