Recommended pipelines (your use case: voice gen + covering vocals in existing songs)
AI cover (local, free)
Stems: UVR5 + Mel-RoFormer→
Convert vocal: RVC via Applio (or seed-vc zero-shot, no training)→
Mix back + reverb
AI cover (cloud, easiest)
Kits.ai — built-in stem split, pro singing clones, commercially safe licensed voices
Voice clone TTS (local M4)
Chatterbox Multilingual (MIT, RU, tops cloning arena)or
Qwen3-TTS (Apache, RU, runs via mlx-audio)
Voice clone TTS (cloud)
Fish Audio ($15/M, RU)·
MiniMax Speech 2.8 (10-s clone, 40+ langs)·
ElevenLabs PVC (best pro clone, premium price)
Russian guide vocal / TTS
F5-TTS + RU finetune (Misha24-10, 5000 h, stress marks — эквиритмика-friendly)·
Silero v5 (utility RU)
Quality score vs cost tier (click a point's card below for details)