Comment créer des vidéos IA à partir de texte

Texte vers vidéo fonctionne mieux avec des prompts qui spécifient sujet + mouvement + caméra + éclairage + style. Commencez par un prompt court, gardez la durée courte et itérez.

Étapes rapides

Une liste rapide que vous pouvez suivre en moins d'une minute.

Ouvrir l'outil
  1. 1
    Ouvrez l'outil Texte vers Vidéo.
  2. 2
    Écrivez un prompt : sujet + action + mouvement de caméra + éclairage + style.
  3. 3
    Choisissez le ratio d'aspect (9:16 pour Shorts, 16:9 pour YouTube).
  4. 4
    Définissez une durée courte pour un mouvement plus propre (commencez par 5s).
  5. 5
    Générez et affinez le prompt (changez une chose à la fois).
Tool options you can choose
Aspect ratio:16:99:164:33:41:121:99:21
Duration:2s–12s4s / 6s / 8s (Audio mode)
Resolution:480p720p1080p
  • Options vary by version: Silent vs Audio.

Exemples du tutoriel (avec prompts et paramètres)

Chaque exemple ci-dessous est présélectionné pour ce guide (non aléatoire).

Example 1

Cinematic neon rain (prompt → video)

Text-to-Videotext-to-video
How to use this example
  1. 1.Open the tool.
  2. 2.Choose mode: Text-to-Video.
  3. 3.No input file needed (prompt-only).
  4. 4.Use the keywords (or full prompt) and pick settings.
  5. 5.Generate and iterate (crop/lighting/prompt) if needed.
Prompt keywords
A cinematic realisma city alleyway in the night rainThe camera begins with a close-up of a puddle: raindrops create concentric ripplesThe camera slowly rises and slides forward (a low-angle push-up)A person in a dark trench coat walks past with their back to the camerathe edge of the coat is lifted by the windrevealing the fabric's texture clearlyTheir footsteps step into puddles
Show full prompt (copy)
A cinematic realism, a city alleyway in the night rain. The camera begins with a close-up of a puddle: raindrops create concentric ripples, and the red and blue reflections of neon signs are slightly distorted on the water's surface. The camera slowly rises and slides forward (a low-angle push-up), water splashing onto the front edge of the lens to create realistic wet halos and tiny droplets. A person in a dark trench coat walks past with their back to the camera; the edge of the coat is lifted by the wind, revealing the fabric's texture clearly. Their footsteps step into puddles, and the splashing water droplets shimmer like tiny silver fragments in the backlight of the streetlights. The background features slightly bokeh of neon lights and mist, with fine rain and fog floating in the air, and volumetric light beams passing through the alleyway to create visible pillars of light. The camera finally focuses on the figure's half-body silhouette from behind, neon light refracting colorful highlights on the raindrops—a tense, restrained, and cinematic atmosphere.

Cinematography: 35mm cinema lens, shallow depth of field, realistic film texture, subtle grain and high dynamic range (HDR), high contrast while preserving shadow detail. Lighting: Neon main light + streetlight backlight, enhanced reflection from wet ground, highlights not overexposed. Movement: Stabilizer slider with slight handheld breathing feel. Color: Cool cyan base tone, accented with red/blue neon.
Settings (used in this example)
Open tool
Example 2

Futuristic rooftop (prompt → video)

Text-to-Videotext-to-video
How to use this example
  1. 1.Open the tool.
  2. 2.Choose mode: Text-to-Video.
  3. 3.No input file needed (prompt-only).
  4. 4.Use the keywords (or full prompt) and pick settings.
  5. 5.Generate and iterate (crop/lighting/prompt) if needed.
Prompt keywords
cinematic blockbusterphotorealNighttime rooftop in a futuristic mega-city during light rainOpen on an extreme close-up of a gloved hand tightening a metal clasp—micro scratcheswet reflectionscrisp detailA distant siren glow pulses across the gloveNeon signs and traffic streams below form anamorphic bokeh
Show full prompt (copy)
cinematic blockbuster, photoreal. Nighttime rooftop in a futuristic mega-city during light rain. Open on an extreme close-up of a gloved hand tightening a metal clasp—micro scratches, wet reflections, crisp detail. A distant siren glow pulses across the glove. Camera whips upward into a smooth stabilized dolly push as the character steps to the edge of the roof. Neon signs and traffic streams below form anamorphic bokeh; rain streaks catch volumetric light beams from a hovering drone. The character turns slightly—only a silhouette—cloak fluttering in the wind, edges rim-lit by cyan and magenta neon. A massive holographic billboard flickers behind them, casting shifting light across drifting mist. End on a heroic wide shot: the figure stands against the skyline as a drone passes close to camera, lens flare blooms, and the city breathes with depth and scale. High dynamic range, rich contrast, fine film grain, premium color grade (cool shadows, warm highlights), smooth motion, no jitter.
Settings (used in this example)
Open tool

Conseils

  • Les mots de mouvement (zoom avant, panoramique, dolly lent) améliorent souvent la sensation de la vidéo.
  • Si la scène tremble, simplifiez le mouvement de caméra et raccourcissez le prompt.
  • Gardez un sujet principal pour réduire la dérive.

FAQ

Pourquoi ma vidéo semble-t-elle aléatoire ou incohérente ?
Les prompts peuvent être trop longs ou conflictuels. Supprimez les détails supplémentaires et gardez un seul sujet et action.
Dois-je utiliser des mots-clés de caméra ?
Oui—les mots-clés de caméra améliorent souvent la qualité cinématographique (gros plan, 35mm, faible profondeur de champ).
Quelle durée est la meilleure ?
Commencez par 5 secondes. Les clips courts sont plus propres et plus faciles à itérer.

Prêt à générer ?

Ouvrez l'outil et réutilisez les prompts/paramètres ci-dessus.

Comment créer des vidéos IA à partir de texte (Texte vers Vidéo)