Skip to content
New Universal Prompt & Editing Workflow: Nano Banana and Seedream 4.0 are now outperforming Leonardo & ChatGPT on visual quality and prompt adherence - outputs are more cinematic and immersive. Now being tested for long promos. [QC-less AutoAI prep : Character-canvas prompt rebuild to enable QC-less in-app show generation by November; content-filter relaxation requests with Google/Bytedance (ETA end-Oct) to allow non-NSFW action and intimate scenes.AutoAI Video: Standardizing on Veo3 for motion generation; improving Veo3 + Seedance prompt templates for smoother, higher-fidelity motion output. Training & Talent Pipeline: 10 film-school outreach drives yielded 80 attendees - 20 trained, 4-5 to onboard as interns;
- Google (Oct 17) & Bytedance post-Diwali workshops scheduled to co-train teams on model best practices.
New Universal Prompt + Editing Workflow - Nano Banana & Seedream 4.0 new’s features of editing images is helping us to generate higher quality images than what we have from Leonardo & ChatGPT. []New images are more cinematic and immersive. There is also much better prompt adherence. Character Canvas Prompt Research - Rebuilding a universal prompt template to generate unique, high quality characters for in-app shows. This is in preparation to launch AutoAI QC-less for multiple in-app shows in November. Filter removal with Google / Bytedance: We’ve raised requests to lighten the children and content policy violations filter for Pocket FM. It will still block NSFW, but allow us to generate images of action sequences, child characters, intimate scenes (no nudity), etc. - ETA is end-October.
AutoAI Video: Doubling down on Veo3 for automated motionVeo3 fast output Seedance output Next steps:
Indian Film School Workshops: Spoke with - Whistling Woods, ZIMA, Xavier’s Institute, etc. - and conducted a free-1 hour workshop for their recent graduates (freelancing in the industry and available to join immediately)~80 recent graduates attended, 40 showed interest in a 2-week workshop with us, 20 have attended our 2-week workshopWe’ll pick 4-5 top performers and onboard as interns - this is a good, robust pipeline for ACDs. Next steps: We’ll repeat this exercise in Dec-Jan to get their third year students to join us when they graduate in March 2026Bytedance & Google teams-led training: Google is doing 2 workshops with us in October, starting 17 October; Bytedance also after DiwaliWe’re using these sessions to get guidance directly from the engineers who have created Nano Banana / Veo3 / Seedance / Seedream on use cases where we struggle to get high quality images.
Seedance Video: Last Frame Return - Instantly reuse last frame of a motion video to extend motion beyond 10s, creating smooth 30–60s continuous scenes with perfect shot and scene continuity. [Seedance Video: First & Last Frame Upload - Generate seamless motion transitions between shots to maintain scene continuity - a fast, efficient alternative to Kling & Pika Labs. [Luma Labs - Useful for reframing (1:1 to 9:16) and style modifications. API being integrated to Playground by 21st Oct [Infinite Talk - Insane lip sync quality. Rolling out $49 plans to each team to add lip sync to their videos. [Veo 3.1 (Previews on X look amazing), Sora 2, Grok Imagine V9
15th October, 2025
Summary
]AutoAI Image Updates
Next steps:
AutoAI Video Updates
Training & Film School Workshop
New Tools
Additions
]
]
& ]
& ]Next up for testing
Want to print your doc?
This is not the way.
This is not the way.
Try clicking the ··· in the right corner or using a keyboard shortcut (
CtrlP
) instead.