A ChatGPT Video Generator With Seedance, Kling and Veo
Add Revana to ChatGPT once and ask for a video in plain words. ChatGPT quotes the price, waits for your yes and hands you the download link a few minutes later. Clips are billed per second from prepaid Revana credit, with no fee for the connection, and a job that fails costs $0.
Add Revana to ChatGPT in four steps
- Add a server. In ChatGPT open Plugins (chatgpt.com/plugins), press Add, then choose Add custom MCP server. Name it Revana and paste
https://revana.si/api/mcpas the server URL. Keep Authentication on OAuth, tick the box that says you understand, and press Create. - Approve the connection. ChatGPT sends you to revana.si to sign in or create an account. Check what the connection may do and press Approve. ChatGPT gets its own Revana key, named after the app; you never copy a key.
- No Add option? Turn on Developer mode under Settings, Apps, Advanced settings, then repeat the first step. In a chat, pick Revana in the composer so ChatGPT can use it.
- Revoke any time. Revoke the ChatGPT key in your Revana dashboard and the connection stops at once; your other keys keep working. The connection reaches only your own Revana account.
Every other app (Claude, Cursor, VS Code, Codex) is on the MCP set-up page.
What to ask for
A short scene from text
Make a 5 second 720p video of a paper boat sailing down a rainy street at night, and tell me the price first.
ChatGPT quotes it, waits for your yes, then makes it. A 10 second Seedance 2.5 clip at 720p comes to $2.10.
Drafts first, final second
Make three 4 second 480p drafts of this prompt with different camera moves, then the best one at 1080p.
Low-resolution drafts keep testing cheap; only the final render costs real money.
Compare the models
Quote the same 10 second 1080p shot on Seedance 2.5, Kling 3.0 and Veo 3.1 Fast, then make the one that costs least.
ChatGPT gets three free quotes and runs one job.
A three-shot ad
Write a 20 second product ad as three shots, make each one, and give me the links.
ChatGPT writes the shot list and renders each clip, quoted before it runs. You get three files to cut together.
Video models you can call from a chat
| モデル | 設定 | 1秒あたり | 備考 |
|---|---|---|---|
| Seedance 2.5 | 480p | $0.09 | Default when you name no model |
| Seedance 2.5 | 720p | $0.21 | Text, a start picture or reference clips; sound can come with the video |
| Seedance 2.5 | 1080p | $0.52 | |
| Veo 3.1 Fast | 720p・音声付き | $0.10 | Google's Veo |
| Kling 3.0 | 1080p・音声付き | $0.17 | Kuaishou's Kling |
| Kling 4.0 | 720p | $0.135 | Clips up to 30 seconds |
| Wan 3.0 | 720p | $0.075 | Low-cost 720p |
| Grok Imagine Video 1.5 | 480p | $0.04 | Rough drafts and quick ideas |
Prices on Revana as of November 2026, the same per-second prices as the Revana API with no extra fee and no seat price. Name a model in your request and ChatGPT passes it on; every model and setting is on the models page, and the AI video cost calculator prices a whole batch.
You approve every dollar
Quote first, then your yes
ChatGPT is told to get a quote and ask before it spends anything. The charge is the quoted price, and an optional cap (up to $100 a job) means it never goes higher. Retrying with the same job id never makes or charges a second clip.
Prepaid credit, no subscription
You top up from $10 in the Revana dashboard and spend at each model's listed price. Credit never expires, and one balance covers images, video, music and voice. Buying credit from the chat stays off unless you allow it when you approve the connection.
A failed clip costs $0
If a route fails, Revana moves the job to another route for that model where it has one. If nothing delivers, the hold is released and you pay nothing, and a prompt the model refuses is not charged either.
The full rules are on the pricing page.
Animate a photo you already have
Paste a public link
In ChatGPT, pictures go in as public https links. Put the photo somewhere public, paste the link and ask for an 8 second vertical clip with a slow push-in; ChatGPT sends it to the video model as the first frame.
Keep a product or character consistent
Add reference pictures of a product or character and ask for several shots. Models that take references, such as Seedance 2.5, keep the look steady across clips. Make one, check it, then make the rest.
The same request under the hood
Every Revana model takes the same fields, so swapping Kling for Veo is a one-word change, and if you later move the flow into code the API request looks the same.
How long a clip takes
Most videos take a few minutes, and long or 1080p clips take longer; ChatGPT waits on Revana's side instead of asking again and again. The job keeps running even if you close the chat: in a new chat, ask ChatGPT to check your last Revana video, or download it from your dashboard. A download link works for about an hour, and you can ask for a fresh one whenever you need it.
FAQ
Can ChatGPT make videos?
Not with these models on its own. Connect Revana as a custom MCP server and ChatGPT can make clips with Seedance 2.5, Kling, Veo and other video models: it quotes the price, asks you, then returns a download link.
How do I add Revana to ChatGPT?
Open chatgpt.com/plugins, press Add, choose Add custom MCP server, name it Revana and paste https://revana.si/api/mcp as the server URL with OAuth, then approve the connection on revana.si. If there is no Add custom MCP server option, turn on Developer mode under Settings, Apps, Advanced settings.
How much does a ChatGPT video cost with Revana?
It is billed per second at each model's Revana price, with no fee for the connection. In November 2026 Seedance 2.5 is $0.21 a second at 720p, so a 10 second clip is $2.10; Grok Imagine Video 1.5 is $0.04 a second at 480p.
How long does a video take?
Most clips take a few minutes, and long or 1080p clips take longer. The job keeps running on Revana even if you close the chat; ask ChatGPT to check your last Revana video, or find it in your Revana dashboard.
Can I animate my own photo?
Yes. In ChatGPT, pictures go in as public https links: put the photo somewhere public, paste the link and ask for a clip. ChatGPT sends it to the video model as the first frame.
Is Revana made by OpenAI?
No. Revana is a separate company and is not affiliated with OpenAI. It runs an MCP server that ChatGPT can connect to, and model names belong to their makers.
What if a clip fails?
You pay nothing. Revana moves a failed job to another route for that model where it has one, and if nothing delivers the hold is released. A prompt the model refuses is not charged either.
API キー 1 つから始める
Revana のすべての言語・画像・動画・音声モデルに使えるキーが 1 つ。プリペイドクレジット、サブスクリプションなし、失敗したリクエストは無料です。