# Av · Avatar Presenter — element 21 of 58 > A face for the camera-shy. Turns text into a presenter on screen. - **Group:** 5 · Creative & Design - **Necessity:** Optional - **Price band:** $$ · $30–150/mo - **Maturity:** Emerging - **Edition:** v2026.Q3 · verified 2026-09-13 ## Leading tools (v2026.Q3) - **HeyGen** — the business standard - **Synthesia** — enterprise polish, audited trust - **Tavus** — real-time conversational api - **D-ID** — developer api veteran - **Mirage (ex-Captions)** — generated actors, not clones ## Our take Great for scaled explainers and localization. Use sparingly where trust matters — audiences can tell. ## Combines with Vi, Vo ## The top 5 — deep dossier (verified 2026-09-13) HeyGen, for almost everyone — it doubled to $200M ARR in eight months (Jun 2026), ships the fastest model cadence in the category (Avatar V, Apr 2026), and its price ladder runs from a free tier to Fortune-100 contracts. Synthesia is the pick when enterprise L&D, compliance, and governance decide: $4B valuation, 140%+ net retention, ISO 42001-certified, and a first-wave signatory of the EU's Article 50 deepfake-labeling code. Tavus wins when the avatar must talk back — real-time conversational video via API. D-ID when you want the cheapest developer API with the longest track record; Mirage (ex-Captions) when a fully generated actor beats cloning anyone at all. One dimension now separates vendors as much as realism does: consent and deepfake policy — all five require consent for likeness cloning, but the depth ranges from Synthesia's audited, moderation-at-creation regime to policies you have to dig for. 1. **HeyGen** (HeyGen (Surreal AI)) — Free 3 videos/mo · Creator $29/mo · Pro $49/mo · Business $149/mo (+$20/seat) · Enterprise custom. Best for: Founders and marketing/content teams who want the most realistic presenter clone with the widest use-case spread — explainers, localization, ads, live avatar APIs — at prosumer prices. Why: The category's growth story: ARR doubled to $200M in eight months (announced Jun 25, 2026), 30M+ users across 196 countries, claimed adoption in 85% of the Fortune 100. Avatar V (Apr 8, 2026) reset the realism bar, and the roadmap keeps compounding — Video Agent, live avatar APIs, Seedance 2.0 integration. Consent is enforced at capture: creating a likeness avatar requires a recorded verbal-consent video with a spoken dynamic passphrase. Watch: Valuation is stale relative to revenue — last priced at $500M (Jun 2024 Series A); no verified 2025–26 round. Credit-based pricing makes heavy use expensive fast, and enterprise governance artifacts (certifications, published audits) trail Synthesia's. [https://www.heygen.com](https://www.heygen.com) 2. **Synthesia** (Synthesia) — Free 10 min/mo · Starter $29/mo ($18 annual) · Creator $89/mo ($64 annual) · Enterprise custom. Best for: Enterprises producing training, onboarding, and internal comms at scale in 140+ languages — where procurement asks for governance paperwork, not demo reels. Why: The enterprise incumbent: $4.0B valuation after a $200M Series E led by GV, 65,000+ customers, 90% of the Fortune 100, 140%+ net revenue retention (Apr 2026). It also owns the trust high ground — first generative-AI company ISO/IEC 42001-certified, moderation embedded at the point of creation, and among the first AI-video companies to sign the EU's Article 50 Code of Practice on labeling AI content (Jul 29, 2026). Roleplay Sessions (Jul 2026) pushes it into interactive coaching. Watch: Priciest per-minute path to custom avatars; creative/social output looks corporate next to HeyGen and Mirage. History cuts both ways: its avatars were used in state propaganda (Freedom House, Oct 2023) — the strict pre-screening regime exists because it was burned first. [https://www.synthesia.io](https://www.synthesia.io) 3. **Tavus** (Tavus) — Free 20 CVI min/mo · Starter $22/mo · Builder $59/mo · Growth $397/mo · Business $975/mo + usage · Enterprise custom. Best for: Product teams embedding a face that converses in real time — AI interviewers, tutors, concierges — via the Conversational Video Interface API. Why: The real-time leader: CVI turns a replica into a sub-second video conversation partner, and the whole company is API-first where HeyGen/Synthesia are editor-first. $40M Series B led by CRV (Nov 2025, $64.2M total) funds the research bet, and the PALs consumer line ('agentic AI humans') previews where presenters become agents. Consent is structural: replica training fails without a spoken consent statement recorded in the footage itself. Watch: Per-minute conversational pricing compounds quickly at scale ($975/mo buys 4,000 minutes). For plain script-to-video output, HeyGen and Synthesia render better-produced results; no disclosed revenue figures. [https://www.tavus.io](https://www.tavus.io) 4. **D-ID** (D-ID (merged with simpleshow, Sep 2025)) — Free trial · Studio & API subscription tiers (pricing table renders client-side, Aug 2026 — verify in-app) · Enterprise custom. Best for: Developers who want the cheapest, longest-serving avatar API — talking-photo videos, translations, and embeddable real-time Agents. Why: The API veteran: 280,000+ developers, 200M+ avatar videos generated, and a clean three-layer line — Studio, Video APIs (V4 Expressive launched Feb 2026), and Realtime Agents with WebRTC streaming and knowledge bases. The Sep 2025 combination with Berlin's simpleshow added an enterprise explainer-video channel. Ethics governance is board-level — advisory board includes Privacy-by-Design creator Ann Cavoukian. Watch: Squeezed from both sides: HeyGen out-ships it on realism, Tavus and Anam on real-time latency. The simpleshow deal followed a funding drought (last major round 2022), and mid-2025 reporting framed simpleshow as the acquirer — the combined entity's trajectory is the open question. Photo-based avatars remain the weakest consent surface in the category. [https://www.d-id.com](https://www.d-id.com) 5. **Mirage (ex-Captions)** (Mirage) — Free · Max $24.99/mo · Scale $69.99–$279.99/mo · Enterprise custom (training-data exclusion). Best for: Creators and performance marketers who want expressive generated actors for social-format video — UGC-style ads and talking clips without filming or cloning anyone. Why: The foundation-model insurgent: rebranded from Captions to Mirage (Sep 4, 2025) around omni-modal models that generate actor, voice, and camera behavior directly from data, then raised $75M (Mar 24, 2026) to keep training. Its consent posture is the interesting one — synthetic 'AI actors' that never existed sidestep likeness risk entirely, while digital twins stay a paid, opt-in feature. Watch: Research-lab ambition, consumer-app revenue: no disclosed ARR, and published trust/consent policy is the thinnest of the five — 'rights, safety, and brand control' language without Synthesia-grade specifics. Long-form corporate output isn't the wheelhouse; social formats are. [https://www.captions.ai](https://www.captions.ai) ### How to choose - If You're producing training, onboarding, or compliance video at enterprise scale — and procurement will ask about governance → Synthesia. ISO 42001, moderation-at-creation, and the EU Article 50 code signature (Jul 2026) are the paperwork nobody else has; 140%+ NRR says enterprises stay. - If You're a founder or marketer making explainers, launch videos, and localized content — realism and speed matter most → HeyGen. Avatar V is the current realism bar, $29/mo gets you a working clone, and the $200M-ARR flywheel funds the fastest model cadence in the category. - If The avatar needs to hold a live conversation inside your product → Tavus CVI (Anam is the latency challenger, D-ID Agents the budget route). Budget for per-minute conversational pricing, not per-video. - If You're shipping UGC-style ads or social clips and nobody's real likeness is needed → Mirage or Argil — fully generated actors carry no consent burden and read native to feeds. Purpose-built ad-actor tools (Arcads, Creatify) live in the ads element, not here. - If Your audience is trust-sensitive — investors, press, grieving customers → Film yourself. Audiences still detect avatars, and from Aug 2, 2026 the EU AI Act requires machine-readable AI labeling anyway (Art. 50). Use avatars for scale, never for sincerity. ### The field (20 more) Argil, Colossyan, AI Studios, Anam, Akool, Vidnoz, Hour One (fading), Elai.io (acquired), UneeQ, Soul Machines (fading), Hedra, Sync, Delphi, LivePortrait, HunyuanVideo-Avatar, Duix (Heygem), SadTalker (fading), VASA-1 (dead), Rephrase.ai (acquired), Movio (renamed) Full dossier data: https://elems.ai/e/av.json --- Source: [elems.ai](https://elems.ai/e/av.html) — the periodic table of the AI-led startup. Data: https://elems.ai/elements.json (CC BY 4.0, cite elems.ai).