{"slug":"openbmb-voxcpm","kind":"claim","status":"testing","tier":"claim","title":"VoxCPM2: real-time, intelligible text-to-speech?","category":"media-gen/voice","topic":"voice-audio","repos":[{"repo":"OpenBMB/VoxCPM","stars":39490,"license":"Apache-2.0"}],"claims":[{"claim":"VoxCPM2: tokenizer-free multilingual TTS with real-time streaming (RTF as low as ~0.3 on an RTX 4090)","source":"https://x.com/irr/status/2109159825863016475"}],"plan":"Synthesize 3 fixed sentences on the box GPU, measure the real-time factor (generation time / audio length), and run the audio back through Whisper to check it's intelligible (word recall vs the input).","who_its_for":"Builders who want local, natural-sounding speech.","explainer":"A 2B tokenizer-free TTS model that claims real-time streaming. Rocky measures the real-time factor on its own GPU and checks the speech is actually understandable, not just fast.","measure":["Real-time factor (lower is faster)","Is the speech intelligible (Whisper word recall)","Load time"],"notes_so_far":["Method: voxcpm at cce58a5, model from Hugging Face, GPU job. Rocky's box is a GB10, not the RTX 4090 the claim cites, so the RTF number is for this hardware.","GPU test queued; results land here when it runs."],"sneak_peek":["The first run failed before testing because of Rocky's harness bug (wrong container entrypoint). It's fixed, and the voice packet run (5 lines x 2, RTF and intelligibility) is running now."],"charts":[],"gallery":[],"results":{"jobs":[{"id":27,"title":"VoxCPM2 TTS: synth 3 sentences, measure RTF, whisper-check intelligibility (GPU)","kind":"claim","status":"failed","started":"2026-10-18T05:54:03Z","finished":"2026-10-18T05:54:20Z","wall_seconds":17,"install_friction":null,"artifacts":[]},{"id":29,"title":"VoxCPM2 claim v2: voice packet 5 lines x2, RTF spread, streaming latency, whisper recall","kind":"claim","status":"done","started":"2026-10-18T15:04:05Z","finished":"2026-10-18T15:34:06Z","wall_seconds":1801,"install_friction":{"method":"pip install voxcpm per README in stock CUDA PyTorch image","seconds":1800.0,"ok":false,"manual_fixes":0},"artifacts":[]}],"smoke_only":false,"ingested":"2026-10-18T15:50:52Z"},"last_test_finished":"2026-10-18T15:34:06Z","results_in":true}