voiceclap-commercial

VoiceCLAP-Commercial (small)#

A fully commercially-licensable voice-text contrastive (CLAP-style) embedding model for speech-emotion and talking-style retrieval. Same 110M dual-tower architecture as laion/voiceclap-small, but trained only on data that permits commercial use — and, perhaps surprisingly, it matches or beats the non-commercial-data model on 4 of 5 benchmarks.

Why this model exists#

The standard VoiceCLAP training mix contains corpora released under CC BY-NC (non-commercial) terms: Expresso and EARS (both Meta, CC BY-NC 4.0), and the bulk of Emilia (the original 101k-hour split is CC BY-NC 4.0). Whether such terms bind the weights of a model trained on that data is a question each user must answer for their own jurisdiction and use; LAION licenses its releases on its own assessment. This model exists for users who prefer not to have to answer it: every corpus in its training lineage permits commercial use.

This model removes every non-commercial source and keeps only commercially usable data. A controlled ablation (below) shows the small model loses nothing by doing so.

Training data (all commercially licensable)#

The non-commercial Emilia clips are filtered out at training time by clip-id (original-Emilia ids look like EN_B00087_S08178_W000004; the retained YODAS clips carry YouTube-style ids). Expresso and EARS are not used. Captions use the __moss_short__ scheme: the MOSS-Audio-8B-Thinking emotion sentence plus one sampled talking-style sentence, 50/50 mixed with the corpus's own caption.

Architecture#

Identical to voiceclap-small — a dual-tower CLAP:

PCL adds 39 learned emotion prototypes and a cross-entropy term on z-scored pseudo-labels derived from emolia's emotion-annotation scalars — a small auxiliary loss that sharpens the emotion subspace.

Results — commercial data costs nothing#

Controlled ablation, all arms trained identically (1 node, __moss_short__ + PCL w=0.2, 15 epochs), evaluated at each arm's best epoch:

This model wins 4 of 5 metrics — emonet top-1, VoiceNet-Emo balanced accuracy (+0.0115), emolia Spearman ρ, and MAEB-voice (+0.009). It trails only on emonet ρ (−0.025, fine-grained intensity ranking on synthetic audio, where the dropped Emilia diversity helped). On the in-domain emolia benchmark and the 8-task MAEB-voice suite it is strictly better than the non-commercial-data model.

Two recovery experiments were tried and discarded: adding VoxCeleb1/2 (CC BY) hurt, and upweighting the existing safe corpora was flat. Plain commercial data is best.

Absolute numbers are from a 1-node training scale used for the controlled ablation; treat them as relative (commercial-vs-noncommercial deltas), not as the maximum achievable with a full multi-node run.

Usage#

import torch, soundfile as sf
from transformers import AutoModel, AutoTokenizer

model = AutoModel.from_pretrained("laion/voiceclap-commercial", trust_remote_code=True).eval()
tok = AutoTokenizer.from_pretrained("laion/voiceclap-commercial")

# audio: raw mono waveform @ 16 kHz
wav, sr = sf.read("clip.wav", dtype="float32")
audio_emb = model.encode_waveform(torch.from_numpy(wav))

# text
t = tok(["a person speaking with quiet pride in their voice"], padding=True, return_tensors="pt")
text_emb = model.encode_text(t["input_ids"], attention_mask=t["attention_mask"])

score = (audio_emb @ text_emb.T).item()  # cosine similarity (both L2-normalised)

Conversion from the training checkpoint was verified functionally against the original open_clip implementation (cosine ≥ 0.99999 on both towers).

License#

CC-BY-4.0 — all training data is commercially usable (Emilia-YODAS CC BY 4.0, LAION's Got Talent, in-house Majestrino), and the architecture/weights carry no non-commercial restriction. This is the distinguishing feature of this model versus the standard VoiceCLAP releases, whose training mix includes CC BY-NC corpora.