Introduction
HeyGen's highest-fidelity digital presenter model with context-matched gestures and native 2K clarity.
| Field | Value |
|---|---|
| Input | Image, Audio |
| Output | Video |
| Model ID | heygen/avatar-6/talking-avatar |
| Publisher | HeyGen |
| Required inputs |
|
| Audio length | 1 to 600 seconds (up to 10 minutes); video duration matches input audio |
| Resolution tiers | 1080p (default, 1080px short edge) or 2k (1620px short edge) |
| Custom dimensions | Optional paired even width and height, 256–3840 pixels each |
| Motion controls | Optional prompt (up to 1,200 chars), expressiveness (default or less), and 1 |
| Current pricing | $0.12/second (1080p) or $0.174/second (2K), rounded up per whole input audio second |
| Failed requests | Not charged |
| Live run verified (2026-11-19) | 3.79s portrait audio at 1080p returned a 1088×1632 MP4 in ~35s |
Key features & highlights
Hushed eye contact with unhurried micro-expressions
In the champagne satin studio sample at 1080p, warm amber studio light rests across the bare shoulder while the speaker delivers a soft, close-range line. Natural script-aware movement pairs subtle head tilts with a slow, confident smile rather than stiff repeated head loops. Supply a clean front-facing portrait and concise motion wording when an intimate late-night cadence matters.
Crisp fabric and facial detail at 1620px short edge
Rendered at 2K on a candlelit night balcony, the dark velvet off-shoulder sample holds soft highlights across skin and evening fabric through a calm spoken monologue. Selecting 2k expands the output short edge to 1620 pixels, suited to close-up portraits and large displays. Pairing steady warm lighting in image_url with gentle posture cues keeps blinks and transitions smooth.
Prompt direction, expressiveness, and motion reference
HeyGen says Avatar 6 examines roughly 30 seconds of surrounding speech context so posture shifts and hand gestures align with phrasing. Guide mood or gaze with up to 1,200 characters in prompt, switch expressiveness to less for calmer delivery, or attach one MP4 or WebM clip through reference_video_urls for animated or illustrated character motion.
Request parameters
| Parameter | Required | Description |
|---|---|---|
image_url | Yes | HeyGen Avatar 6: portrait photo or illustration (up to 32 MB). at least 1 characters |
audio_url | Yes | HeyGen Avatar 6: speech audio driving lip sync and clip length. at least 1 characters |
prompt | No | HeyGen Avatar 6: optional plain-text motion and gesture direction (up to 1,200 chars). at least 1 characters · at most 1200 characters |
resolution | No | HeyGen Avatar 6: 1080p or 2k output tier (changes per-second rate). |
expressiveness | No | HeyGen Avatar 6: default for natural gestures or less for calmer delivery. |
reference_video_urls | No | HeyGen Avatar 6: optional single reference video (≥ 2s; illustrated/animated on standard workspaces). at least 1 items · at most 1 items |
width | No | HeyGen Avatar 6: optional even output width in pixels (paired with height). |
height | No | HeyGen Avatar 6: optional even output height in pixels (paired with width). |
Pricing
| Specification | Price |
|---|---|
| $0.12 / second |
| $0.174 / second |
Use cases
Close-up brand and executive films
Pair a single studio headshot with recorded narration at 2K (1620px short edge) to produce crisp founder updates or product launches without booking a camera crew.
Intimate late-night character monologues
Combine atmospheric adult portraits—like our off-shoulder champagne satin studio clip or candlelit velvet balcony sample—with hushed voiceovers for dramatic previews and narrative vignettes.
Long-form lessons and studio briefings
Drive continuous talking-head segments up to 10 minutes from a single audio track, relying on phrase-aware posture and gesture timing instead of short repetitive loops.
Portrait and widescreen custom layouts
Keep the uploaded portrait proportions—as in our verified 1088×1632 vertical test—or specify exact even width and height pairs from 256 to 3840 pixels for fixed editorial slots.
Animated and illustrated character clips
Attach a 2-second or longer MP4 or WebM motion clip in reference_video_urls on standard workspaces to guide rhythmic cadence for illustrated or animated speakers.
Prompt guide & pro tips
Choose an authorized portrait
Upload a clear JPEG, PNG, or WebP portrait (up to 32 MB) of a consenting adult you have permission to depict. Frame lighting, wardrobe, and setting in the source image.
Prepare clean spoken audio
Provide an MP3 or WAV recording between 1 and 600 seconds in
audio_url. Review pacing and pauses beforehand, because whole input audio seconds set the charge.Set resolution and motion cues
Pick 1080p or 2k, optionally add a plain-text prompt up to 1,200 characters, and keep expressiveness at default or set less for quieter delivery.
Verify output and download MP4
Optionally pair even width and height values (256–3840 pixels) or keep the portrait ratio. Inspect lip sync, eye line, and framing, then save your finished MP4.
Native 1080p and 2K talking portraits from photo and speech
heygen/avatar-6/talking-avatar animates a single portrait image and spoken audio recording into an asynchronous 1080p or 2K MP4. HeyGen says Avatar 6 analyzes roughly 30 seconds of surrounding speech context so posture changes, hand movements, and facial nuance follow phrasing rather than repeating short motion loops.
Each input carries a clear role: image_url sets the speaker's identity, wardrobe, lighting, and scene; audio_url supplies the spoken words and clip length (1 to 600 seconds); and optional prompt, expressiveness, and reference_video_urls refine physical delivery.
Parameters
| Parameter | Required | Type & Limits | Purpose |
|---|---|---|---|
image_url | Yes | URI (JPEG, PNG, WebP, max 32 MB) | Clear portrait photo or illustration of one adult subject; output keeps its aspect ratio by default. |
audio_url | Yes | URI (MP3 or WAV, 1–600s) | Spoken audio track driving lip sync, gesture timing, and final clip duration. |
prompt | No | String (1–1,200 chars) | Plain-language cues for expression, gaze, head tilts, posture, and hand movement. |
resolution | No | 1080p (default) or 2k | Output tier: 1080p (1080px short edge) or 2k (1620px short edge); sets the per-second rate. |
expressiveness | No | default (default) or less | Gesture amplitude: default for natural script-aware motion, less for calmer delivery. No price change. |
reference_video_urls | No | Array of 1 URI (MP4/WebM, >= 2s) | Optional motion/style reference clip; animated or illustrated characters on standard workspaces. |
width | No | Even integer (256–3840) | Optional exact output width in pixels; must be paired with height. |
height | No | Even integer (256–3840) | Optional exact output height in pixels; must be paired with width. |
Pricing and cost examples
Successful runs are billed per whole second of input audio (rounded up, 1-second minimum, 600-second maximum). Changing expressiveness, prompt, or width/height does not alter the tier rate, and failed requests are not charged.
| Resolution | Per-second rate | 10-second audio | 60-second audio |
|---|---|---|---|
1080p (default) | $0.12 / second | $1.20 | $7.20 |
2k | $0.174 / second | $1.74 | $10.44 |
Listen to the full recording before submitting so silence tails or extra takes do not inflate billed seconds.
Steering expressiveness, motion prompts, and reference video
Expressiveness and text direction
Keep spoken words inside audio_url and reserve prompt for physical delivery—for example: "She speaks softly with warm eye contact, subtle head tilts, and a confident slow smile." Establish wardrobe, background, and lighting in image_url rather than expecting prompt to redraw the scene.
Leave expressiveness at default when you want Avatar 6 to match gesture energy to spoken emphasis. Switch to less for newsroom briefings, quiet late-night monologues, or tight close-ups where minimal torso movement reads cleaner.
Using reference_video_urls
Pass a single MP4 or WebM clip of at least 2 seconds in reference_video_urls when you want an animated or illustrated character to echo a particular rhythm. Realistic human reference videos require dedicated consent approval and are rejected on standard workspaces.
Verified live run and sample notes
In our 2026-11-19 live test, a 3.79-second portrait audio track at 1080p finished in roughly 35 seconds and returned a 1088×1632 MP4 preserving portrait proportions. Because generation is asynchronous, use status polling or webhooks in production pipelines.
- Champagne studio whisper (
1080p): Soft studio lighting on an off-shoulder champagne satin portrait paired with a quiet close-range line, showing warm eye contact and gentle head tilts aligned with speech. - Velvet balcony monologue (
2k): Dark velvet off-shoulder evening portrait on a candlelit night balcony rendered at 1620px short edge, preserving fabric sheen, steady blinks, and calm posture shifts.
Choosing between Avatar 6, Avatar 6 Turbo, and Avatar IV
- Pick HeyGen Avatar 6 when close-up facial detail, 2K clarity, and unprompted micro-expressions are your primary goals.
- Pick HeyGen Avatar 6 Turbo when running bulk personalized outreach, high-volume localized variants, or rapid script drafts where quicker turnaround and lower per-second rates matter more than fine close-up nuance.
- Pick HeyGen Avatar IV if you specifically need
background_image_urlscene swapping or three-level (low/medium/high) expression controls at up to 1080p.
Consent, likeness rights, and permitted inputs
Use portraits and voices of consenting adults you are legally authorized to depict, and never impersonate another person. Verify rights to both image_url and audio_url prior to submission, and download completed MP4 outputs promptly once verified.
API quick start
# Generate the idempotency key once; reuse it when retrying after a lost response.
export SPICY_IDEMPOTENCY_KEY="${SPICY_IDEMPOTENCY_KEY:-$(uuidgen)}"
curl -X POST https://api.spicyapi.ai/api/v1/jobs/createTask \
-H "Authorization: Bearer $SPICY_API_KEY" \
-H "Content-Type: application/json" \
-H "Idempotency-Key: $SPICY_IDEMPOTENCY_KEY" \
-d '{
"model": "heygen/avatar-6/talking-avatar",
"input": {
"audio_url": "https://example.com/reference.mp3",
"expressiveness": "default",
"image_url": "https://example.com/reference.png",
"resolution": "1080p"
}
}'
# Copy data.taskId from the response into TASK_ID, then poll the status.
TASK_ID="TASK_ID_FROM_RESPONSE"
curl "https://api.spicyapi.ai/api/v1/jobs/recordInfo?taskId=$TASK_ID" \
-H "Authorization: Bearer $SPICY_API_KEY"
See the API reference for authentication, the complete schema, polling, and error handling.



