|Docs

Turn a blog post URL into a video

Turn a blog post, article or newsletter issue into a video: blog to video, article to video, text to video from a URL. Your agent reads the post, writes a short script, and shows it to you first. Then it voices the script, lays stock b-roll and text slides under it, adds captions, and puts music under the voice that dips when the voice speaks.

Use invideo for this when you want the result on an editable timeline: you can swap one shot, fix one caption, or re-voice one line without starting over.

Use your own posts, or posts you have the right to adapt. The agent never publishes the video. You post it.

Quickest path

Set up your agent first with invideo for AI agents. Then give it this prompt.

Turn https://acme.com/blog/onboarding-tips into a 60-second video.Read the post, write a script in my voice, and show it to me before youspend anything. Then voice it, use stock b-roll and text slides for thekey points, add captions and soft music under the voice. 16:9 and 9:16.

In a chat app or a terminal, the prompt is the same. The agent reads the URL with its own browser or fetch tool.

What you need

These are the inputs the agent asks for.

  • The post's URL, or the text pasted into the chat if the post is behind a login.
  • The length: 30, 60 or 90 seconds.
  • A voice. The agent finds one with search_library and sends you the preview.
  • The shape: 16:9 for YouTube and your site, 9:16 for Reels, Shorts and TikTok.
  • Your logo and brand colors, if you want them on the slides.
  • A music mood, or no music.

Steps

The agent works on a branch, so the main timeline stays clean until you approve.

  1. Read the post

    invideo does not fetch web pages. The agent reads the URL with its own browser or fetch tool, outside invideo. Expected result: the post's title, headings and main points in the chat.

  2. Write the script and wait for your yes

    The agent writes a short script from the post: a hook, three or four key points, and a closing line. It marks which point each slide shows. It shows you the script before it spends anything. Expected result: your yes, or your changes.

  3. Voice the script

    The agent runs find_models with the job in words, picks one, and reads its inputs with get_model. A good pick is elevenlabs_text_to_speech.elevenlabs_tts_v4. It picks a voice from search_library in elevenlabs_text_to_speech_voices and sends you the preview. It calls generate with dry_run: true, the script as prompt and that voice_id, and tells you the price. Then it calls generate with max_credits set to that price and an idempotency_key. Expected result: a sequence URL with the full script as audio. The agent reads that URL with read and never calls generate again to check.

  4. Get the word times

    The agent prices audio.transcript on the voice's sequence URL with dry_run, starts it with max_credits set to that price, and calls again after retry_after_s for the words. Expected result: every word with its start and end time, which set the captions and the slide changes.

  5. Find b-roll and build the timeline

    The agent searches search_library in stock_video with words from each point, and brings the picks in with add_media and stock. It makes a timeline with new_timeline, creates a branch with timeline_history, places the voice with insert_media, and the shots with place_media. It adds text slides for the key points with add_text. Expected result: a new state_token and a shot or slide under every line.

  6. Add captions and music

    The agent adds the words with add_captions. It finds music in stock_music, places it on A2 (Music) with insert_media, and adds ducker to that track with add_audio_effect. A new timeline ducks under A1 (Voice) by default, so the ducker follows the voice with no more setup. The agent sets set_sidechain_source only when the voice sits on another track. Expected result: music that drops under the voice and comes back between lines.

  7. Check, then export

    The agent runs capture at each slide and check_timeline, and checks the voice and music with audio.loudness. Then it runs export with format video and target_lufs. Expected result: an export card with status completed.

The terminal version of the same job:

invideo project new --title "Onboarding tips video"invideo agent register <project-url> --name <name>export INVIDEO_AS=<name>invideo gen models <project-url> --query "narration voice from a script"invideo library search <project-url> \  --library elevenlabs_text_to_speech_voices --query "warm narrator"invideo gen model <project-url> \  --model elevenlabs_text_to_speech.elevenlabs_tts_v4invideo gen run <page-url> \  --model elevenlabs_text_to_speech.elevenlabs_tts_v4 \  --dry-run -i prompt=@script.txt -i voice_id=<voice-id>invideo gen run <page-url> \  --model elevenlabs_text_to_speech.elevenlabs_tts_v4 \  --max-credits <price> --idempotency-key onboarding-voice-v1 \  -i prompt=@script.txt -i voice_id=<voice-id>invideo media audio.transcript <sequence-url> --dry-runinvideo media audio.transcript <sequence-url> --max-credits <price>invideo library search <project-url> --library stock_video \  --query "team onboarding laptop"invideo library search <project-url> --library stock_music \  --query "soft upbeat background"invideo asset add <project-url> --stock <stock-id>invideo edit new <project-url> --name "Blog video 16x9"invideo edit show <edit-url> --jsoninvideo branch new <edit-url> --name blog-v1 --token <state_token>invideo edit show <branch-url> --jsoninvideo edit change <branch-url> --timeline T1 --token <state_token> \  --operations @build.jsoninvideo edit capture <branch-url> --at 2,15,30,55 --out ./framesinvideo edit check <branch-url>invideo edit export <branch-url> --format video --quality 1080 \  --target-lufs -14 --jsoninvideo status <card-url> --waitinvideo notebook show <card-url> --out ./out

Where the values come from:

  • <project-url> comes from invideo project new, and <edit-url> from invideo edit new. <name> is the name you choose in invideo agent register. <page-url> is a notebook page from invideo notebook show <project-url>.
  • <voice-id> comes from the voice search. Each <price> is the dry run right before it: .estimated_credits for the voice, credits for the transcript.
  • <sequence-url> is the voice's sequence URL, which the paid voice run returns.
  • <stock-id> is a result's id from invideo library search.
  • <state_token> comes from invideo edit show <edit-url>, and on the branch from invideo edit show <branch-url>; the main token is stale there. <branch-url> is the .branch_url that invideo branch new returns.
  • <card-url> is .sequences[0] of what the export returns. invideo status waits and does not download; invideo notebook show saves the file as ./out/export.mp4.

The file ./build.json holds a bare list of operations. On the MCP path, the same list goes in the operations field of edit_timeline. It adds a text slide and the first caption on V3 (Titles), and the ducker on A2 (Music). A1 is the voice track, and a new timeline ducks under A1 by default.

[  { "op": "add_text", "track": "V3", "start_f": 90, "end_f": 240,    "text": "Tip 1: Send the welcome email before day one",    "style": { "font_size": 56, "color": "#ffffff",      "align": "center" } },  { "op": "add_captions", "track": "V3",    "slides": [      { "text": "New hires decide in a week", "start_f": 0,        "end_f": 60 }    ] },  { "op": "add_audio_effect", "target": "A2",    "target_kind": "track", "effect": "ducker" }]

If the voice sits on another track, the agent adds set_sidechain_source to the same list, with timeline T1 and that track's id from get_timeline. It works in the same call as add_audio_effect.

For the 9:16 version, the agent creates a second branch, sets the size with set_timeline_settings, and fills the frame on each shot with fill_frame.

Cost and time

Edits and exports are free. Generation and transcripts cost credits. Paid stock costs a small credit charge, or one of your workspace's iStock licenses. Your agent runs a dry run first and tells you the price before it spends anything.

  • Free: reading the post, the script, every edit, the ducker, capture, export.
  • Costs credits: the TTS voice, audio.transcript on the voice, each stock shot and the stock music.

One video takes under an hour, most of it your review of the script. See Credits for prices.

If something goes wrong

Most problems are a script that drifts from the post, or timing.

SymptomFix
get_model or generate refuses the model as not foundModels retire. Run find_models with the job in words, use the best match, and tell the person which model you used.
Exit code 6, or status is heldA safety review, not a refusal. Wait, then run invideo status <url> --wait again. Do not resend. If it ends failed, change the input the review flagged.
The agent cannot open the URLThe page needs a login or blocks bots. Paste the text into the chat.
The script says something the post does notAsk the agent to quote the post line behind each claim, and cut what has none.
The music does not drop under the voiceThe ducker follows A1 by default. If the voice is on another track, point it there with set_sidechain_source.
A slide changes mid-sentenceMove the slide's edges to word times from the transcript with trim_clip.
timeout on generateDo not call it again. read the thread or call status.
CLI exit 3The edit changed since your read, or the token is from main. Run invideo edit show <branch-url> and use its state_token.
The ./out folder is empty after statusstatus waits and does not download. Run invideo notebook show <card-url> --out ./out.
CLI exit 5Credits ran out. Stop and tell the person.
Avoid: voicing the script before the person reads itInstead: show the script first. The voice costs credits.
Avoid: adapting someone else's articleInstead: use the person's own posts, or ask them to confirm the rights.

Prompts to try

Each prompt is a real variation of this job.

Turn my last three posts into three 30-second Shorts, 9:16, same voice,captions in the middle of the frame.
Make a video from this newsletter issue. No voice: text slides andcaptions only, with music. Pasted text below.
Turn this listicle into a 90-second video with one stock shot per tipand a title card for each number.

When to use something else

invideo does not read web pages, so the reading happens in the agent's own browser or fetch tool. invideo does not publish to your blog, YouTube or social accounts. For a talking presenter instead of b-roll, see the avatar style in the video sales letter page. For a whiteboard drawing style, see the whiteboard explainer page.