Tool index · Voice
Voice,
and what each one does.
Every option for this job in one place: 3 listed tools read off the vendor's own page, 18 alternatives, and 23 free or open-source replacements. The label on each row is what it is.
If you came for ElevenLabs
the short answer
ElevenLabs is the default because the API is good and it handles dubbing and low-latency streaming. If you only need a finished voiceover, Speechify costs less; it is consumer-first, not an API. If you are not locked in: Kokoro-82M runs on a CPU and costs nothing, OpenVoice is the fastest open clone to a usable result, and XTTS-v2 is the one to fine-tune. Pay for ElevenLabs when you need the API, not for a one-off voiceover.
Alternatives
18 tools
| Tool | |
|---|---|
| Resemble AI Cloning and real-time API. | Open → |
| Cartesia Sonic, built for low-latency agents. | Open → |
| Deepgram STT plus Aura TTS. The API is the product, not the seat. | Open → |
| LMNT Fast, clean TTS. | Open → |
| Hume AI Expressive speech with emotion controls. | Open → |
| PlayHT Cloning and API TTS. | Open → |
| Amazon Polly Per-character TTS. The AWS free tier covers the first year. | Open → |
| Google Cloud TTS Studio voices are the strong part. | Open → |
| AssemblyAI | Open → |
| Rime Low-latency conversational TTS. | Open → |
| WellSaid | Open → |
| Camb.ai Dubbing and voice cloning across languages. | Open → |
| Fish Audio Cloning with a marketplace of voices. | Open → |
| MiniMax Speech Fast multilingual TTS. | Open → |
| Azure AI Speech Enterprise TTS with custom neural voice. | Open → |
| OpenAI TTS Clean TTS behind the API. | Open → |
| LOVO Voice and dubbing. | Open → |
| PlayHT / Play.ai Voice cloning and agents. | Open → |
Free and open source
23 replacements
Kokoro-82M
82M-parameter TTS that runs on CPU.
Repository → VoiceChatterbox
Emotion control and voice cloning from Resemble, released open.
Repository → VoiceFish Speech
Multilingual cloning with a hosted API if you do not want to run weights.
Repository → VoiceF5-TTS
Fast cloning from a short reference clip. Good enough for a 15-second VO.
Repository → VoiceCosyVoice 2
Alibaba's streaming TTS. Strong on Chinese and code-switching.
Repository → VoicePiper
Raspberry-Pi-grade TTS. Ugly next to ElevenLabs, instant and free.
Repository → VoiceXTTS-v2
Clones a voice from six seconds. Weights still run after the company closed.
Repository → VoiceStyleTTS2
Natural prosody on a single GPU. The strongest open English voice.
Repository → VoiceOrpheus TTS
LLM-backed speech with emotion tags. Built for realtime agents.
Repository → VoiceDia
Dialogue and non-verbal sounds from one model. 1.6B parameters.
Repository → VoiceZonos
Multilingual cloning with a quality slider you can hear.
Repository → VoiceCSM
Conversational speech model. The closest open match to a real conversation.
Repository → VoiceIndexTTS
Chinese-English cloning with precise duration control for dubbing.
Repository → VoiceOpenVoice
Instant cloning with tone and accent control. The fastest open clone to a usable result.
Repository → VoiceMegaTTS3
ByteDance cloning model. Strong on Mandarin and English with a small footprint.
Repository → VoiceSpark-TTS
LLM-backed TTS with controllable pitch and gender.
Repository → VoiceHiggs Audio
High-fidelity cloning and multi-speaker dialogue from one model.
Repository → VoiceVibeVoice
Long-form, multi-speaker podcast audio without the seams.
Repository → VoiceKyutai TTS
Streaming TTS built for realtime agents, not batch renders.
Repository → VoiceNeuTTS Air
Clones on a laptop CPU. The offline fallback when the API is down.
Repository → VoiceSeed-VC
Zero-shot voice conversion, so the performance stays yours.
Repository → VoiceRVC
The voice-conversion standard for singing and dubbing.
Repository → VoiceBark
Text-prompted audio with laughs and breaths. Weird, expressive, free.
Repository →Where this fits
4 buying moments
Recreate this
A teardown names the shots, the beats and the hook.
Rebuild the post →Production lineDirect and generate the film
Story to stitch: look, original AI cast, shot board, generation, consistency and the final cut.
Direct the film →Formats, 699 formatsProduce this format
A format is a shot list with a known cost.
Produce the format →StudioThe full production stack
One box: the actor, the voice, the cut and the bed, in the order a 30-second spot is actually made..
Open the stack →