The LensCowboy Pipeline is an AI content generation platform that automates image, video, music, and 3D generation. Each subscriber gets a Google Sheet that acts as the control surface — you configure what to generate, and the pipeline runs it automatically.
The pipeline supports multiple generation modes (image-only, image-to-video, text-to-video, video restyle, music-only, and more) and works across leading AI image, video, music, and post-processing models.
- Google Sheet — configure prompts, vendors, modes, settings across several tabs.
- Pipeline — reads your sheet, resolves prompts, selects vendors, dispatches jobs.
- AI Vendors — each job routes to the appropriate vendor based on your Platform/Vendor/Model cascade.
- Google Drive — generated media uploads to your project folder, organised by date.
Runs can be triggered manually from the sheet menu (Pipeline → Run Now), from the Settings page, or automatically on a schedule.
| Tab | Purpose |
|---|---|
| Daily | One row per day. Controls generation mode, vendor selection, aspect ratio, duration, resolution, feature toggles (enhancer, upscale, color grade). |
| Weekly Direction | Weekly creative direction — subject, style, environment, palette. Referenced by prompt templates via {Subject}, {Style}, etc. |
| Prompt | Platform-specific prompt templates. Each row is a platform entry with image/video/music/restyle prompt columns in Template, Custom, or JSON format. |
| Post | Key-value settings for post-processing: watermark, model name overlay, Drive folder IDs, music volume, LUT settings. |
| PAL | PAL generation tab — one row per shot. Video restyle, frame restyle, I2V from PAL previz exports. |
| 2D | 2D editing pipeline — chained operations (BG removal, upscale, restyle, composite) with dependency resolution. |
| Log | Append-only audit log of every successful run. PAL AI Render results also write here. |
Set the Generation Mode column on the Daily tab.
| Mode | What it does |
|---|---|
| Image to Video | Generate an image, then generate a video from that image. |
| Text to Video | Generate a video directly from the prompt (no image step). |
| Image Only | Generate or grade an image. No video, no music. |
| Video Restyle | Restyle an existing video (from Source Video column). |
| Music Only | Generate music or sound effects only. |
| Recipe | Gemini multimodal — up to 3 reference images for guided generation. |
| Extract Prompt | Vision LLM extracts a generation prompt from an existing image. |
| LUT Extract Only | Extract a .cube LUT from a reference image for colour grading. |
| PAL | Reads the PAL tab, dispatches to video restyle vendors (3-pass workflow). |
- Template — Prompt tab's template column. Variables like
{Subject},{Style},{Environment}resolve from Weekly Direction. - Custom — Prompt tab's custom column. Free-form text, still supports
{Variable}substitution. - JSON — Prompt tab's JSON column. Structured format for complex multi-field prompts. Use Transform menu items to convert JSON to resolved text.
Image and video prompts resolve independently — you can use Template for images and JSON for video on the same run.
Enable the prompt enhancer to rewrite resolved prompts through an LLM before sending them to the generation vendor. Set Prompt Enhancer on the Daily tab to Gemini, Claude, or ChatGPT.
Then enable per-prompt with Enhance Image Prompt and/or Enhance Video Prompt columns.
The enhancer adds cinematic depth, atmosphere, and visual detail while preserving the original prompt's specifics. In Test Run, both the original and enhanced prompts display side-by-side so you can verify before committing to a real run.
- Platform — the provider routing to the chosen model (selectable from the dropdown on each row)
- Vendor — the model provider (e.g. ByteDance, Lightricks, Alibaba)
- Model — the specific model name
The same model may be available on multiple platforms — each with its own API key and pricing.
See Vendor Costs for live pricing and Model Endpoints for the full catalogue.
Run Pipeline → Install Dropdowns from the sheet menu to install cascading validations on the Daily, 2D, and PAL tabs. Select Platform → Vendor dropdown updates → Model dropdown updates.
Dropdown data is fetched from the server and cached in the sheet. When new models are added, re-run Install Dropdowns.
Run Pipeline → Test Run to preview without calling any vendor APIs (except the prompt enhancer, which makes a real LLM call so you can see the rewrite).
The sidebar shows:
- Config — mode, vendors, aspect ratio, duration, resolution
- Vendor Status — which vendors are activated and have keys
- Resolved Prompts — exact prompts that would be sent, including enhancement diff (original vs enhanced)
- Cost Estimate — from the model registry
- Warnings — missing keys, empty prompts, configuration issues (e.g. Enhance on but Prompt Enhancer empty)
- Sheet menu — Pipeline → Run Now. Runs today's row immediately.
- Settings page — RUN NOW button with slot selector. Shows live progress steps.
- Daily scheduling — automatic daily runs at your configured time.
During a run, the Settings page Runs tab shows real-time progress with step-by-step updates. You can cancel a running job from the Runs tab.
Access at app.lenscowboy.com/settings. The settings page manages:
- Project Slots — connect Google Sheets and Drive folders. Each slot is a separate project.
- Vendor Activation — toggle vendors on/off and enter API keys. Keys are stored encrypted per-tenant.
- Run History — past runs with status, cost, error details.
- RUN NOW — trigger immediate runs with slot and mode selection.
- LUT Extract — extract a .cube LUT from a reference image using the
LUT Extract Onlymode. - LUT Application — apply .cube LUTs to generated images and videos.
- 16-bit output — supported for HDR workflows. Watermark is skipped on 16-bit output to preserve bit depth.
Music runs alongside video (skipped entirely in Image Only mode). Set the Music Mode column:
generate— generate music from a promptsource— use a source music file from your Drive folderdisabled— no music
Music, sound effects, and text-to-speech generation are all supported — pick the model and version from the dropdown on each row.
The 2D tab supports chained image and video operations with automatic dependency resolution:
- Image — BG Removal, Object Removal, Roto Mask, Inpaint, Outpaint, Restyle, Upscale, Composite, Apply LUT, Reformat
- Video — Video BG Removal, Image to Video
- Generation — Image Only, Image to Video (same vendor cascade as Daily)
Each row can reference another row's output as its input: generate → edit → upscale → generate video.
PAL mode runs from the PAL tab when the Daily tab's Generation Mode is set to PAL. 3-pass workflow:
- Pass 1 — Video Restyle: restyle the PAL previz video. Output saved to
Output Video URL. - Pass 2 — Frame Restyle: extract the first frame from the restyle output and restyle it. Output saved to
Output Restyle Frame URL. - Pass 3 — I2V: use the restyled frame as the starting frame for image-to-video. Output saved to
Output Restyle Video URL.
Passes 2 and 3 are optional, controlled by Restyle First Frame and Restyle Video Model columns. Restyle prompt resolves from the Prompt tab per Restyle Prompt Mode.
Test Run shows estimated per-run cost based on vendor, resolution, and duration. Live pricing for every model is on the Vendor Costs page.
Test Run shows warnings
Common causes: missing API keys (activate the vendor on Settings), empty prompts (Prompt tab), vendor not activated, or Enhance on but Prompt Enhancer blank.
Run fails with 402 / billing error
Vendor API needs credits. The error includes a link to the vendor's billing dashboard. Top up and re-run.
Video not appearing in Drive
Check your project slot has a valid Drive folder ID on Settings. If fields are empty, click Create Project to scaffold the folder structure.
Dropdowns not showing new models
Re-run Pipeline → Install Dropdowns. Fetches the latest catalogue from the server.