D-ID (Expressive V4)
Enterprise AI agents with expressive avatars and native RAG
Latence TTFR
~450ms
temps réel
Coût / minute
à partir de $3.000/min
selon l'abonnement souscrit
Qualité visuelle
8/10
score estimé
Protocoles
WebRTC, REST
Direction émotionnelle & mise en scène
V4 Expressive expose des sentiments et une voix configurable, mais la preuve officielle consultée doit être séparée entre la vidéo async et l’agent streaming.
Pilotage API expliciteParamètres et mécanisme
`sentiment_id`, `sentiment`, `voice_config`, `stability`, `style` et gestion de queue des speaks.
Limite à garder visible
Valider avec D-ID la disponibilité exacte de chaque paramètre dans le mode temps réel ciblé avant de le classer candidat Où est Ava ?
Personnalisation de l'avatar
RAG / Base de connaissances
Native 'Knowledge' API: connect custom knowledge bases (PDF, text). Agent queries them in real-time during conversation.
Comportement & personnalité
System prompt configuration for personality, tone, response style. Full agent persona definition.
Langage corporel & gestes
V4 Expressive: adaptive body language generated to match voice inflection and emotional context. Natural posture.
Expressions faciales
V4 Expressive: emotion and micro-expression control via sentiment tags (joy, seriousness, surprise).
Voix & clonage vocal
ElevenLabs + Microsoft Azure TTS partnerships. Hundreds of multilingual voices. Pitch, speed, accent control. Voice cloning available.
Fine-tuning du persona
Full agent persona: system prompt + knowledge base + voice + visual identity. Consistent role maintenance throughout interaction.
Entraînement de l'avatar
Bonnes pratiques
- 01.Frontal uniform lighting, no shadows
- 02.Neutral fixed background
- 03.Stable camera at eye level
- 04.Natural speech with pauses
- 05.Avoid excessive head movements
- 06.Clean audio recording (no background noise)
- 07.No face occlusions (sunglasses, hands)
Analyse API
Protocoles
SDKs
Sessions simultanées
Plan-dependent
Limites de débit
BYO-S3 available on Pro+ plans
Fonctionnalités clés
- Knowledge API for native RAG
- Agents API for full conversational agent management
- Expressive V4 emotion tags
- ElevenLabs + Azure TTS integration
- WebRTC real-time streaming
Contraintes API
- Expressive V4 requires specific credits (not all stock avatars)
- No on-premise standard option
- Data stored on D-ID infrastructure by default (BYO-S3 on Pro+)
- Celebrity face moderation
Modèle tarifaire
| Plan | Prix | Minutes incluses | Dépassement |
|---|---|---|---|
| Trial | Free (14 days) | Limited | N/A |
| Build | $18/mo | 32 min streaming | ~$0.56/min |
| Launch | $40/mo | 90 min streaming | ~$0.44/min |
| Scale | $158/mo | 400 min streaming | ~$0.40/min |
| Enterprise | Custom | Custom | Custom |
Coûts cachés / à surveiller
- Expressive V4 credits premium
- BYO-S3 only on Pro+
- 20–30% discount on annual commitment
Souveraineté & Hébergement
Score de souveraineté
Hébergement
AWS US-East-1
GDPR
YesOn-premise
NoDétail souveraineté
Israeli company. AWS US-East-1. No EU hosting. GDPR principles respected but no EU datacenter.
Contraintes & Limites
- No manual hand/arm gesture control
- US hosting only (AWS US-East-1)
- Expressive V4 not available on all stock avatars
- Celebrity face moderation
- No on-premise standard option
Pertinence pour GamiWays
Score
7/10
Strong native RAG and expressive emotion control make D-ID compelling for educational agents. V4 Expressive is a significant upgrade. Main limitation: US-only hosting and higher cost per minute vs. Simli/bitHuman.