Speech to Video API: Script or Audio to Video

Q: Can I send audio instead of a script?

Yes, on both models. Put the file's public URL in audio_url and leave out prompt and voice. Audio can be up to 600 seconds.

Short answer: A speech-to-video API takes a script or an audio file and returns a video of a person saying it, with the lips in sync. On HiAPI there are two: HeyGen Avatar IV turns one photo of your choice into the presenter, and HeyGen Avatar V gives you 497 preset presenters with transparent-background and caption options. Both cost $0.15 per second of video, so a 10-second clip is $1.50 either way, and both accept either a typed script (read by one of 102 voices) or your own recording. Try it now: open the Avatar IV model page, upload a portrait, type a line, and the person starts talking. New accounts get 200 free credits.

The two models at a glance

	HeyGen Avatar IV	HeyGen Avatar V
Model id	`heygen-avatar-iv`	`heygen-avatar-v`
Presenter	Your photo (`image_url`)	One of 497 presets (`avatar`)
Drive it with	`prompt` (script, up to 5,000 characters) or `audio_url` (up to 600 s)	Same
Voices for scripts	102 presets, default Warm Pro Narrator	Same 102
Resolution	720p or 1080p, same price	720p, 1080p or 4k
Aspect ratios	16:9, 9:16, 4:5, 5:4, 1:1, auto	16:9, 9:16 and more
Extras	Colour or image background fill	Transparent WebM, SRT captions
Price	$0.15 / second	$0.15 / second

Live HiAPI prices, checked October 16, 2026.

You never send a duration. With a script, length follows the words and the voice's pace; with audio, the video runs as long as the recording. Estimating cost means estimating the read-aloud time.

From a script: Avatar IV with one photo

A 9-second 1080p explainer from a single portrait, $1.35:

Hi, I'm Lin. With one photo and a short script, I can present your product or
record a lesson, with natural lip-sync and expressions.

Generate with this script → (upload your photo, voice Cute Chloe - Friendly, 1080p, 16:9)

heygen-avatar-iv, 1080p, 16:9, 9 seconds, $1.35. Sweater, hair and room come from the photo; lips, blinks and nods follow the voice.

{
  "model": "heygen-avatar-iv",
  "input": {
    "image_url": "https://your-cdn.example.com/portrait.jpg",
    "prompt": "Hi, I'm Lin. With one photo and a short script, I can present your product or record a lesson, with natural lip-sync and expressions.",
    "voice": "Cute Chloe - Friendly",
    "resolution": "1080p",
    "aspect_ratio": "16:9"
  }
}

From a script: Avatar V with a preset presenter

No photo needed. Daniel in an office, 1080p, about 8 seconds, $1.20:

Hi, I'm Daniel. Send me your quarterly numbers, and I'll walk your team
through what changed and what to do next.

Generate with this script → (pick an avatar and voice on the page, 1080p, 16:9)

heygen-avatar-v, 1080p, 16:9, about 8 seconds, $1.20.

From your own audio

Replace prompt and voice with audio_url. The lips follow your recording, which can be up to 600 seconds, and the price is still $0.15 per second of output:

{
  "model": "heygen-avatar-v",
  "input": {
    "avatar": "Abigail Office Front",
    "audio_url": "https://your-cdn.example.com/voiceover.mp3",
    "resolution": "1080p",
    "aspect_ratio": "9:16"
  }
}

Use this when a voice actor has already recorded the line, when you're dubbing an existing track, or when the speech comes from a text-to-speech model you prefer.

Which to choose

The presenter must be a specific real person, a founder, a teacher, a customer: Avatar IV with their photo.
No photo, or you'd rather not use one: Avatar V's presets.
Transparent background to composite over slides, or an SRT caption file: Avatar V.
4K: Avatar V. Avatar IV tops out at 1080p.
Vertical clips from a portrait photo: Avatar IV with aspect_ratio: "9:16" or auto.
Price: identical, so choose on the above.

What it costs

Length	Price
5 seconds	$0.75
10 seconds	$1.50
30 seconds	$4.50
60 seconds	$9.00

$10 buys about 66 seconds, enough to test one photo or preset with several scripts. Trim the warm-up line: every sentence you cut comes off the bill. Failed tasks are refunded.

Integration in three steps

POST /v1/tasks with "model": "heygen-avatar-iv" or "heygen-avatar-v" and the input above. The schema is strict: a stray duration field gets HTTP 400.
Poll GET /v1/tasks/{taskId} or receive a callback, until status is success (treat both fail and failed as failure).
Download output[0].url; the link expires.