Cycling through all six. Tap any point to stop.

Measured on six things

Audio-driven AI video generator that turns one image and ...

Security

HTTPS + HSTS in place, but no CSP, frame protection, X-Content-Type-Options, Referrer-Policy, Permissions-Policy, SPF or DMARC. No SOC 2/ISO/pen-test claims. Table stakes only.

Screenshots

1 / 3

OmniHuman 2.0

About this app

OmniHuman 2.0 is an audio-driven AI video generator from ByteDance that transforms a single image and audio track into a 720p digital human video up to 5 minutes long. Unlike typical video tools that start with text, OmniHuman 2.0 lets audio drive the performance—lip movements, intonation, and emotion all sync seamlessly. You can add an optional prompt to control actions, camera moves, and background, giving you director-level control. Standout features include audio comprehension (the character can act on what's said), emotional performance, multi-person scenes, and support for animals, cartoons, and anime characters. It's ideal for marketers, filmmakers, musicians, educators, and content creators who want to produce talking or singing avatar videos without a camera crew. Pricing is credit-based: 2 credits per second of video, with 20 free credits to start (enough for one 10-second clip). Credit packs start at $19.90 for 100 credits, and monthly plans are available. Compared to Omn...

Key features

1Audio-driven avatar generation: one image + one audio track → 720p video up to 5 minutes, with lip sync, intonation and emotion following the audio.
2Optional prompt control for actions, camera moves and background, plus audio comprehension so the character can act on what's said.
3Multi-person scenes (choose who speaks) and support for animals, cartoons, anime and pixel-art characters.
4Transparent credit pricing: 2 credits/second, 20 free credits at signup, one-time packs from $19.90 and monthly plans from $29.90.

Target customer profile

Drawn from the product itself, not from a survey.

Demographic

Marketing and e-commerce teams

Pain points

Creating consistent presenter videos for multiple products is time-consuming and costly.

Primary needs

Quickly generate personalized video ads with a consistent virtual presenter.

Demographic

Film, short-drama, and video creators

Pain points

Shooting dialogue scenes requires actors, sets, and expensive equipment.

Primary needs

Produce character-driven scenes from a single image and audio with directable actions.

Demographic

Musicians and music-video makers

Pain points

Producing music videos with performers is expensive and logistically complex.

Primary needs

Turn songs into engaging performance videos with synced lip movements and emotional expression.

Demographic

Educators and podcasters

Pain points

Creating video content from audio requires additional recording and editing.

Primary needs

Convert audio lessons or podcasts into video with a virtual presenter up to 5 minutes long.

Scout Review

Written by AI from measured evidence, scored out of 100.

OmniHuman 2.0 is a polished, fast front-end to ByteDance's audio-driven avatar model, sold through transparent credit pricing with a real free trial. The landing page and studio panel are well-crafted, the site is genuinely quick, and the feature set — 5-minute clips, multi-person scenes, animal/cartoon avatars, audio comprehension — is a credible differentiator against D-ID, Synthesia and HeyGen. The gaps are the ones you'd expect from a young reseller skin: no security hardening beyond HTTPS/HSTS, no compliance claims, no accessibility statement, and no visible generation progress. Worth a try on the free credits; go in knowing the moat is UX, not model.

Measure by measure

Usability

Required labels, file-type/size hints, live per-second cost estimate and 20 free credits make the path to a first clip short. Missing a visible generation progress indicator and deeper docs.

Design

Cohesive dark-navy/cyan system, strong hero hierarchy, and a studio panel that cleanly separates required inputs from the optional prompt. Docked for dense text walls and tiny feature-card body copy.

Speed

Lighthouse mobile 95/100 (LCP 1.6s, CLS 0, TBT 140ms) and desktop 99/100. The marketing site is genuinely fast; no lag complaints in collected evidence.

Security

HTTPS + HSTS in place, but no CSP, frame protection, X-Content-Type-Options, Referrer-Policy, Permissions-Policy, SPF or DMARC. No SOC 2/ISO/pen-test claims. Table stakes only.

Accessibility

Lighthouse a11y 100/100 with only two minor failing audits (label accessible names, table headers). No WCAG statement or multi-language support found, which caps it below the top of the band.

Growth

Real problem, real wedge (audio-driven performance, 5-min clips, multi-person/animal support) and transparent credit pricing — but it's a reseller skin on BytePlus's model, so the moat is thin.

Improvement areas

Security hardening: add CSP, frame protection, X-Content-Type-Options, Referrer-Policy, Permissions-Policy, SPF and DMARC.
Publish a security/compliance page (data retention for uploaded faces and audio, SOC 2 or pen-test status).
Add a generation progress indicator and ETA to the studio preview player.
Improve homepage copy scannability and add an accessibility statement plus multi-language support.

Design & Usability

The homepage is a competent, cohesive dark-navy/cyan build with a strong hero and a studio panel that actually separates required inputs (image, audio) from the optional prompt — with file-type and size hints right where you need them. The live '2 credits / second' estimate is a genuinely thoughtful touch. Where it slips: the 'What is OmniHuman 2.0?' section is a dense text wall, feature-card body copy is small enough to squint at, and the preview player has no visible generation progress indicator, so after you hit Generate you're left guessing whether it's rendering or dead. Onboarding is otherwise short — 20 free credits gets you a 10-second clip without a card.

Speed & Security

Speed is the standout: Lighthouse mobile 95/100 with LCP 1.6s, CLS 0 and TBT 140ms, and desktop 99/100 — the marketing site is genuinely snappy and there are no lag complaints in the collected evidence. Security is the weak flank. HTTPS and HSTS are enforced, but the header probe found no CSP, no frame protection, no X-Content-Type-Options, no Referrer-Policy, no Permissions-Policy, and the DNS shows no SPF or DMARC — 4 of 12 checklist points. There's no SOC 2, ISO, pen-test or bug bounty claim anywhere on the site. For a tool that ingests your face and voice, that's a thin public security story.

Accessibility & Growth

Accessibility scores 100/100 on Lighthouse's automated audit, with only two minor failing checks (visible labels without matching accessible names, large-table cells without headers). There's no WCAG statement and the site is English-only, which keeps it out of the top of the band. Growth is the interesting one: the problem is real — turning audio into a performing avatar without a camera crew — and the wedge is specific: audio-driven performance, up to 5-minute single-run clips, multi-person scenes, and support for animals and cartoons, which D-ID, Synthesia and HeyGen don't all match. Pricing is transparent and competitive. The honest limit: this is a storefront on BytePlus's OmniHuman 2.0 model, so the differentiation lives in UX and price, not in technology the incumbents can't copy.

Conclusion

If you need talking-avatar video and want to test audio-driven performance without a subscription commitment, the 20 free credits are a fair audition. If you're putting faces and voices of real people through it at scale, wait for a security page and a compliance claim before you sign a team up. And if you're betting a business on it, remember the model underneath belongs to BytePlus, not to omnihumanai.net.

Suggestions for improvement

Add a visible generation progress indicator and ETA to the preview player so users know it's rendering.
Ship a security page with CSP, DMARC/SPF and a stated data-retention policy for uploaded faces and audio.
Break the 'What is OmniHuman 2.0?' text wall into scannable sub-sections or an accordion.
Publish an accessibility statement and consider a second language — the model is global, the site isn't.

How we score

Alternatives comparison

Named competitors, point by point. Nobody paid to appear here or to be left out.

	OmniHuman 2.0	D-ID	Synthesia	HeyGen
Primary input mode	Audio-driven (image + audio + optional prompt)	Text or audio to talking avatar	Text to avatar video (script-based)	Text to talking avatar video
Max single-generation length	Up to 5 minutes (300s) at 720p	Short clips (typically under a minute per render)	Long-form via scene stitching	Long-form via scene stitching
Multi-person / non-human avatars	Yes — multi-person scenes, animals, cartoons, anime, pixel art	Single presenter focus	Multiple presenters via scenes; human avatars	Multiple avatars; human-focused
Free tier / trial	20 free credits (~10s clip)	Free trial credits available	Free demo video; paid plans for full access	Free tier with limited credits
Entry pricing	$19.90 one-time (100 credits) / $29.90 per month	Subscription tiers starting in the low tens per month	Subscription, higher entry point for teams	Subscription tiers from ~$24/month