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: classify the 10 new audio.cpp-gguf folders as needing a newer runtime (#13393)

Daniel Han20030550f8a8
diff --git a/studio/backend/tests/test_audio_cpp_umbrella_catalog.py b/studio/backend/tests/test_audio_cpp_umbrella_catalog.py
index dd5ab2ef6bb..49c7e918486 100644
--- a/studio/backend/tests/test_audio_cpp_umbrella_catalog.py
+++ b/studio/backend/tests/test_audio_cpp_umbrella_catalog.py
@@ -14,11 +14,12 @@ import pytest
from core.inference import audio_cpp_models as acm
from core.inference.audio_cpp_spec_families import AUDIO_CPP_SPEC_FAMILIES
-# Folder -> (family its GGUF header names, Audio pages), audio-cpp/audio.cpp-gguf at e36610ac.
+# Folder -> (family its GGUF header names, Audio pages), audio-cpp/audio.cpp-gguf at f861648c.
UMBRELLA = {
"ACE-Step1.5-GGUF": ("ace_step", ["music"]),
"Apollo-GGUF": ("apollo", []),
"AudioSR-GGUF": ("audiosr", []),
+ "AST-AudioSet-GGUF": ("ast_audioset", []),
"BS-RoFormer-ep368-GGUF": ("bs_roformer", ["separate"]),
"Breeze-TTS-2-GGUF": ("breeze_tts", ["speak", "clone"]),
"Canary-180M-Flash-GGUF": ("canary_asr", ["transcribe"]),
@@ -26,6 +27,7 @@ UMBRELLA = {
"Chatterbox-Turbo-GGUF": ("chatterbox_turbo", ["speak"]),
"Citrinet-ASR-GGUF": ("citrinet_asr", ["transcribe"]),
"Cohere-Transcribe-GGUF": ("cohere_asr", ["transcribe"]),
+ "CED-GGUF": ("ced", []),
"Confucius4-TTS-GGUF": ("confucius4_tts", ["clone"]),
"ControlFoley-GGUF": ("controlfoley", ["music"]),
"CosyVoice3-GGUF": ("cosyvoice3", ["clone"]),
@@ -37,6 +39,7 @@ UMBRELLA = {
"FireRedAudio-GGUF": ("firered_audio", ["clone", "edit"]),
"FireRedTTS3-Base-GGUF": ("fireredtts3", ["clone"]),
"FireRedTTS3-Instruct-GGUF": ("fireredtts3", ["clone"]),
+ "FireRedVAD-GGUF": ("firered_vad", []),
"Fish-Audio-S2-Pro-GGUF": ("fish_audio", ["speak", "clone"]),
"Fun-ASR-Nano-2512-GGUF": ("fun_asr_nano", ["transcribe"]),
"GigaAM-ASR-GGUF": ("gigaam_asr", ["transcribe"]),
@@ -47,6 +50,7 @@ UMBRELLA = {
"Higgs-Audio-v3-STT-GGUF": ("higgs_audio_stt", ["transcribe"]),
"Higgs-Audio-v3-TTS-4B-GGUF": ("higgs_audio_tts", ["speak", "clone"]),
"Hviske-v5.3-GGUF": ("hviske_asr", ["transcribe"]),
+ "Hviske-v6-GGUF": ("hviske_asr_v6", []),
"IndexTTS2-GGUF": ("index_tts2", ["clone"]),
"IndexTTS2.5-GGUF": ("index_tts2", ["clone"]),
"Inflect-Micro-v2-GGUF": ("inflect_v2", ["speak"]),
@@ -67,10 +71,12 @@ UMBRELLA = {
"MeanVC2-GGUF": ("meanvc2", ["convert"]),
"Mel-Band-RoFormer-GGUF": ("mel_band_roformer", ["separate"]),
"MiDashengLM-Gen-GGUF": ("midashenglm_gen", ["music"]),
+ "MicroWakeWord-GGUF": ("micro_wake_word", []),
"MiniMax-H3-Q4-GGUF": ("minimax_h3", []),
"MioCodec-25Hz-44.1kHz-v2-GGUF": ("miocodec", []),
"MioTTS-1.7B-GGUF": ("miotts", ["clone"]),
"Moonshine-Streaming-GGUF": ("moonshine_asr", ["transcribe"]),
+ "MossFormer2-GGUF": ("mossformer2", []),
"MuScriptor-Small-GGUF": ("muscriptor", []),
"Nemotron-3.5-ASR-Streaming-0.6B-GGUF": ("nemotron_asr", ["transcribe"]),
"NeuTTS-2E-GGUF": ("neutts", ["speak"]),
@@ -80,6 +86,7 @@ UMBRELLA = {
"OmniVoice-GGUF": ("omnivoice", ["speak", "clone"]),
"Parakeet-TDT-0.6B-v3-GGUF": ("parakeet_tdt", ["transcribe"]),
"PersonaPlex-GGUF": ("personaplex", []),
+ "Phonon-2-GGUF": ("parakeet_tdt", []),
"Piper-TTS-GGUF": ("piper_tts", ["speak"]),
"PocketTTS-GGUF": ("pocket_tts", ["speak", "clone"]),
"PulseVAD-GGUF": ("pulsevad", []),
@@ -93,6 +100,7 @@ UMBRELLA = {
"RVC-GGUF": ("rvc", ["convert"]),
"Samsone-GGUF": ("samsone", []),
"SeedVC-MLX-GGUF": ("seed_vc", ["convert"]),
+ "Sherpa-KWS-GGUF": ("sherpa_kws", []),
"Sidon-GGUF": ("sidon", []),
"Smart-Turn-v3-GGUF": ("smart_turn", []),
"Sortformer-Diar-4spk-v1-GGUF": ("sortformer_diar", []),
@@ -100,6 +108,7 @@ UMBRELLA = {
"Stable-Audio-3-Small-Music-GGUF": ("stable_audio", ["music"]),
"Stable-Audio-3-Small-SFX-GGUF": ("stable_audio", ["music"]),
"Supertonic-3-GGUF": ("supertonic", ["speak"]),
+ "TF-GridNet-GGUF": ("tf_gridnet", []),
"Tone-Color-VC-GGUF": ("tone_color_vc", ["convert"]),
"UniverSR-GGUF": ("universr", []),
"Vevo2-GGUF": ("vevo2", ["clone", "edit", "convert"]),
@@ -107,6 +116,7 @@ UMBRELLA = {
"VibeVoice-ASR-GGUF": ("vibevoice_asr", ["transcribe"]),
"VoxCPM1-GGUF": ("voxcpm1", ["speak", "clone"]),
"VoxCPM2-GGUF": ("voxcpm2", ["speak", "clone"]),
+ "Voxtral-4B-TTS-2603-GGUF": ("voxtral_tts", []),
"Voxtral-Mini-4B-Realtime-2602-GGUF": ("voxtral_realtime", ["transcribe"]),
}
@@ -134,7 +144,12 @@ def pinned_runtime(monkeypatch):
)
[email protected]("folder", sorted(UMBRELLA))
+# Variants of a family the pinned runtime knows that only a newer runtime loads: the pickers leave
+# them out, but a load by id still resolves through the family, so they skip the per-folder check.
+NEWER_RUNTIME_VARIANTS = {"Phonon-2-GGUF"}
+
+
[email protected]("folder", sorted(set(UMBRELLA) - NEWER_RUNTIME_VARIANTS))
def test_each_folder_lands_on_its_pages(folder):
family, pages = UMBRELLA[folder]
spec = {"tasks": SPEC_TASKS[family]} if family in SPEC_TASKS else None
diff --git a/studio/frontend/src/features/audio/audio-cpp-catalog.ts b/studio/frontend/src/features/audio/audio-cpp-catalog.ts
index fb028f789d9..02f4d9974e9 100644
--- a/studio/frontend/src/features/audio/audio-cpp-catalog.ts
+++ b/studio/frontend/src/features/audio/audio-cpp-catalog.ts
@@ -211,6 +211,16 @@ export const AUDIO_CPP_UNOFFERED_FOLDERS: Readonly<Record<string, string>> = {
"OWSM-GGUF": "needs a newer audio runtime",
"Sidon-GGUF": "needs a newer audio runtime",
"Smart-Turn-v3-GGUF": "needs a newer audio runtime",
+ "AST-AudioSet-GGUF": "needs a newer audio runtime",
+ "CED-GGUF": "needs a newer audio runtime",
+ "FireRedVAD-GGUF": "needs a newer audio runtime",
+ "Hviske-v6-GGUF": "needs a newer audio runtime",
+ "MicroWakeWord-GGUF": "needs a newer audio runtime",
+ "MossFormer2-GGUF": "needs a newer audio runtime",
+ "Phonon-2-GGUF": "needs a newer audio runtime",
+ "Sherpa-KWS-GGUF": "needs a newer audio runtime",
+ "TF-GridNet-GGUF": "needs a newer audio runtime",
+ "Voxtral-4B-TTS-2603-GGUF": "needs a newer audio runtime",
"MioCodec-25Hz-44.1kHz-v2-GGUF": "the codec MioTTS downloads and loads with",
"Qwen3-ForcedAligner-0.6B-GGUF": "the aligner Transcribe uses for Qwen3-ASR timestamps",
"MiniMax-H3-Q4-GGUF": "an audio and video package Studio cannot load yet",
diff --git a/studio/frontend/tests/audio-cpp-catalog.test.ts b/studio/frontend/tests/audio-cpp-catalog.test.ts
index b731fd856a5..9b1aee93861 100644
--- a/studio/frontend/tests/audio-cpp-catalog.test.ts
+++ b/studio/frontend/tests/audio-cpp-catalog.test.ts
@@ -145,7 +145,7 @@ 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, 18);
+ assert.equal(Object.keys(AUDIO_CPP_UNOFFERED_FOLDERS).length, 28);
for (const [folder, reason] of Object.entries(AUDIO_CPP_UNOFFERED_FOLDERS)) {
assert.equal(audioCppModelFor(`${AUDIO_CPP_REPO}/${folder}`), null, folder);
assert.match(folder, /-GGUF$/);