Fish Audiodrama-3
fishaudio/drama-3
Text to speechFP8 / BF16CUDA
1 × B300drama-3-single-b300-engine
unverifiedAcceptance INCONCLUSIVE
INFERMATRIXLAB / MODEL CATALOG
Inference recipes.
Choose a model. Match your hardware. Get the commands.
6 models / 6 engines
TASK
HARDWARE
ENGINE
PRECISION
DEPLOYMENT
Deployment and performance acceptance are tracked separately.
Explore models 6
6 recipes · Source-pinned configurations
Fish Audio
Liquid AILFM2.5-350M
LiquidAI/LFM2.5-350M
Choice scoringFP16CUDA
1 × L40Ssystem1-omni
unverifiedAcceptance INCONCLUSIVE
ByteDanceOuro-1.4B
ByteDance/Ouro-1.4B
Text generationBF16CUDA
1 × H20-3evllm-rlt
unverifiedAcceptance INCONCLUSIVE
QwenQwen-Image-2.1-Turbo
Qwen/Qwen-Image-2.1-Turbo
Text to imageBF16CUDA
4 × B300 SXM6 ACvLLM-Omni
unverifiedAcceptance INCONCLUSIVE
QwenQwen3-TTS-12Hz-1.7B-CustomVoice
Qwen/Qwen3-TTS-12Hz-1.7B-CustomVoice
Text to speechBF16ASCEND
1 × Ascend 910B1qwen3-tts-engine
unverifiedAcceptance INCONCLUSIVE
QwenQwen3-TTS-12Hz-1.7B-CustomVoice-4bit
mlx-community/Qwen3-TTS-12Hz-1.7B-CustomVoice-4bit
Text to speech4-bitMLX
1 × M4 (16 GB unified memory)vllm-omni-mlx
unverifiedAcceptance INCONCLUSIVE
Only the listed model and hardware combinations are documented. Review the source and verification scope before deployment.