The production pipeline for the vertical animated series 'Lesnaya Ulitsa' utilizes cloud-based AI models to generate 1080x1920 resolution episodes lasting 23–26 seconds. The workflow involves creating a script, generating character reference sheets, producing images via GPT Image or Nano Banana Pro, and animating scenes using models like Kling 4.0.
The team tested various models including Wan, Kling, Veo, and Seedance between October 29 and October 31, 2026. Kling 4.0 was selected for animation due to its performance, costing 54 Higgsfield credits per scene. The process prioritizes consistency by using fixed character reference sheets and manual human oversight at each stage of production.
Audio production relies on ElevenLabs eleven_v4 for voice synthesis, with a custom casting process selecting specific voices for characters, such as 'Babushka Valya' for the character Zina. The team uses Voice Changer eleven_multilingual_sts_v2 to ensure character voice consistency, as native AI-generated voices often do not match established character profiles.
Technical quality control includes checking for visual artifacts, lip-sync accuracy, and audio levels at approximately -14 LUFS. The team uses Scribe for speech recognition to verify that dialogue matches the script. If a scene fails the automated or human review, it is discarded or re-generated based on specific error logs.
