Sulat.com
AI models
$10 off the fastest DeepSeek V4.1 Flash, Kimi K3 and GLM 5.3 from Synthetic
Vercel AI Gateway logo

Model details

HeyGen Video

HeyGen Video is designed around a single guiding idea: the written brief is the shot. Rather than improvising beyond what the user specifies, the model constructs the subject, setting, lighting, and sound in one pass from a natural-language description, leaving unspecified choices to default behaviour. This literal approach to prompting rewards detail. The prompt field accepts up to roughly 32,000 characters, and guidelines recommend a few hundred to a few thousand characters over a one-line idea, with longer briefs giving noticeably better fidelity to the requested composition, wardrobe, camera, and audio cues.

HeyGen as a company has been building text-driven video tools since its founding in late 2020, combining proprietary video models with third-party language and audio engines to make natural-language video creation accessible without traditional editing. That product lineage frames HeyGen Video as a tool for users who want precise, brief-driven control over generative footage—particularly useful when a creator needs a specific mood, camera setup, wardrobe, or sonic environment spelled out explicitly. The model's literal interpretation of prompts makes it a strong fit for teams that prefer writing detailed shot lists to iterating against unpredictable model improvisation.

Vercel AI Gatewayheygen/heygen-video-1

Quick Info

Powered by
Provider
Vercel AI Gateway
Model key
heygen/heygen-video-1
Release date
Oct 9, 2026
Last updated
Oct 9, 2026
Input modalities
Output modalities
Capabilities

Limits

Output tokens
0 tokens
Context window
0 tokens

Latest news about HeyGen Video

Vercel AI Gateway

Coverage

According to a MindStudio pricing analysis dated October 2, 2026, HeyGen Video is HeyGen's first general-purpose video generation model, positioned as a budget-friendly alternative rather than a flagship cinematic model. Launch pricing through end of October runs 1 cent per second for 480p output and roughly 1.5 cents per second for 768p, both at a 50% discount, with regular rates of 2 cents and about 3 cents per second respectively after the sale ends. Access is API-first at launch, with ComfyUI integration and a native HeyGen web interface planned as later additions. The post characterizes the model as a fine-tune on top of Minimax H3 rather than a model trained from scratch, implying its output behavior will closely track other H3 fine-tunes already circulating in the market. The pitch centers on speed and cost rather than top-tier fidelity.

Vercel AI Gateway

CoverageDocumentation

HeyGen's official developer catalog for HeyGen Video 1.0 documents a prompt-to-video general-purpose model with per-second list pricing of $0.03/s at 768p with audio. Its Arena Elo is normalized to 1000, ahead of Seedance 2.0 (955), H3 Max (952), H3 Max Turbo (920), H3 Max balanced (860), Kling 3.0 Pro (857) and Veo 3.1 (744). In inference benchmarks it renders a 10-second image-to-video clip in 3.7 seconds, faster than H3 Max Turbo (4.1s) and H3 Max (8.3s). Per-axis quality scores for HeyGen Video include True to image 65%, Follows prompt 62%, Natural motion 63%, Timing 66%, Physics 59%, Camera motion 47%, Consistency 43%, Detail 46%, Sound and voices 57%, Music 45%, and No extras 57%, drawn from 16 to 115 ranked answers. The catalog lists the model alongside H3 Max, H3 Max Turbo, Seedance 2.0, Kling 3.0 Pro and Veo 3.1 on a price-versus-quality scatter, positioning it as the highest-rated entry on both axes.

Videos about HeyGen Video