Dumpling Does Nothing: a five-episode AI pet comedy, and what it really cost
We made an AI-generated capybara comedy series with AutoShort: credits per episode, the retries, and the weak view counts, each with its read date. Run 2026-10-23; numbers read up to 2026-10-28.
Format: Animal comedy, five planned episodes, narrated, under a minute each, vertical 9:16. Everything shown in the videos is AI-generated. Numbers are from our own accounts and are small; each carries the date it was read, because platform numbers change as a video ages. This page makes no promise of views.
What we made
Dumpling Does Nothing is an animated comedy about a capybara whose only goal is to do nothing, and a coconut that keeps interrupting. Every image, clip and voice is AI-generated, and every post we made carries the platform AI label.
We planned a five-episode season, made one character sheet for Dumpling so the same animal appears in every shot, confirmed the season plan (it shows a price per episode before anything is charged), then rendered episode by episode. Each episode got a free quote first, and the create call had to repeat the quoted credit amount, so there were no surprise charges.
Honest scope note: these episodes were created through AutoShort's series API, not typed into an MCP chat. So this page shows what a series costs and what it returns, not an MCP transcript. The MCP series tools are the lighter route to the same idea: creating a series is free and each scheduled video is charged when it renders. The prompts linked at the bottom start there.
Credits and cost, episode by episode
| What | Number | As of | Source file : line |
|---|---|---|---|
| Character sheet for Dumpling (5 images) | 10 credits | run 2026-10-23 | qa-dumpling-ep1.md:9 |
| Season plan price, confirmed before any episode was charged (5 episodes, 480p) | 1,544 credits quoted, 0 charged at confirm | run 2026-10-23 | qa-dumpling-ep1.md:13 |
| Episode 1, rendered once | 304 credits, no refund | run 2026-10-23 | qa-dumpling-ep1.md:68 |
| Episode 2, first render: two shots fell back to still pictures | 328 charged, 128 refunded, net 200 credits | run 2026-10-23 | qa-dumpling-ep2.md:13 |
| Episode 2, re-render: all five shots animated | 328 credits, no refund | run 2026-10-23 | qa-dumpling-ep2-v2.md:11 |
| Episode 3 | 304 credits, no refund | run 2026-10-26 | qa-dumpling-ep3.md:15 |
| Episode 4, first render: failed our QA (a second animal appeared), not posted | 304 credits, no refund | run 2026-10-27 | qa-dumpling-ep4.md:16 |
| Episode 4, retry with one wording change in the plan | 304 credits, no refund | run 2026-10-27 | posts-2026-10-27-ep4b.md:13 |
| Episode 5 | 304 credits, no refund | run 2026-10-27 | qa-dumpling-ep5.md:14 |
| Provider cost we measured for episode 1 (clips 0.400 USD, plus stills and sheet) | about 0.43 to 0.49 USD for the episode, about 0.45 to 0.51 USD with the sheet | run 2026-10-23 | qa-dumpling-ep1.md:69 |
| Provider cost we measured for episode 3 | 0.4003 USD (clips) | run 2026-10-26 | qa-dumpling-ep3.md:17 |
| Provider cost for the two episode 4 attempts together | about 0.9 USD | run 2026-10-27 | posts-2026-10-27-ep4b.md:23 |
Total credits (sheet 10 + ep1 304 + ep2 200 + ep2 re-render 328 + ep3 304 + ep4 304 + ep4 retry 304 + ep5 304): 2,058 credits, which is $41.16 at the $0.02-per-credit list rate. Plan prices differ; see what things cost.
Credits are what the product charged. The USD figures are our own provider cost, measured as the change in a spend counter around each render; they are not an invoice and not the price a customer pays. The total is a sum of the rows above (7 renders for 5 episodes), priced at the list rate below.
What the platforms showed
| What | Number | As of | Source file : line |
|---|---|---|---|
| YouTube, episode 1 (channel @autoshort-org): views | 5 views, average view 0:08 of 0:35 | 2026-10-24 11:15 ICT, about 13.8 h after posting | metrics-ep1-2026-10-24.md:8 |
| Instagram, episode 1: views | 153 views, 98.7% from non-followers, 2 likes, 0 follows | 2026-10-25 about 10:20 ICT, about 37 h after posting | metrics-ep1-2026-10-25-48h.md:11 |
| TikTok, episode 1: views | 0 views | 2026-10-25 about 10:20 ICT, about 37 h after posting | metrics-ep1-2026-10-25-48h.md:12 |
| Facebook Page, episode 1: views | 1 view, watched 1 s | 2026-10-25 about 10:20 ICT, about 23 h after posting | metrics-ep1-2026-10-25-48h.md:13 |
| YouTube, episode 2 (new channel @dumplingdoesnothing): views | 0 views, no traffic or retention data | 2026-10-26 08:20-08:45 ICT, about 12.3 h after posting | measurements-2026-10-26-evening-posts.md:10 |
| Instagram, episode 2: views on the reels grid | 128 views | 2026-10-28, read 06:40-09:40 ICT | measurements-2026-10-28-autoshort.md:23 |
| Instagram, episode 3: views on the reels grid | 143 views | 2026-10-28, read 06:40-09:40 ICT | measurements-2026-10-28-autoshort.md:23 |
| Facebook Page, episode 3 | reach 180, 213 views, average watch 3 s | 2026-10-28, read 06:40-09:40 ICT | measurements-2026-10-28-autoshort.md:26 |
| Series channel @dumplingdoesnothing, clips and episodes posted 26-27 Oct | 23, 0, 4, 2, 1 and 3 views; the channel median was 5 | 2026-10-28, read 06:40-09:40 ICT | measurements-2026-10-28-autoshort.md:17 |
| Clicks on our tracked links, all social posts of those days | 80 clicks, 0 signups, 0 paid | 2026-10-28 14:05 ICT | measurements-2026-10-28-autoshort.md:38 |
Views include our own: reading analytics and verifying posts adds plays that were not subtracted, and one YouTube read showed one person looping a Short. Numbers are small, so treat them as signal, not proof. Platform numbers change as videos age; the "as of" column is the read time.
What we learned
- Cost was predictable. The quote matched the season price table every time, and a failed clip is refunded automatically; a failed QA is not, because the render did complete. That is a full second render.
- When the provider fails on a shot, the product substitutes a still picture and refunds those credits. Decide whether a still is acceptable before paying for a re-render.
- Distribution, not the video, was the bottleneck: a new channel got no test audience at all, while Instagram gave strangers views in the usual range for that account. We could not separate channel age from the creative with one pair.
- A 35-second episode averaged 0:08 on YouTube, so most viewers left in the first seconds. Nothing about the series produced a signup in the days we measured.
What we would do differently
- Put the funniest frame and the question in the first second, and start with an 8 to 12 second cold open before releasing a 35-second episode.
- Use one channel for the whole series instead of splitting the audience across two.
- Write "one single capybara alone" into any shot where a second animal appeared; that one wording change fixed the extra animal on retry.
- Run the same episode with two different openings and compare, instead of judging a series by one post.
The public videos
Links as recorded in our publishing ledger. Platforms can remove or restrict a video later.
Try the same kind of prompt
Try the same idea with a prompt you can read before it spends anything. Drafts and cost estimates are free; a render is charged only after you approve it.
Pet comedy prompts and recipeConnect AutoShort to Claude, ChatGPT, Cursor or Gemini CLIFree script preview
A prompt does not make a video perform. It makes the script and the cost visible before you spend credits. Results of ours are above.
Data sources
Files in our internal research folders that every number on this page comes from:
- growth-ops/research/animal-comedy-series-2026-10-23/qa-dumpling-ep1.md, qa-dumpling-ep2.md, qa-dumpling-ep2-v2.md, qa-dumpling-ep3.md, qa-dumpling-ep4.md, qa-dumpling-ep5.md
- growth-ops/research/animal-comedy-series-2026-10-23/posts-2026-10-27-ep4b.md
- growth-ops/research/animal-comedy-series-2026-10-23/metrics-ep1-2026-10-24.md, metrics-ep1-2026-10-25-48h.md, measurements-2026-10-26-evening-posts.md
- growth-ops/research/measurements-2026-10-28-autoshort.md
- growth-ops/ledger/published.md (public links)