HeyGen: HeyGen Voice Turbo

HeyGen Voice Turbo is a text-to-speech model from HeyGen, listed on api14 under the model id heygen/heygen-voice-1-turbo. It sits in the speech category and is positioned as a lower-latency version of HeyGen Voice, the company's core text-to-speech model. If you already know HeyGen for its avatar and video tooling, this is the same vendor's voice synthesis line, packaged here as a single speech endpoint you can call alongside your other models.

The model takes text as its only input and returns speech as its only output. It ships with the same 10 default voices as HeyGen Voice: nadine, tabitha, bram, gareth, noelle, meredith, sabrina, lucian, weston and vincent. If those defaults are not enough, you can bring your own HeyGen key to use voice clones, which lets you generate speech in a voice you have already set up on HeyGen's side. There is no published context length for this model, so treat the text you send as a single utterance or script segment rather than something with a documented token budget. The model exposes no supported request parameters beyond what the endpoint itself requires.

The obvious use cases follow from text in, speech out, with latency as the selling point. Narration and voiceover for short video clips, spoken versions of articles or summaries, audio for product walkthroughs, and voice prompts inside interactive applications are all reasonable fits. Because it is the turbo variant, it suits pipelines where a person is waiting on the audio, such as an assistant that speaks a reply or a tool that reads a result back. The voice clone option matters if you need brand or character consistency across many clips rather than a generic narrator.

You call it through api14's OpenAI-compatible API using the model id heygen/heygen-voice-1-turbo, the same way you would call any other speech model on the gateway. Point your existing OpenAI client at the api14 base URL, send your text as the input, and select one of the ten voices. Since the model lists no supported parameters, keep your request minimal and avoid assuming sampling controls, speed settings or format options that are not documented. If you need voice clones, supply your own HeyGen key as described by the provider.

There are a couple of cases where this model may not be the right pick. It only accepts text and only produces speech, so it cannot transcribe audio, translate speech or generate music, and it will not help with image or video generation even though HeyGen is known for avatars. There is also no published context length, which makes it a poor choice for very long documents in a single call; chunking your script into shorter segments is the safer approach. If you need a model with documented parameter controls or a defined context window, check the other speech models on api14 before committing.