What to Build With the $300 Vertex AI Credit
Practical project ideas for your $300 Google Cloud credit: text with Gemini, images, Veo video, Lyria music and TTS voice, each with a quick how-to.
You claimed the $300 credit, and now what? Vertex AI gives you Google's own models for text, images, video, music, voice and transcription from a single account. Here are practical projects for each one, with a short pointer on how to start. If you have not claimed the credit yet, begin with Google Cloud $300 Free Credit for Vertex AI: How to Claim It.
Before you start: two things to know
First, a model ID changes every few months, so do not copy one from a blog post. List what exists with gcloud ai model-garden models list --model-filter=gemini, and use veo, lyria or imagen as the filter for the others. Pick the newest version and confirm it with one tiny call before building on it. A listed model can still return an error if it is not enabled for your project, in which case skip it.
Second, the location differs. Text, voice, music and image models live in the global location, and Veo lives in us-central1. A text request goes to an endpoint of this shape:
https://aiplatform.googleapis.com/v1/projects/PROJECT_ID/locations/global/publishers/google/models/MODEL_ID:generateContent
Remember that the credit covers Google's own models on Vertex AI, not partner models such as Claude and not the Gemini API through AI Studio.
Text with Gemini
Gemini 4 is the most capable language model in the list, and Gemini 3.8 Flash is the fast, cheap one for everyday tasks.
- Document summarizer: feed it long reports or contracts and ask for the five points that matter.
- Email and reply drafter: paste an incoming message and get a draft answer in your tone.
- Code helper: explain an unfamiliar function, or ask for a test for code you wrote.
- Script writer: the first step of a video: topic in, narration script out.
How to start: use Flash for experiments, because it is the cheap option, and move to Gemini 4 only for tasks where quality matters. Send a generateContent request, or use a Python script that connects with your service account key.
Transcription with Gemini Transcribe
- Meeting notes: turn a recording into text, then ask Gemini for decisions and action items.
- Interview archive: transcribe a set of recordings so you can search them.
- Lecture notes: transcribe a class, then summarize it into study cards.
How to start: send a short recording first and check the text before processing a long one.
Images with Nano Banana 2.1
- Blog and social banners: describe the scene and the mood, get a banner image.
- Video thumbnails: generate a few variations and pick the clearest one.
- Frames for video: create a starting image that you later animate with Veo.
- Mock-ups: visualize a product idea before spending time on design.
How to start: write the prompt as a specific scene, not a topic. Who is in it, where, in what style. Generate several versions and compare.
Video with Veo 3.1
- Short promo clips: a few seconds of realistic footage from a text description.
- Background footage: a looping scene for a presentation or a website header.
- Animate a still: take an image from Nano Banana and bring it to life.
How to start: Veo works through a long-running request (predictLongRunning), so you submit the job and then wait for the result. It lives in the us-central1 location. If you want the output saved to a storage bucket, the setup skill has an option for that. Start with one short clip and a precise description before you try anything bigger.
Voice with Gemini 3.8 TTS
- Narrating your own articles: turn a blog post into audio.
- Voiceover for videos: natural-sounding narration for the script you wrote with Gemini.
- Audio for learners: read lesson material aloud, which helps people who prefer listening to reading.
How to start: take one paragraph, generate the audio, and listen before converting the whole text.
Music with Lyria 3 Pro
- Background music: original music for a video, matched to a description such as calm, warm, no lyrics.
- Intro and outro jingles: a short signature sound for a podcast or channel.
How to start: describe mood, instruments and pace in a sentence, and generate a couple of takes to compare.
Putting it together: one video from one account
- Gemini writes the script.
- Nano Banana makes the opening image or thumbnail.
- Veo shoots the footage.
- Gemini TTS narrates the script.
- Lyria composes the background music.
Every step runs on the same project and the same credit.
Keep an eye on the credit
Experiments are cheap until you run a big batch, and video generation is the place where it is easiest to overspend by repeating attempts. Check your balance on the credits page, keep budget alerts on, and remember that the credit ends after 90 days. The full list of precautions is in how to claim the $300 credit without surprise charges. And for the account setup itself, go back to the first post in this series.